[LIVRO] Challenges and Opportunities in Agrometeorology - Farmacognosia (2025)

Prévia do material em texto

Challenges and Opportunities in Agrometeorology.S.D. Attri l L.S. Rathore l M.V.K. Sivakumar lS.K. DashEditorsChallenges and Opportunitiesin AgrometeorologyEditorsDr. S.D. AttriIndia Meteorological DepartmentMinistry of Earth SciencesLodi RoadNew Delhi-110003Indiasdattri@gmail.comDr. M.V.K. SivakumarWorld MeteorologicalOrganization Climateand Water Programme ClimatePrediction and Adaptation Branch7bis, Avenue de la Paix1211 GenevaSwitzerlandmsivakumar@wmo.intDr. L.S. RathoreIndia Meteorological DepartmentMinistry of Earth SciencesLodi RoadNew Delhi-110003Indialsrathore@ncmrwf.gov.inProf. S.K. DashCentre for Atmospheric SciencesIndian Institute of Technology DelhiNew Delhi-110016Indiaskdash@cas.iitd.ac.inISBN 978-3-642-19359-0 e-ISBN 978-3-642-19360-6DOI 10.1007/978-3-642-19360-6Springer Heidelberg Dordrecht London New YorkLibrary of Congress Control Number: 2011934440# Springer-Verlag Berlin Heidelberg 2011This work is subject to copyright. All rights are reserved, whether the whole or part of the material isconcerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publicationor parts thereof is permitted only under the provisions of the German Copyright Law of September 9,1965, in its current version, and permission for use must always be obtained from Springer. Violationsare liable to prosecution under the German Copyright Law.The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,even in the absence of a specific statement, that such names are exempt from the relevant protectivelaws and regulations and therefore free for general use.Cover design: deblikPrinted on acid-free paperSpringer is part of Springer Science+Business Media (www.springer.com)v.PrefaceThe global food security and sustainable agriculture are the key challenges beforethe scientific community in the present era of enhanced climate variability, rapidlyrising population and dwindling resources. Agriculture is intimately tied to weatherand climate influencing every aspect from long term planning to tactical decisionsin day-to-day management operations. Agrometeorology has a vital role to play inincreasing agricultural production in a sustainable manner using state-of-art tech-nology and resources efficiently. It is the responsibility of the meteorologists toadvise the farming community well in advance to take full advantage of benevolentweather and precautions against malevolent weather to minimize losses. Uncertain-ties of weather and climate pose a major threat to food security of the world ingeneral and developing countries, in particular. Asia in recent years has madeconsiderable progress in the field of agriculture. However, in order to keep pacewith the increasing population, the growth in agricultural production should besustainable. The problem, therefore, has to be addressed collectively by scientists,planners and the society as a whole.In view of need for increasing agricultural productivity to meet the demand ofrapidly growing population and coping with enhanced uncertainties and risks inagriculture, Agrometeorology is facing lot of challenges as well as opportunities forachieving the path of sustainability. Indian Meteorological Society in associationwith World Meteorological Organization, India Meteorological Department, Min-istry of Earth Sciences, Department of Science and Technology and Department ofSpace, Government of India organized and International Conference (INTROMET2009) on “Challenges and Opportunities in Agrometeorology” during 23–25 Feb-ruary 2009 in New Delhi, India. The conference was participated by about 300experts from India and 20 from abroad (USA, Korea, Egypt, Ukraine, Italy,Philippines, South Africa, China and Switzerland) including International organi-zation like WMO.The INTROMET-2009 was organized with the specific objectives to focus onthe above issues and draw attention of global agrometeorological community,administrators and policy makers to debate and devise improved methods andviitechniques for better prediction, preparedness and mitigation of the adverse weatherimpacts and aware of the possible impact, consequences and mitigation measures tosustain food security. The scientific programme was deliberated through followingeight sub-themes in addition to opening and closing session wherein 45 oralpresentations were made:l Weather Forecastingl Monsoon Variability and Crop Productionl Operational Agrometeorologyl Agromet. Information Systeml Adaptation to Climate Changel Risk Evaluation and Managementl Crop Weather Relationshipl Extreme Weather EventsFurther, 63 short oral presentations were also made on above themes along withposter display. A special session was organized to share the wisdom of VeteranScientists on “Role of IMS in addressing Challenges in Weather and ClimateService”.All the participants in the conference took part actively in discussion on thesepapers and to develop several useful recommendations for all organizationsinvolved in providing agrometeorological services to farmers to cope up withagrometeorological risk management, particularly the National Meteorologicaland Hydrological Services. The main recommendations emerged from the Confer-ence are summarized as under:l Set up a comprehensive meteorological observation system ranging surfaceincluding Agromet., upper air, radar, satellite etc. for weather forecast up todistrict/taluka level and possibly at village level.l Development user oriented meteorological information system keeping in viewregion-specific requirements of varied users including farming community.l Establishment of mechanism for greater collaboration/feedback between theproviders of information and users and also between meteorologists and agricul-ture scientists.l Develop action plan at district level for climate change, identity hot spots andpromote inter-disciplinary collaboration to enable effective mitigation ofimpacts in all sectors of economy.l Greater role in International arena through the establishment of Regional Cli-mate Centre with association of WMO and other International Organization.l Review of Agromet curriculum in Agricultural Universities with emphasis onAgromet services, Outreach and Human Resource Development.Selected papers have been edited and compiled in form of this book. As Editorsof this volume, we are highly thankful to all the authors for their efforts andcooperation in bringing out this publication. We are also grateful to the WorldMeteorological Organization and various Ministries/Departments of the Govern-ment of India like Ministry of Earth Sciences, Ministry of Science and Technology,viii PrefaceIndia Meteorological Department, and Department of Space for providing financialsupport and encouragement. Our special thanks are to the Springer for this publica-tion and Mr Subhash Khurana and Mr Dinesh Khanna for the assistance.SD AttriLS RathoreMVK SivakumarSK DashEditorsPreface ixIndian Meteorological SocietyThe Indian Meteorological Society (IMS) established in 1956 has more than 2,000members at present. The society has been able to reach not only to meteorologicalcommunity but also amongst a wide spectrum of scientists of allied fieldsfrom more than 50 national and international organizations. It carries out itsactivities from HQ office in Delhi as well as through its 17 Chapters located atdifferent places in India viz. Ahmedabad, Pune, Mumbai, Kolkata, Chennai,Nagpur, Visakhapatnam, Bhopal, Bhubaneswar, Bangalore, Hyderabad, Cochin,Thiruvananthapuram, Guwahati, Noida, Varanasi and Other Places.The IMS activities are related to encouragement and expansioncm (Fig. 2.9a). Simulated rainfall over Bihar, West Bengal and Orissais 1–15 cm which is less than the corresponding observed value of 10–25 cm.However, precipitation simulated by RegCM3 at 30 km exactly matches with theobserved values. Simulated rainfall over North-West India at both 55 km and30 km lies between 1 and 15 cm. These are close to the observed rainfall value of0–15 cm for the year 2002. The observed and simulated mean rainfall in themonth of July at 55 and 30 km resolutions in the year 2003 are depicted inFigs. 2.10a–c respectively. Over west coast of India, the simulated rainfall is80–120 cm for all the three cases. Comparing with observed rainfall over thecentral India, we found that higher resolution of 30 km predicted much accuraterainfall than 55 km resolution.Figure 2.11a–c show the monthly mean rainfall of IMD and those simulated byRegCM3 with 55 and 30 km resolutions for the year 2002 respectively. The spatialdistribution of rainfall shows differences in pattern for model simulated and IMDvalues, but some similarities exists in the zones of maximum and minimum.Simulated precipitation range rainfall for both of the resolutions over the westcoast is 200–300 cm (Fig. 2.11b, c) which is similar to the corresponding observedvalue (Fig. 2.11a). Over Bihar and Gangetic west Bengal rainfall values liesbetween 100 and 200 cm for both the observed and RegCM3 simulated with30 km resolution while with 55 km resolution these rainfall values lies between50 and 100 cm. This shows that the RegCM3 simulated rainfall with 55 km is lessthan that observed over Gangetic plain. However, higher resolution of RegCM3 at30 km depicts nearly same distribution of rainfall. Figure 2.12a–c show the monthlymean rainfall of IMD and RegCM3 simulated with 55 and 30 km respectively for30N20N10N70E 80E 90E 70E 80E 90E 70E 80E 90E30N20N10N30N20N10N0 15 30 45 80 75 90 105 120a b cFig. 2.10 July 2003 mean rainfall (cm) at the grids over Indian landmass (a) IMD 1-degree,(b) RegCM3 55 km and (c) RegCM3 30 km26 S.K. Dash et al.the year 2003. Over west coast of India, the observed rain and that simulated byRegCM3 with 30 km resolution are more than 200 cm which is reasonably good incomparison to 55 km resolution value lying between 100 and 200 cm. Simulatedrainfall values are up to 200 cm over Gangetic plain and up to 100 cm over theNorth-West India with both the resolutions of model.Table 2.1 shows monthly mean rainfall values simulated by RegCM3 with 55and 30 km resolutions using 9-ensemble members and IMD rainfall values for theyears 2002 and 2003. In 2002, the area weighted average JJAS rain amountare 70.9 cm, 69.0 cm and 70.6 cm in case of 30 km, 55 km and IMD respectively.In 2003 these values are 90.7 cm and 88.3 cm with 30 and 55 km resolutionsrespectively while the corresponding observed value is 89.4 cm. Simulated rainfallby RegCM3 with 30 km resolution shows an overestimate of about 0.4% while with30N20N10N30N20N10N30N20N10N70E 80E 90E 70E 80E 90E35030025020015010050070E 80E 90Ea bcFig. 2.11 JJAS 2002 mean rainfall (cm) at the grids over Indian landmass (a) IMD 1�, (b)RegCM3 55 km and (c) RegCM3 30 km2 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 2755 km resolution rainfall value is underestimated by about 2.3% in 2002. While in2003, simulated rain amount is overestimated by 1.4% and underestimated by 1.3%in 30 and 55 km model resolutions respectively. These results indicate that area30N20N10N30N20N10N30N20N10N70E 80E0 50 100 150 200 250 300 350 40090E70E 80E 90E70E 80E 90Ea bcFig. 2.12 JJAS 2003 mean rainfall (cm) at the grids over Indian landmass (a) IMD 1�, (b)RegCM3 55 km and (c) RegCM3 30 km28 S.K. Dash et al.Table2.1ComparisonofRegCM3simulatedrainfall(cm)atresolutions55and30kmwithobservedrainfallofIMDfortheyears2002and200320022003MonthsIMDRegCM3-55kmRegCM3-30kmIMDRegCM3-55kmRegCM3-30kmMagnitude(cm)% departurefromnormalMagnitude(cm)% departurefromnormalMagnitude(cm)% departurefromnormalMagnitude(cm)% departurefromnormalMagnitude(cm)% departurefromnormalMagnitude(cm)% departurefromnormalJune16.2104.215.9102.216102.616.9108.719.1122.718.3117.4July14.248.313.947.314.950.531.4106.730.2102.530.9105August24.895.723.18924.393.724.694.92388.724.695.1September15.489.616.193.416.998.216.59616.193.416.998.2JJAS70.680.268.978.37281.789.4101.588.3100.290.71032 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 29weighted average rainfall over India is better simulated by the RegCM3 at higherresolution of 30 km and are more accuracy in comparison to 55 km resolution.Figures 2.13 and 2.14 show the daily standardised rainfall simulated by nineensembles individually with 30 and 55 km resolutions respectively in the years0.00.20.40.60.81.01.21.41.61.82.001-Jun-0208-Jun-0215-Jun-0222-Jun-0229-Jun-0206-Jul-0213-Jul-0220-Jul-0227-Jul-0203-Aug-0210-Aug-0217-Aug-0224-Aug-0231-Aug-0207-Sep-0214-Sep-0221-Sep-0228-Sep-02DatesRainfall (cm)Member1Member2Member3Member4Member5Member6Member7Member8Member9IMD0.00.20.40.60.81.01.21.41.61.82.001-Jun-0308-Jun-0315-Jun-0322-Jun-0329-Jun-0306-Jul-0313-Jul-0320-Jul-0327-Jul-0303-Aug-0310-Aug-0317-Aug-0324-Aug-0331-Aug-0307-Sep-0314-Sep-0321-Sep-0328-Sep-03Rainfall (cm)DatesMember1Member2Member3Member4Member5Member6Member7Member8Member9IMDabFig. 2.13 Daily standardised rainfall series simulated by individual ensembles at 30 km resolutioncompared with that of IMD (a) 2002 (b) 200330 S.K. Dash et al.2002 and 2003. Figure 2.15 shows the daily accumulated rainfall simulated byRegCM3 with 30 and 55 km along with IMD daily rainfall time series, starting from1st June up to 30th September. These figures demonstrate the advantages of usinghigher resolution of 30 km compared to 55 km.0.00.20.40.60.81.01.21.41.61.82.001-Jun-0208-Jun-0215-Jun-0222-Jun-0229-Jun-0206-Jul-0213-Jul-0220-Jul-0227-Jul-0203-Aug-0210-Aug-0217-Aug-0224-Aug-0231-Aug-0207-Sep-0214-Sep-0221-Sep-0228-Sep-02DatesRainfall (cm)Member1Member2Member3Member4Member5Member6Member7Member8Member9IMD0.00.20.40.60.81.01.21.41.61.82.001-Jun-0308-Jun-0315-Jun-0322-Jun-0329-Jun-0306-Jul-0313-Jul-0320-Jul-0327-Jul-0303-Aug-0310-Aug-0317-Aug-0324-Aug-0331-Aug-0307-Sep-0314-Sep-0321-Sep-0328-Sep-03DatesRainfall (cm)Member1Member2Member3Member4Member5member6Member7Member8Member9IMDabFig. 2.14 Daily standardised rainfall series simulated by individual ensembles at 55 km resolutioncompared withthat of IMD (a) 2002 (b) 20032 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 310.00.20.40.60.81.01.21.41.61.82.01-Jun-028-Jun-0215-Jun-0222-Jun-0229-Jun-026-Jul-0213-Jul-0220-Jul-0227-Jul-023-Aug-0210-Aug-0217-Aug-0224-Aug-0231-Aug-027-Sep-0214-Sep-0221-Sep-0228-Sep-02DatesRainfall (cm)RegCM3 (30km)RegCM3 (55km)IMD0.00.20.40.60.81.01.21.41.61.82.01-Jun-038-Jun-0315-Jun-0322-Jun-0329-Jun-036-Jul-0313-Jul-0320-Jul-0327-Jul-033-Aug-0310-Aug-0317-Aug-0324-Aug-0331-Aug-037-Sep-0314-Sep-0321-Sep-0328-Sep-03DatesRainfall (cm)RegCM3 (30km)RegCM3 (55km)IMDabFig. 2.15 Daily ensemble mean rainfall simulated at 30 and 55 km of RegCM3 and IMD rainfallfrom 1st June to 30th September (a) 2002 and (b) 200332 S.K. Dash et al.2.7 ConclusionsIn this study characteristics of summer monsoon circulation, rainfall and tempera-ture simulated by RegCM3 at two different resolutions 55 and 30 km in thecontrasting monsoon years 2002 and 2003 are compared in detail. The initial andboundary conditions for model integration are obtained from the analyses of NCEP/NCAR. Also the physical parameterization schemes of the two simulations aresame. Results indicate that in both the years 2002 and 2003, higher resolutionRegCM3 at 30 km predicts much closer wind to NCEP/NCAR reanalysis field at850 hPa than that at 55 km model resolution. Also across the peninsula the strengthof the easterly jet is well simulated by RegCM3 at 30 km resolution in both theyears 2002 and 2003. In general, the maximum strengths of the Tibetan anticycloneand the easterly jet over peninsula are more in the year 2003 compared to those in2002 in both the simulations of RegCM3 at 55 and 30 km. The temperatures at500 hPa based on both the simulations of RegCM3 are close to NCEP/NCARreanalysis in 2002 and 2003. Comparison of rainfall shows that higher resolution ofRegCM3 at 30 km simulates much better rainfall magnitude and distribution closeto that of IMD in both 2002 and 2003. The percentage rainfall departure fromnormal is minimum at 55 km model resolution in July of both the years 2002 and2003. These percentage values are much closer to observed IMD at 30 km modelresolution. Comparison of daily rainfall simulated by individual ensembles showsthat at 30 km model resolution the time series are much closer to each other thanin 55 km model resolution. In most of the months 55 km model simulationunderestimates rainfall values in comparison to 30 km model simulation. In sum,results of this study demonstrate the advantages of using higher resolution RegCM3in simulating the characteristics of summer monsoon over India. However, it isessential to conduct more detailed study by considering large number of contrastingyears and also examining other important features of Indian summer monsoonincluding intra seasonal variabilities.ReferencesAzadi M, Mohanty UC, Madan OP, Padmanabhamurti B (2001) Prediction of precipitationassociated with a western disturbance using a high-resolution regional model – role ofparameterization of physical processes. Meteorol Appl 7:317–326Bhaskar Rao DV, Ashok K, Yamagata T (2004) A numerical simulation study of the Indiansummer monsoon of 1994 using NCARMM5. J Meteorol Soc Jpn 82(6):1755–1775Bhaskaran BR, Jones G,Murphy JM, NoguerM (1996) Simulations of the Indian summermonsoonusing a nested regional climate model: domain size experiments. Clim Dyn 12:573–578Dash SK, Shekhar MS, Singh GP (2006) Simulation of Indian summer monsoon circulation andrainfall using RegCM3. Theor Appl Climatol 86(1-4):161–172Dash SK, Kulkarani MA, Mohanty UC, Prasad K (2009) Changes in the characteristics of rainevents in India. J Geophys Res 114:12. doi:10.1029/2008JD 0105722 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 33Gadgil S, Srinivasan J, Nanjundiah RS, Kumar KK, Munot AA, Kumar KR (2002) On forecastingthe Indian summer monsoon: the intriguing season of 2002. Curr Sci 4:394–403Giorgi F, Marinucci MR, Bates GT (1993a) Development of a second generation regional climatemodel (RegCM2). Part I: boundary-layer and radiative transfer processes. Mon Wea Rev 121:2794–2813Giorgi F, Marinucci MR, Bates GT, De Canio G (1993b) Development of a second-generationregional climate model (RegCM2). Part II: convective processes and assimilation of lateralboundary conditions. Mon Wea Rev 121:2814–2832Grell GA, Dudhia J, Stauffer D (1994) A description of the fifth-generation Penn-State/NCARmesoscale model. NCAR technical report note TN-398, National Center for AtmosphericResearch, Boulder, Colorado, USAJi Y, Vernekar AD (1997) Simulation of the Asian summer monsoons of 1987 and 1988 with aregional model nested in a global GCM. J Clim 10:1965–1979Kalsi SR, Hatwar HR, Subramanian SK, Rajeevan M, Jayanthi N, Shyamala B, Jenamani RK(2004) IMD Monograph, Synop Meteorol No. 2/2004, pp 105Rajeevan M, Pai DS, Dikshi SK, Kelkar RR (2004) IMD’s new operational models for long-rangeforecast of southwest monsoon rainfall over India and their verification for 2003. Curr Sci86(No.3):422–31Rajeevan M, Bhate J, Kale JD, Lal B (2005) Development of a high resolution daily griddedrainfall data for the Indian region. IMD Met Monogr, Climatol No. 22/2005, pp 27Rajeevan M, Jyoti B, Kale JD, Lal B (2006) Development of a high resolution daily griddedrainfall data for the Indian region: analysis of break and active monsoon spells. Curr Sci91(3):296–306Shekhar MS, Dash SK (2005) Effect of Tibetan spring snow on the Indian summer monsooncirculation and associated rainfall. Curr Sci 88(No. 11):1840–1844Shepard D (1968) A two-dimensional interpolation function for irregularly spaced data, Proc. 23ACM Nat’l Conf., Princeton, NJ, Brandon/Systems Press, 517–524Singh GP, Jai-Ho Oh, Jin-Young K, Ok-Yeon K (2006) Sensitivity of summer monsoon precipi-tation over east Asia to convective parameterization schemes in RegCM3. SOLA 2:29–3234 S.K. Dash et al.Chapter 3Simulation of Heavy Rainfall in Associationwith Extreme Weather Events: Impacton AgricultureU.C. Mohanty, S. Pattanayak, A.J. Litta, A. Routray,and O. Krishna KishoreAbstract The extreme weather systems like thunderstorms, tropical cyclones,heavy rainfall and monsoon depression evolve through different scales of pro-cesses. The genesis and development of weather events involve complex interac-tion mechanism of mesoscale convective organization embedded in large-scalecirculation. The skill of prediction of rainfall associated with extreme weatherevents as well as its impact on agriculture has been demonstrated through theclimatological concept and the numerical simulations with the WRF modelingsystems.3.1 IntroductionThe tropical weather and climate systems are more dominated by physical forcing(complex land-air-ocean interaction processes) than the dynamical forcing unlike inthe mid-latitude, and hence weather prediction in tropics is more challengingtask for the meteorological community. Although there have been significantimprovements in numerical weather prediction (NWP) in last three decades, theprediction of extreme weather events is still a difficult and challenging task for theoperational forecasters, particularly in the developing nations. Convective activityis one of the major processes in the atmosphere influencing the local and large-scaleweather in the tropics. The Southeast Asia is mainly dominated by the summermonsoon and the tropical ocean regions act as main reservoirs of heat and moisturein supplying the necessary energy to the establishment and maintenance of the largescale circulation andassociated monsoon activity over Indian subcontinent. A lateU.C. Mohanty (*) • S. Pattanayak • A.J. Litta • A. Routray • O. Krishna KishoreCentre for Atmospheric Sciences, Indian Institute of Technology, Delhi, Hauz Khas, New Delhi110016, Indiae-mail: mohanty@cas.iitd.ernet.in; sujata05@gmail.com; ajlitta@gmail.com; ashishroutray.iitd@gmail.com; osurikishore@gmail.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_3, # Springer-Verlag Berlin Heidelberg 201135or early onset of the monsoon or an untimely break period in the monsoon may havedevastating effects on agriculture even if the mean seasonal rainfall is normal.At the same time, meso-convective activity embedded with large-scale/synopticcirculations lead to extreme weather events like intense tropical cyclones (TCs),severe thunderstorms (STSs), and monsoon depressions (MDs) etc. have time andagain contributed a large amount of precipitation in terms of rain, snow or hailwhich has a very hazardous impact on Indian agriculture.Weather forecasting is important for each and every agricultural operation. Theforecast of the sowing rains is one event for which the weather forecaster provides acritical input for the agriculture. At other stages of crop growth a combination ofweather elements like rain, temperature, cloudiness, humidity, solar radiation,wind, evaporation and soil moisture become crucial for good crop production.Strong winds, heavy and torrential rains and the worst of all the storm surges arethe most devastated elements of tropical cyclones, causing a lot of damage in thecoastal region. Also, the agriculture of low laying delta regions are worst affectedby costal inundation and intrusion of saline water due to storm surges. At the sametime, farmers use various techniques for the mitigation of dry season e.g. changingsowing dates, using different types of seeds, storage of water, multiple cropping,relay cropping and intercropping etc. Hail is one of the most shocking natural perilsfor farmers. Hail occurs overnight whereas other perils like drought give farmer’stime to agonize and cope. Although there are numerous hail factors influencing cropdamage (kinetic energy, number of hailstones of relatively large diameter, momen-tum, mass, mean diameter), damage appears to best correlated with maximumhailstone size. Timely and reasonably accurate weather/climate forecast of thesehigh impact weather events can significantly reduce the loss of lives, damage toproperties and an awareness of the events to the general public as well as theplanners for scheduling their action to a great extent. Hence, as far as the modelforecast is concerned, high-resolution non-hydrostatic mesoscale models areexpected to provide possibly the best forecast.A brief description of the modeling system used in this study is given inSect. 3.2. The simulation of severe thunderstorms is presented in Sect. 3.3. Thesimulations of intense tropical cyclones are provided in Sect. 3.4. Section 3.5 dealswith the simulation of heavy rainfall and monsoon depression and the conclusionsin Sect. 3.6.3.2 Modeling SystemThe Weather Research and Forecasting (WRF) Model is a next-generation meso-scale forecast model that will be used to advance the understanding and theprediction of mesoscale convective systems. It features multiple dynamical cores,a 3-dimensional variational (3DVAR) data assimilation system, and a softwarearchitecture allowing for computational parallelism and system extensibility. TheWRF model will be used for a wide range of applications, from idealized research36 U.C. Mohanty et al.to operational forecasting, across scales ranging from meters to thousands ofkilometers. WRF can resolve the small-scale weather features such as front, local-ized convection, hurricane core, and topographic effect much better than the globalmodel.The Weather Research and Forecasting (WRF) model contains two dynamiccores: the Non-hydrostatic Mesoscale Model (NMM – Janjic 2003) core, developedat the National Centers for Environmental Prediction (NCEP) and the AdvancedResearch WRF (ARW – Skamarock et al. 2005) core, developed at the NationalCenter forAtmospheric Research (NCAR). NMM runs are initialized through the samebasic mechanism as the ARW runs: the WRF preprocessing system (WPS) readsGRIB data from an initializing model and interpolates it onto the target WRFdomain grid. However, the functionality of the WPS had to be expanded to handlethe horizontal staggering, map projection, and vertical coordinate used by theNMM, as each is distinct from its ARW counterpart. The NMM is a fully com-pressible, non-hydrostatic mesoscale model with a hydrostatic option. The modeluses a terrain following hybrid sigma-pressure vertical coordinate. NMM modelsurfaces are terrain-following sigma surfaces near the ground, purely isobaric abovea prescribed pressure value (typically about 420 hPa), and relax from terrainfollowing to isobaric over the intervening depth. Further details of the verticalcoordinate can be found in Janjic (2003). While ARW model’s vertical coordinateis a terrain-following hydrostatic pressure coordinate.NMM model uses a forward-backward scheme for horizontally propagating fastwaves, implicit scheme for vertically propagating sound waves, Adams-Bashforthscheme for horizontal advection, and Crank-Nicholson scheme for vertical advec-tion. The same time step is used for all terms. The dynamics conserve a number offirst and second order quantities including energy and entropy (Janjic 1984), whileARWmodel uses higher order numerics. This includes the Runge-Kutta second andthird order time integration schemes, and second to sixth order advection schemesin both horizontal and vertical directions. It uses a time-split small step for acousticand gravity-wave modes. The dynamics conserves scalar variables. Both modelssupport a variety of capabilities, which include real-data simulations, full physicsoptions, non-hydrostatic and hydrostatic (runtime option), one-way static nestingand applications ranging from meters to thousands of kilometers. Table 3.1 gives abrief illustration of the two WRF dynamic systems.Along with WRF, a three dimensional variational data assimilation system wasdeveloped based on fifth generation Mesoscale Model (MM5) 3DVAR system(Barker et al. 2004). The analysis system must be able to properly resolve allvariables and scales of importance of the assimilation system. This formidableproblem will be addressed incrementally by analyzing scales and variables, whichwe currently know how to do and developing techniques to perform the analysis atsmaller scales and with additional analysis variables. 3DVAR analysis scheme iscapable of analyzing most standard meteorological data, and more recentlyincluded a number of non-traditional data types (e.g. SSM/I radiances and radarradial velocity). The WRF assimilation research will strive to develop the “forward3 Simulation of Heavy Rainfall in Association with Extreme Weather Events 37models” (and their adjoints) by which the WRF model fields can be transformedand interpolated to the observation locations and values for each of these data, andfor these operators to be portable to alternative assimilation algorithms (e.g.alternative 3DVAR systems, 4DVAR, ensemble-based DA, etc.).Initial and boundary conditions for all the following cases are derived from6-hourly global final analysis (FNL) at 1.00 � 1.00 grids generated by NationalCenter for Environmental Prediction (NCEP)’s global forecast system (GFS).Table 3.2 briefly describes the model configuration of different convectiveactivities.3.3 Simulation of Severe ThunderstormThunderstorm is a mesoscale system of space scale a few kilometers to a couple of100 km and time scale of less than an hour to several hours.Severe thunderstormsform and move from northwest to southeast over the Eastern and NortheasternTable 3.1 A brief illustration of the two WRF dynamic systemsARW NMMTerrain following sigma vertical coordinate Hybrid sigma to pressure vertical coordinateArakawa C-grid Arakawa E-gridThird order Runge-Kutta time-split differencingAdams-Bashforth time differencing withtime splittingConserves mass, entropy and scalars using up tosixth order spatial differencing equation forfluxes (fifth order upwind diff. is default)Conserves kinetic energy, enstrophy andmomentum using Second orderdifferencing equationNCAR physics package Eta/NAM physicsNoah unified land-surface model Noah unified land-surface modelTable 3.2 Model configuration of different convective activityThunderstorm Tropical cycloneHeavy rainfall/monsoondepressionRadiationparameterization GFDL/GFDL GFDL/GFDL Rrtm/dudhiaSurface layerparameterizationJanjic similarityscheme NCEP GFS scheme Janjic similarity schemeLand surfaceparameterizationNMM Land surfaceschemeNMM Land surfaceschemeNoah land surfaceschemeCumulusparameterizationGrell-Devenyiensemble schemeSimplified –Arakawa-SchubertGrell-Devenyi ensembleschemePBLparameterization Mellor-Yamada-Janjic YSU YSUMicrophysics Ferrier scheme Ferrier scheme Ferrier scheme38 U.C. Mohanty et al.states of India (i.e., Gangetic West Bengal, Jharkhand, Orissa, Assam and partsof Bihar) during the pre-monsoon season (April–May). They are locally called“Kal-baishakhi” or “Nor’westers”. Strong heating of landmass during mid-dayinitiates convection over Chhotanagpur Plateau, which moves southeast and getsintensified by mixing with warm moist air mass from head Bay of Bengal (BoB).It produces heavy rain showers, lightning, thunder, hail-storms, dust-storms, surfacewind squalls, down-bursts and tornadoes. The northwest India gets convective dust-storms called locally Andhi in this season. These severe thunderstorms associatedwith thunder, squall lines, lightning, torrential rain and hail cause extensive loss inagriculture, damage to property and also loss of life.The casualties reported due to lightning associated with thunderstorms in thisregion are the highest in the world. The strong wind produced by the thunderstormis a real threat to aviation. The highest numbers of aviation hazards are reportedduring occurrence of these thunderstorms. In India, 72% of tornadoes are associatedwith Nor’westers. India is among the countries in the world having large frequencyof hail. Reviewing the annual reports of India Meteorological Department from1982 to 1989, Nizamuddin (1993) finds that there were 228 hail days (about 29 peryear) of moderate to severe intensity. Hail size comparable to mangoes, lemons andtennis balls has been observed. Eliot (1899) found that out of 597 hailstorms inIndia 153 yielded hailstones of diameter 3 cm or greater. These events killed 250persons and caused extensive damage to winter wheat crops. These severethunderstorms have significant socio-economic impact in the eastern and northeast-ern parts of the country. An accurate location specific and timely prediction isrequired to avoid loss of lives and property due to strong winds and heavy precipi-tation associated with these sever weather system.One of the main difficult tasks in weather prediction is the thunderstormforecasting mainly because of its rather small temporal spatial extension and theinherent non-linearity of their dynamics and physics (Orlanski 1975). Accuratesimulation requires knowledge about “where” and “when” storms will develop andhow they will evolve. The high resolution non-hydrostatic mesoscale models withsophisticated parameterization schemes for the important physical processes wouldbe very useful tool for reasonably accurate prediction of these severe thunderstorms(Weiss et al. 2007). Several studies related to the simulation of severe thunderstormevents using WRF-NMM model have been performed (Kain et al. 2006; Litta andMohanty 2008; Litta et al. 2009). In the present study, an attempt is made tosimulate a severe thunderstorm event that occurred over Kolkata (22.65� N,88.45� E) on 21 May 2007, using WRF-NMM model and validated the modelresults with observational data.The occurrence of pre and post monsoon thunderstorms over Indian continent isa special feature. Thunderstorms are associated with heavy rainfall during shortduration of 2–3 h. It is useful for vegetations when it is in hot summer season.A severe thunderstorm occurred over Kolkata on 21 May 2007 at 1100 UTC istaken here for the present study. A squall line was reported over MO Kolkata at1100 UTC from northwesterly direction with maximum wind speed of 19 ms�1 andlasted for 1 min. This intense convective event produced 20 mm rainfall over3 Simulation of Heavy Rainfall in Association with Extreme Weather Events 39Kolkata. For this event, synoptic chart analysis shows low-level cyclonic circula-tion lies over Jharkhand/Gangetic West Bengal (GWB) and extended up to midtropospheric levels. Coastal winds of Orissa/GWB are westerly to southeasterly.Good moisture incursion over East and North-East states was noticed (Mohantyet al. 2007). In the present simulation study, the model was integrated for a period of24 h, starting from 21 May 2007 at 0000 UTC as initial values. A single domainwith 3-km horizontal spatial resolution was configured.Kolkata DWR composite radar reflectivity pictures on 21 May 2007 from 0800to 1100 UTC is shown in Fig. 3.1. By analyzing DWR pictures, scattered echoes aredeveloped near Dumka at 0800 UTC and moving south eastwards at 0900 UTC.This echo is intensified into a squall line (30 km north of Kolkata) at 1000 UTC.This squall line moved further in southeast direction. NMM model simulatedcomposite radar reflectivity on 21 May 2007 from 0800 UTC to 1100 UTC isshown in Fig. 3.2. By analyzing NMMmodel simulated composite radar reflectivitypictures, scattered echoes developed north-west of Kolkata at 0800 UTC. This echowas moving south eastwards at 0900 UTC and intensified at 1000 UTC. This echomoved further in southeast direction. The model well simulated this squall lineFig. 3.1 Kolkata DWR composite radar reflectivity pictures from 0800 to 1100 UTC on 21 May200740 U.C. Mohanty et al.movement with simulated composite radar reflectivity fields. Variation of convec-tion in the atmosphere depends upon dynamics as well as thermodynamic instabil-ity indices. A number of stability indices are devised in order to detect the likelyoccurrence of thunderstorms. The NMM model simulated skew-t plot andcorresponding stability indices of Kolkata (22.65� N, 88.45� E) at 1100 UTC on21 May 2007 is illustrated in Fig. 3.3. The skew-t plot shows that the atmospherewas convectively unstable at 1100 UTC. The NMM model simulated CAPEvalue (2314 Jkg�1) is high and is a favorable condition for severe thunderstorms.The model simulated TT index at 1100 UTC is showing a very high value (46).Examination of all the model simulated stability indices clearly indicated thatNMM model has well captured the instability of the atmosphere at 1100 UTC forthe occurrence of a severe thunderstorm.Figure 3.4a shows the inter-comparison of observed (AWS) and NMM modelsimulated diurnal variation of surface temperature (�C) over Kolkata valid forRadar Reflectivity (8UTC)Radar Reflectivity (10UTC) Radar Reflectivity (11UTC)Radar Reflectivity (9UTC)24N23.7N23.4N23.1N22.8N22.5N22.2N21.9N21.6N21.3N21N24N23.7N23.4N23.1N22.8N22.5N22.2N21.9N21.6N21.3N21N24N23.7N23.4N23.1N22.8N22.5N22.2N21.9N21.6N21.3N21N24N23.7N23.4N23.1N22.8N22.5N22.2N21.9N21.6N21.3N21N86.4E 86.7E 87E 87.3E 87.6E 87.9E 88.2E 88.5E 88.8E 89.1E 89.4E 86.4E 86.7E 87E 87.3E 87.6E 87.9E88.2E 88.5E 88.8E 89.1E 89.4E86.4E 86.7E 87E 87.3E 87.6E 87.9E 88.2E 88.5E 88.8E 89.1E 89.4E 86.4E 86.7E 87E 87.3E 87.6E 87.9E 88.2E 88.5E 88.8E 89.1E 89.4E3025201510551015200Fig. 3.2 WRF-NMM simulated composite radar reflectivity pictures from 0800 to 1100 UTC on21 May 20073 Simulation of Heavy Rainfall in Association with Extreme Weather Events 4121 May 2007 at 0000 UTC to 22 May 2007 at 0000 UTC. From the figure, we canclearly see that model captured the variation with drop in temperature at 1000 UTC,1 h before the observed. Observed temperature showed a sudden drop from 30�C to24�C at 1100 UTC whereas the model simulation shows a drop from 33�C to 27�Cat 1000 UTC. Figure 3.4b shows the inter-comparison of observed and NMMmodelsimulated accumulated progressive rainfall at Kolkata valid for 21 May 2007 at0000 UTC to 22 May 2007 at 0000 UTC. Model has able to capture 18.5 mm ofrainfall at 1000 UTC, which is close to the actual observation (20.0 mm). Model haspredicted the rainfall at 1000 UTC, which is 1 h prior to the actual thunderstormoccurrence (1100 UTC). Relative humidity at surface level has also been taken intoaccount, as it is an essential factor in intense convection. Model has captured(Fig. 3.4c) the rising of relative humidity values during the model simulatedthunderstorm hour as in the observation. The observed relative humidity valuespeaked from 69% to 97% at 1100 UTC whereas model showed a sharp rise fromaround 56% to 86% at 1000 UTC, which is 1 h prior to the observed.RHmb10025200300400500600700800900100050 75 100(%)−40 −30 −20 −10Temperature (°C)0 10 20 30 40K 31TT 46PW(cm) 5SurfaceMost UnstablePress(mb) 950 Temp(°C) 28.6Dewp(°C) 25.5q c(K) 368q c(K) 369LI −6LI −6CAPE(J) 2096CAPE(J) 2314CIN(J) 94CIN(J) 26Mixing Ratio (g/kg)1 2 3 4 5 0 10 15 20 25 303540LCLFig. 3.3 The NMM model simulated skew-t plot of Kolkata at 1100 UTC on 21 May 200742 U.C. Mohanty et al.2123252729313335373900Z02Z04Z06Z08Z10Z12Z14Z16Z18Z20Z22Z00ZTime (UTC)Temperature (C)OBSNMM3040506070809010011000Z02Z04Z06Z08Z10Z12Z14Z16Z18Z20Z22Z00ZTime (UTC)Relativehumidity (%)OBSNMMa051015202500Z02Z04Z06Z08Z10Z12Z14Z16Z18Z20Z22Z00ZTime (UTC)Accumulated Rainfall (mm)OBSNMMbcFig. 3.4 (a–c) The inter-comparison of observed and WRF-NMM model simulated diurnalvariation of (a) temperature (�C) (b) accumulated rainfall (mm) (c) relative humidity (%) overKolkata valid for 21 May 2007 at 0000 UTC to 22 May 2007 at 0000 UTC3 Simulation of Heavy Rainfall in Association with Extreme Weather Events 433.4 Simulation of Tropical CyclonesTropical cyclones are among nature’s most violent manifestations and potentially adeadly meteorological phenomenon. These are large low-pressure storms that formover the tropical oceans of the Earth’s low latitudes, typically between 30�N and30�S. The Bay of Bengal is a potentially energetic region for the development ofcyclonic storms and about 7% of the global annual tropical storms forms over thisregion. Moreover, the Bay of Bengal storms are exceptionally devastating, espe-cially when they cross the land (Pattanayak and Mohanty 2008). Thus the Bay ofBengal tropical cyclone is the deadliest natural hazard in the Indian sub-continent.It has significant socio-economic impact on countries bordering the Bay of Bengal,especially India, Bangladesh and Myanmar. At the same time, Arabian Seacontributes 2% of the global annual tropical storms. Therefore, reasonably accurateprediction of these storms is important to avoid the loss of lives. The trend of thetropical cyclones is increasing from 1905 to recent years and there has been a net0.8�C increase in SST during this period. Also there is a drastic increase in the TCtrend from 1995 onwards (Holland and Webster 2007). Bell and Chelliah (2006)provide a detailed summary of natural variability for long period tropical cyclones.Tropical cyclones cause variety of damages. The major causes of damage arestrong wind, storm surge and heavy precipitation. Strong wind can cause damage totall structures, crops, power supply and communication systems. It also causes lossof life and generates devastating storm surges. Storm surge causes coastal flooding,damages to property, loss of life and intrusion of saline water in coastal areas.Intrusion of saline water in turn causes loss of soil fertility in coastal areas. Whileeffects of strong wind and storm surge are concentrated within few tens ofkilometers from the coastline, heavy rainfall often affects hundreds of kilometersfrom the coast. Flooding due to heavy rainfall may occur even after the storm haslost its hurricane intensity. In general, heavy precipitation follows the storm and themagnitude of precipitation varies with the intensity of the storm. Heavy rainfallcauses loss of life, destruction of vegetation, crops and livestock and contaminationof water supply. Hence, there is a direct impact on agriculture with the severe landfalling tropical cyclones. The loss of life and property due to tropical cyclonedisaster can be significantly reduced by reasonably accurate forecast of the stormand associated surge. In the present study, an attempt is made to simulate severetropical cyclones that occurred over Bay of Bengal (Case I: Mala) and Arabian Sea(Case II: Gonu), using WRF-NMM model and validated the model results withobservational data.3.4.1 Case-IThe cyclone Mala has developed over the warm tropical ocean, near 9.5� N, 90.5� E,around 0300 UTC 25 April 2006 with the central mean sea level pressure of 996 hPaand the maximum sustainable wind of 25 kts. The system remained at that stage for44 U.C. Mohanty et al.further 6 h i.e. up to 0900 UTC 25 April 2006. By 0900 UTC 25 April 2006, it wasturned into deep depression stage. Then the system became cyclonic storm after1200 UTC 25 April 2006. At 0000 UTC 26 April 2006, the clear-cut cyclonic stormwith the center of the storm at 10.5� N and 89.0� E with the central pressure of994 hPa and maximum surface wind of 45 kts was observed. The system remainedin cyclonic stage up to 0000 UTC 27 April 2006. Then around 0300 UTC 27 April2006, it became severe cyclonic storm with central pressure of 990 hPa and themaximum sustainable wind of 55 kts. The system became very severe cyclonicstorm (VSCS) by 1200 UTC 27 April 2006 with the central mean sea level pressureof 984 hPa and the maximum surface wind of 65 kts. The storm remained in VSCSfor a period of 42 h i.e. up to 0600 UTC 29 April 2006. The maximum observedcentral pressure was 954 hPa with the pressure drop of 52 hPa. The observedmaximum sustainable surface wind was 100 kts. The very severe cyclonic stormcrossed the Arakan coast at about 100 km south to Sandoway around 0700 UTC 29April 2006. The system remained on the land for further 12 h and caused a lot ofdevastation in the underlying coastal areas.Figure 3.5a, b represents the day-3 forecast of mean sea level pressure and windat 850 hPa valid at 0000 UTC 29 April 2006. At the Day-3 forecast, modelsimulation shows that, the storm moved northeastward from (13.7� N/91.2� E) to(17.7� N/93.2� E) in last 24 h with a MSLP of 960 hPa which is nearly matchingwith that of the observation. The maximum observed MSLP was 954 hPa with thepressure drop of 52 hPa. The WRF-NMM model simulated the maximum MSLPof 960 hPa with the pressure drop of 45 hPa. Figure 3.6a, b represents the 24 haccumulatedprecipitation valid at 0300 UTC 29 April 2006. Figure 3.6a represents24 h accumulated precipitation as a merged analysis of Tropical Rainfall MeasuringMission (TRMM), TMI and rain gauges observations carried out by NationalAeronautics and Space Administration (NASA) valid for 0300 UTC 29 AprilFig. 3.5 Simulation of mean sea level pressure & wind at 850 hPa for case-I (Mala) (a) MSLPfor Day-3 and (b) wind for Day-33 Simulation of Heavy Rainfall in Association with Extreme Weather Events 452006. Figure 3.6b represents the Day-3 forecast of accumulated precipitation validat 0300 UTC 29 April 2006 from WRF-NMM. The observed precipitation is about40 cm in Day-3 whereas the model could simulate the precipitation of 32 cm. Thetrack of the cyclone as obtained with the model simulations from different initialconditions are evaluated and compared with the best-fit track obtained from IndiaMeteorological Department (IMD). Figure 3.7 represents the track of the cycloneMala as obtained with model simulations from different initial conditions. Resultsshow that, in each case the cyclone moves to the Arakan coast, whatever the initialcondition is being chosen. The vector displacement errors (VDEs) are also calcu-lated at every 12 h interval and the detailed of the VDEs are provided in Table 3.3.3.4.2 Case-IIThe tropical storm Gonu has developed as a depression over the east centralArabian Sea with center near lat 15.0� N, long 68.0� E at 1800 UTC 01 June2007. It moved westwards and intensified into a cyclonic storm at 0900 UTC 02June 2007 near lat 15.0� N, long 67.0� E. It remained in that stage for 15 h i.e. up to0000 UTC 03 June 2007. By 0000 UTC 03 June 2007, it intensified into a severecyclonic storm with the central pressure of 988 hPa and centered at lat 15.5� N, long66.5� E and the storm remained in that stage for next 18 h i.e. up to 1500 UTC 03June 2007. Continuing its northwestward movement, it further intensified into avery severe cyclonic storm by 1800 UTC 03 June 2007 and lay centered at lat 18.0�N, long 66.0� E with the central pressure of 980 hPa. It sustained in that stage fornext 18 h i.e. up to 1500 UTC 04 June 2007. By 1500 UTC 04 June 2007, the systemFig. 3.6 (a and b) 24 h accumulated precipitation valid at 0300 UTC 29 April 2006. (a) Observedfrom TRMM and (b) simulated with WRF-NMM (Day-3)46 U.C. Mohanty et al.moved west-northwestwards and further intensified as a super cyclonic storm andlay centered at lat 20.0� N, long 64.0� E with the minimum central pressure of920 hPa. It remained in the super cyclonic storm stage for the next 6 h i.e. up to0000 UTC 05 June 2007. Then the storm further moved in northwestward directionand weakened into a very severe cyclonic storm by 2100 UTC 04 June 2007 andlay centered over northwest Arabian Sea at lat 20.5� N, long 63.5� E with theminimum central pressure of 935 hPa. The storm remained in that stage for the next48 h i.e. up to 2100 UTC 06 June 2007. Then the storm gradually weakened, movednorthwestward and crossed the Makran coast near lat 25.0� N, long 58.0� E between0300 and 0400 UTC 07 June 2007 as a cyclonic storm.24N22N20N18N16N14N12N10N8N6N76E 78E 80E 82E 84E 86E 88E 90E 92E 94E 96E 98E 100EFig. 3.7 Track of the cyclone MalaTable 3.3 Vector displacement errors for case-I (Mala)Initial time of model integration 00 h 00 h 12 h 36 h 48 hMala-1(2500) 118.0 118.0 98.3 148.1 202.1Mala-2(2512) 80.8 80.8 71.1 248.2 372.3Mala-3(2600) 22.2 22.2 74.5 228.9 295.6Mala-4(2612) 35.1 35.1 115.9 323.6 546.6Mala-5(2700) 55.5 55.5 124.1 323.6 571.4Mean error 62.3 62.3 96. 7 254.4 397.63 Simulation of Heavy Rainfall in Association with Extreme Weather Events 47Figure 3.8a, b represents the day-3 forecast of mean sea level pressure and windat 850 hPa valid at 0000 UTC 06 June 2007. The model simulation shows thatthe storm moved northwestward with the minimum central pressure of 981 hPa. Theobserved central pressure at that time was 970 hPa. The observed maximumsustained surface wind at that time was 77 kts, whereas the model could simulatethe maximum wind of 66 kts. Figure 3.9 represents the 24 h accumulatedFig. 3.8 Simulation of mean sea level pressure & wind at 850 hPa for case-II (Gonu), (a) MSLPfor Day-3 and (b) wind for Day-3Fig. 3.9 (a and b) 24 h accumulated precipitation valid at 0300 UTC 06 June 2007. (a) Observedfrom TRMM and (b) simulated with WRF-NMM (Day-3)48 U.C. Mohanty et al.precipitation valid at 0300 UTC of 06 June 2007. Figure 3.9a represents 24 hTRMM accumulated precipitation valid at 0300 UTC of 06 June 2007. Figure 3.9brepresents the Day-3 forecast of accumulated precipitation valid at 0300 UTC 06June 2007 from WRF-NMM. The observed precipitation is about 16 cm in Day-3whereas the model could simulate the precipitation of 14 cm. The track of thecyclone as obtained with both the model simulations from different initialconditions are evaluated and compared with the best-fit track obtained from IMD.Figure 3.10 represents the track of the cyclone Gonu as obtained with modelsimulations from different initial conditions. The vector displacement errors(VDEs) are also calculated at every 12 h interval and the detailed of the VDEsare provided in Table 3.4.28N26N24N22N20N18N16N14N12N10N8N56E 58E 60E 62E 64E 66E 68E 70E 72E 74EFig. 3.10 Track of the cyclone GonuTable 3.4 Vector displacement errors for case-II (Gonu)Initial time of model integration 00 h 12 h 24 h 36 h 48 hGonu-1(0200) 100.0 95.0 88.5 128.1 142.0Gonu-2(0212) 60.8 70.0 88.3 124.2 272.3Gonu-3(0300) 32.0 65.4 140.7 188.0 225.0Gonu-4(0312) 59.1 102.3 168.0 202.4 325.4Gonu-5(0400) 50.1 84.1 125.1 223.6 371.0Mean error 60.4 83.36 122.12 173.26 267.143 Simulation of Heavy Rainfall in Association with Extreme Weather Events 493.5 Simulation of Heavy Rainfall EventsSouth west (SW) monsoon is the main feature in the climate of India as well as theprincipal denominator of the prosperity of the country and the agro-economy. Mostof the country as a whole receives 70% of the total annual rainfall except southernparts of peninsula, especially Tamil Nadu (Parthasarathy 1984) during this SWmonsoon season. Hence, it is not only nourishes the kharif crops but enriches allsources of irrigation to enable cultivation of a wide variety of crops during rabi andpre-kharif season. As a result all sections of people starting from a farmer to thescientist and from a common man to the policy makers at the highest level are keento know detail about the variation of weather of the season, its causes and effectsespecially on agriculture. Vulnerability and agricultural sustainability are primarilylocal issues, and depend critically on the amount and temporal distribution ofrainfall received over a region.Heavy rainfall pose a serious threat to many sectors particularly, agriculturalsector. According Dash et al. (2009), the frequencies of moderate and low rain daysconsidered over the entire country have significantly decreased in the last halfcentury. On the basis of the duration of rain events it is inferred that long spellsshow a significant decreasing trend over India as a whole while short and dry spellsindicate an increasing tendency with 5% significance. Field flooding during theplant growing season causes crop losses due to low oxygen levels in the soil,increased susceptibility to root diseases, and increased soil compaction due to theuse of heavy farm equipment on wet soils. If the flood hit just as farmers start toharvest the crops, the losses to be very large. The flooding severely erodes uplandsoils where erosion puts some farmers out of business. The flooding also causes anincrease in runoff and leaching of agricultural chemicals into surface water andgroundwater. Another impact of heavy rainfall is that wet conditions at harvest timeresultin reduced quality of many crops. Vegetable and fruit crops are sensitive toeven short-term, minor stresses, and as such are particularly vulnerable to weatherextremes. Therefore, an accurate estimate of frequency, intensity and distribution ofthese events can significantly aid policy planning.Over the Indian monsoon region in particular, the better simulation of weatherevents like heavy rainfall and monsoon depressions are important and routinelyhave resulted in flooding and significant loss to agriculture, life and property duringthe Indian monsoon. But the current mesoscale models have limited success insimulating these events over the region (Routray et al. 2005). The forecast per-formance of the mesoscale models critically depends on the quality of initialconditions (Pielke et al. 2006). Typically, large scale global analyses which providethe initial condition to the mesoscale models have limitations such as coarseresolution and inadequate representation of localized mesoscale features. There-fore, assimilation approaches like 3-dimensional variational data assimilation(3DVAR) that ingest local observations are important to develop improvedanalyses (Daley 1991) and hence the forecast. Vinod Kumar et al. (2007) adoptedfour dimensional data assimilation (FDDA) and surface data assimilation to study50 U.C. Mohanty et al.tropical depressions over Bay of Bengal. The results suggested that improvement ofmonsoon depression simulations over Bay of Bengal were equivalent or better thanthat of increasing the model resolution from 30 to 10 km grid spacing. The mainpurpose of this study is to demonstrate the ability of the WRF-ARW mesoscalemodel in simulating the heavy rainfall events like Mumbai heavy rainfall (26–27July 2005) and monsoon depressions (Case I: 2–4 August 2006 and Case II: 21–23June 2007) with and without assimilation of Indian conventional and non-conven-tional observations using 3DVAR assimilation system.3.5.1 Mumbai Heavy RainfallThis was a record-braking rain evnt, Santacruz received 94.4 cm within 24 h and thestudy by Jenamani et al. (2006) described details of the event. To simulate thislocalized and intense rainfall event, two numerical experiments with (3DV) andwithout (CNTL) data assimilation are carried out. For both the experiments, themodel integrated 36 h from 0000 UTC of 26 July 2005 up to 1200 UTC of 27 July2005. Figure 3.11 shows the temporal evolution of model and observed cumulativerainfall over Santacruz and Colaba IMD recording stations along 72.5� E. It isclearly observed that the record-breaking rainfall at Santacruz is significantlyimproved in the 3DV simulation (more than 78 cm) compared with the CNTLsimulation (36 cm). In the same time, the 3DV experiment is also well simulatedlow rainfall amount over Colaba as compared to the CNTL simulation and trend iswell match with the observations. Figure 3.12 shows the 24 h accumulated precipi-tation as obtained from CNTL and 3DV experiments and satellite estimates rainfall(TRMM) along with the IMD ground based rain gauge observations at and aroundMumbai. The orientation and distribution of rainfall is well simulated in the 3DV010203040506070809010003Z 26Jul06Z 09Z 12Z 15Z 18Z 21Z 00Z 27 Jul03Z Time (UTC)Cumulative rainfall (cm)Santacruz (obs)CNTL3DVColaba (obs)CNTL3DVFig. 3.11 Observed and simulated 3-hourly interval cumulative rainfall (cm) over Santacruz andColaba on 26–27 July 20053 Simulation of Heavy Rainfall in Association with Extreme Weather Events 51experiment (Fig. 3.12d) and feature is very much match with the satellite estimatedrainfall pattern (Fig. 3.12b). However, the CNTL experiment (Fig. 3.12c) shows thepatch of maximum rain (50–60 cm) very much away from the observed locationwith a location error 120 km northeast of Santacruz, Mumbai. Further, we evaluatethe model performance with 3DVAR based initial condition for the evolution ofvarious meteorological parameters during this heavy rain event. Figure 3.13 showsthe three hourly vertical cross sections (Time-pressure) of vorticity (Fig. 3.13a, b)and vertical velocity (Fig. 3.13c, d) obtained from CNTL and 3DV simulationstarting from 0000 UTC 26 to 06 UTC 27 July 2005. Vertical cross section ofvorticity obtained from 3DV simulation (Fig. 3.13b) exhibits structure of strongcyclonic vorticity (30–36 � 10�5 s�1) at low and upper level. The feature is notnoticed in the CNTL simulation (maximum vorticity is 21 � 10–5 s�1 only at upperlevel). In the 3DV simulation, the positive vorticity increased gradually fromdeveloping stage to mature stage and then start decreasing after 1800 UTC.Fig. 3.12 (a) IMD observed rainfall (mm) and 24 h accumulated rainfall (cm) for (b) TRMM(c) CNTL and (d) 3DV valid at 0300 UTC of 27 July 200552 U.C. Mohanty et al.The maximum cyclonic vorticity is found during the mature stage and extended upto upper atmosphere. This shows the gradual growth of the convective systemwhich is well represented in the 3DV simulation. The strong negative vorticity(12–24 � 10�5 s�1) at upper atmosphere around 200 hPa pressure level and abovethrough out this period. The maximum upward velocity is 2.2 ms�1 obtained from3DV simulation (Fig. 3.13d) and found at mid-level (around 400 hPa) and upperatmosphere (around 200 hPa) during the mature stage (1200–1800 UTC). However,the maximum vertical velocity from CNTL simulation is 0.8 ms�1 at upperatmosphere on 0000 UTC 27 July 2005. Similarly, the vertical cross section ofdivergence field (figures not provided) obtained from the 3DV simulation showsstrong low-level convergence and upper-level divergence through out the period.The convergence is found from 0600 to 1800 UTC up to 600 hPa level withmaximum value of 20 � 10�5 s�1 at 1500 UTC. It can be inferred that the maxi-mum convergence and vorticity precede and maximum vertical velocity follows themature stage.Vorticity (x10-5s-1)Pressure (hPa)Vertical velocity (x10-5s-1)Pressure (hPa)Time Timea bc dFig. 3.13 (a–d) Time-pressure cross section of vorticity (10�5 s�1) at Santacruz (19.11�N;72.85�E) for (a) CNTL and (b) 3DV. Similarly, (c and d) are same as (a and b) but for verticalvelocity (m/s)3 Simulation of Heavy Rainfall in Association with Extreme Weather Events 533.5.2 Monsoon DepressionsFrom the above case study, it is clear that the 3DVAR experiment improve theprediction of rainfall with the assimilation of conventional and non-conventionaldata. An attempt is also made here to study the impact of DWR data along with theGTS observation on the simulation of movement of monsoon depressions andassociated heavy rainfall. In this purpose, three different numerical experimentsare carried out such as CNTL (no assimilation), 3DV_GTS (only GTS observation)and 3DV_DWR (GTS and DWR). The model is integrated 54 h in all experimentsfrom 0000 UTC of 2 August 2006 (case-1) and 21 June 2007 (case-2).From Fig. 3.14, the 24 h accumulated rainfall for case-1 is well represented theintensity (150–250 mm) as well as spatial distribution i.e. maximum rainfall overthe land and wide spread rainfall over oceanic region by the 3DV_DWR simulation(Fig. 3.14d). However, the TRMM (Fig. 3.14a) show maximum rainfall overoceanic regions. The Kalpana-1 satellite cloud image shows intense and wide spreadconvective clouds over the land as well as oceanic regions (figure not shown) whichshow good consensus with the rainfall simulated by the 3DV_DWR. Similarly, incase-2 (Fig. 3.14e–h), the spatial and temporal distribution of simulated rainfall from3DV_DWR experiment (Fig. 3.14h) indicates that assimilation of DWR data resultedin a good simulation of rainfall during day-1. The CNTL experiment underestimatesthe rainfall (by 50–100 mm) over land. Another typical characteristic of the MD(Sikka andGadgil 1980) is that heavier rainfalls are chiefly concentrated in thewestern and south western sectors of the depression and the 24 h rainfall generallyranges between 250 and 350 mm. These features are well resolved in the 3DV_DWRruns for both the cases. It is also noticed that the amount and spatial distribution ofrainfall is improved in the 3DV_GTS simulations as compared to CNTL.The spatial (Lat. 15�–25�N; Long. 75�–90�E) correlation co-efficient (CC) andRMSE of rainfall between TRMM and the model outputs are calculated over theland (mask out the ocean part) for day-1. The RMSE values are found less in3DV_DWR simulation (30.35 in case-1 and 22.79 in case-2) as compared with the3DV_GTS (34.85 in case-1 and 37.40 in case-2) and CNTL (40.46 in case-1 and41.37 in case-2) simulations. Similarly, the CC is improved in the 3DV_DWR (0.74in case-1 and 0.77 in case-2) as compared to 3DV_GTS (0.56 in case-1 and 0.51 incase-2) and CNTL (0.32 in case-1 and 0.20 in case-2) experiments. This studyclearly depicts the capability of DWR radial velocity and reflectivity in the predic-tion of quantitative prediction of rainfall. Figure 3.15a, b represents the modelsimulated track of two MDs from all the experiments along with observed track.The mean vector displacement errors (VDEs) with 6-h interval are shown inFig. 3.15c. It is observed that (Fig. 3.15c) the mean initial position error is leastin the case of 3DV_DWR (60 km). It is well known that the reduction in initialposition error causes better track prediction (Holland 1984). Therefore, the track ofthe systems is improved in 3DV_DWR experiment with least VDEs. However, thetrack of the MDs is also improved in the 3DV_GTS simulation as compared toCNTL simulated track. The 24-h (48 h) forecast errors (in kms) are 127 (93) in54 U.C. Mohanty et al.Fig. 3.14 (a–h) 24 h accumulated precipitation (cm) for day-1 (a) TRMM (b) CNTL(c) 3DV_GTS and (d) 3DV_DWR valid at 0300 UTC of 03 August 2006. Similarly, (e–h) aresame as (a–d) respectively but for case-2 valid at 0300 UTC of 22 June 20073 Simulation of Heavy Rainfall in Association with Extreme Weather Events 55 Case-1 (2-4 August 2006) Case-2 (21-23 June 2007) Mean VDEs194168170249 274 317 35649230912598 115 153223 26124534820961 52751171271401062139301002003004005006000 6 12 18 24 30 36 42 48Forecast HourMean VDEs (km)CNTL3DV_GTS3DV_DWRabcFig. 3.15 (a–c) Hourly track (observed and simulated) and VDEs from CNTL; 3DV_GTS and3DV_DWR for (a) track for case-1 (b) same as (a) but for case-2 (c) mean VDEs for case-1 andcase-256 U.C. Mohanty et al.3DV_DWR while for 3DV_GTS and CNTL the errors are 223 (209) and 274 (309)kms respectively. Overall results show that the forecast errors are reduced signifi-cantly in assimilation experiments particularly with DWR data as compared withthe other experiments.3.6 ConclusionsThe model simulation studies of severe thunderstorm, tropical cyclone, heavyrainfall and monsoon depression lead to the following broad conclusions:The WRF-NMM model simulated meteorological parameters such as relativehumidity, rainfall and temperature are consistent with each other and all are ingood agreement with the observation even though 1 h time lag exists. From themodel simulated spatial plots of composite radar reflectivity, we can clearly see thesquall line movement as in DWR images. The model simulated thermodynamicderivatives of stability indices are good enough with the values indicating a deeperlayer around Kolkata favorable for intense convective activity. Thus the dynamicand thermo-dynamic properties of the atmosphere are well captured byWRF-NMMfor the occurrence of a severe thunderstorm over Kolkata on 21 May 2007 at 1000UTC, and agree reasonably well with the observationsThe WRF-NMM model also could simulate most of the features of the cyclonesMala and Gonu with reasonable accuracy. The intensity of the tropical cyclones interms of MSLP and maximum sustainable wind illustrates that the modelunderestimates the intensity of the storm. The pattern of distribution of precipita-tion is reasonably well predicted for both the cyclone cases. The track forecast withWRF-NMM is found to be reasonably accurate over this part of the world. Themean vector displacement error clearly demonstrates the forecast skill of the WRF-NMM in terms of track prediction. However, intensity of cyclone need furtherresearch for better performance.The WRF-ARW model with improved initial condition is able to simulate withreasonably good accuracy the amount, intensity, timing and spatial distribution ofthis unusual rain event as compared to CNTL simulation. From the 3DV simulation,the analyses of the dynamical parameters at the location of heavy precipitationrevealed that the maximum convergence and vorticity precede the mature stage,however the maximum vertical velocity follows it. The CNTL simulation failed torepresent these types of features during the simulation period.The northwestward movement and location of the depression are wellrepresented in the assimilation experiments mainly in the 3DV_DWR and agreewith the IMD observations. The model precipitation fields compared well withthe TRMM rainfall estimates and the Kalpana-1 visible satellite convection imag-ery which also reflected in the statistical skill scores. The MDs track is also wellsimulated by the 3DV_DWR experiment, reasonably match with the IMD observedtrack. The VDEs are significantly improved in the assimilation experiments ascompared to CNTL simulation, however, the values are found less in the DWR3 Simulation of Heavy Rainfall in Association with Extreme Weather Events 57assimilation experiment for both the cases. The intensity and structure of the MDs isthus better simulated with the DWR assimilation.Overall, WRF modeling systems are able to broadly reproduce several featuresof these intense convective activities leading to extreme weather events and heavyrainfall in tropics. However, more realistic initial conditions with advanced dataassimilation techniques of observational data from various platforms is required forthe better prediction of these extreme weather events with state-of-the-art meso-scale models.ReferencesBarker DM, Huang W, Guo YR, Xiao Q (2004) A three-dimensional variational (3DVAR) dataassimilation system for use with MM5: implementation and initial results. Mon Weather Rev132:897–914Bell GD, Chelliah M (2006) Leading tropical modes associated with interannual and multidecadalfluctuations in north Atlantic hurricane activity. J Climate 19:590–612Daley Roger (1991) Atmospheric data analysis, Cambridge atmospheric and space science series.Cambridge University Press, Cambridge, p 457Dash SK, Kulkarni MA, Mohanty UC, Prasad K (2009) Changes in the characteristics of rainevents in India. J Geophys Res 114:12Eliot J (1899) Hailstorm in India during the period 1883–1897 with a discussion on their distribu-tion. I Met Mem 6:4Holland GJ (1984) Tropical cyclones in the Australian/Southwest Pacific region. II: Hurricane.Aust Meteor Mag 32:17–33Holland GJ, Webster PJ (2007) Heightened tropical cyclone activity in the North Atlantic: naturalvariability or climate trend? Philos Trans R Soc A 365:2605–2716Janjic ZI (1984) Non–linear advection schemes and energy cascade on semi–staggered grids. MonWeather Rev 112:1234–1245Janjic ZI (2003) A nonhydrostatic model based on a new approach. Meteorol Atmos Phys82:271–285Jenamani R, Bhan SC, Kalsi SR (2006) Observational/forecasting aspects of the meteorologicalevent that caused a record highest rainfall in Mumbai. Curr Sci 90:1344–1362Kain JS, Weiss SJ, Levit JJ, Baldwin ME, Bright DR (2006) Examination of convection-allowingconfigurations of the WRF model for the prediction ofsevere convective weather: the SPC/NSSL spring program 2004. Weather Forecast 21(2):167Kumar V, Chandrasekar A, Alapaty K, Niyogi D (2007) The impact of assimilating soil moisture,surface temperature, and humidity and the traditional four dimensional data assimilation on thesimulation of a monsoon depression over India using a mesoscale model. J Appl MeteorolClimatol 47:1393–1412Litta AJ, Mohanty UC (2008) Simulation of a severe thunderstorm event during the field experi-ment of STORM programme 2006, using WRF-NMM model. Curr Sci 95(2):204–215Litta AJ, Mohanty UC, Bhan SC (2009) Numerical simulation of a Tornado over Ludhiana (India)using WRF-NMM model. Meteorol Appl 17:64–75Mohanty UC, Sikka DR, Madan OP, Pareek RS, Kiran Prasad S, Litta AJ et al (2007) Weathersummary pilot experiment of Severe Thunderstorms-Observational and Regional Modeling(STORM) Programme – 2007, KharagpurNizamuddin S (1993) Hail occurrences in India. Weather 48:90–92Orlanski I (1975) A rational subdivision of scales for atmospheric processes. Bull Am MeteorolSoc 56:527–53058 U.C. Mohanty et al.Parthasarathy B (1984) Interannual and long term variability of Indian summer monsoon rainfall.Proc Indian Acad Sci Earth Planet Sci 93:371–385Pattanayak S, Mohanty UC (2008) A comparative study on performance of MM5 and WRFmodels in simulation of tropical cyclones over Indian seas. Curr Sci 95(7):923–936Pielke RA Sr, Matsui T, Leoncini G, Nobis T, Nair U, Lu E, Eastman J, Kumar S, Peters-Lidard C,Tian Y, Walko R (2006) A new paradigm for parameterizations in numerical weather predic-tion and other atmospheric models. Natl Weather Dig 30:93–99Routray A, Mohanty UC, Das AK, Sam NV (2005) Study of heavy rainfall event over the west-coast of India using analysis nudging in MM5 during ARMEX-I. Mausam 56:107–120Sikka DR, Gadgil S (1980) On the maximum cloud zone and ITCZ over the Indian longitudesduring the southwest monsoon. Mon Weather Rev 108:1122–1135Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Wang W, Powers JG (2005)A description of the advanced research WRF version 2, NCAR tech note, NCAR/TN–468 +STR, 88 ppWeiss SJ, Kain JS, Bright DR, Levit JJ, Carbin GW, Pyle ME, Janjic ZI, Ferrier BS, Du J,Weisman ML, Xue M (2007) The NOAAHazardous weather testbed: collaborative testing ofensemble and convection-allowing WRF modelsand subsequent transfer to operations at theStorm Prediction Center. 22nd Conference Weather Analysis Forecasting/18th Conference onNumerical Weather Prediction, Salt Lake City, Utah, American Meteorological Society,CDROM 6B.43 Simulation of Heavy Rainfall in Association with Extreme Weather Events 59.Chapter 4Representation of Uncertainties in SeasonalMonsoon Predictions Using a GlobalClimate ModelSarat C. KarAbstract A seasonal prediction system is being developed at NCMRWF underthe Seasonal Prediction and Application to Society (SeaPrAS) programme. Ensem-ble integrations have been carried out using the Indian global Climate model(In-GLM1) for 23 monsoon seasons with observed, climatological and persistedsea surface temperature forcing. It is found that the model simulated climatology isreasonably good, however, inter-member ensemble spread of rainfall is quite largeover the Indian monsoon region. Examination revealed that the inter-membervariance is not purely due to internal dynamics. For making seasonal predictions,these uncertainties and systematic errors have to be represented. There are alsolarge uncertainties in the SST predictions from various models and a methodhas been developed to incorporate these uncertainties while making seasonalpredictions using prescribed SSTs. After establishing the usability of the modelfor seasonal predictions, the prediction system is being tested for real time formonsoon seasons since 2006.4.1 IntroductionRainfall during the Indian summer monsoon season (June, July August andSeptember) show considerable interannual and intraseasonal variability. Predictionof the Indian summer monsoon both in interannual time scale and intraseasonaltimescale are quite important for the economy of the subcontinent as the Indianeconomy is largely dependent on agriculture. Prediction of the seasonal meanmonsoon at least one season in advance is one of the most important and challeng-ing problems in tropical climate. The seasonal mean monsoon circulation in theS.C. Kar (*)National Centre for Medium Range Weather Forecasting, Ministry of Earth Sciences, A-50,Sector-62, Noida, UP, Indiae-mail: sckar@ncmrwf.gov.inS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_4, # Springer-Verlag Berlin Heidelberg 201161tropics is potentially more predictable. This is because the low-frequency compo-nent of the tropical variability is primarily forced by slowly varying boundaryforcing, which evolves on a slower time scale than that of the weather systemsthemselves. Influence of the El Nino Southern Oscillation (ENSO) on the Indianmonsoon rainfall has been investigated (Sikka 1980; Shukla 1987). Goswami(1998) had examined interannual variation of Indian summer monsoon in a GCMand had evaluated the role of external conditions versus internal feedbacks. Karet al. (2001) had examined long-term simulations of a global model with observedand climatological SST. Goswami and Xavier (2005) have examined the dynamicsof internally generated interannual variability of Indian summer monsoon in aGCM. Kang et al. (2004) have estimated the potential predictability of summermonsoon rainfall in a dynamical seasonal prediction system with systematic errorcorrection. NCMRWF, over the years, has successfully developed and made oper-ational a deterministic medium-range weather forecasting system. There has beendemand from several user agencies for extended range prediction (monthly andseasonal time-scales). Whereas efforts have been made to improve operationalmonsoon prediction at IMD using statistical methods, dynamical extended-rangeprediction efforts have continued at several organizations in India including that atNCMRWF.In a stand-alone atmospheric model, various land surface processes are coupledwith atmosphere through the land-surface process parameterization schemes, butthe Sea Surface Temperatures (SSTs) are prescribed as external conditions. Largeuncertainties exist in the forecasted SSTs from other SST prediction models.Therefore, it is essential to incorporate these uncertainties in the seasonal predictionmodels when real-time forecast runs are made in seasonal timescales. It is alsoimportant to understand the structure and behaviour of the internal variability,relate it to the systematic errors of the model, and represent these uncertainties inthe final seasonal predictions. Main objective of this report is to document thestrength and weaknesses of the two-tire system (a global model forced withprescribed SSTs) used in the seasonal prediction system. Section 4.2 of this reportdescribes the seasonal prediction system, the model and data used. In Sect. 4.3,results are presented with discussion. Section 4.4 concludes the report.4.2 The Seasonal Prediction SystemSeasonal Prediction & Application to Society (SeaPrAS) programme at NCMRWFhas been initiated for preparing seasonal predictions from users’ point of view. Theobjective is to improve the capacity in India’s resource management to cope withthe impacts of climate variability. SeaPrAS shall be a platform for policymakers &resources managers to have access to, and make use of, information generated byclimate prediction models. It is expected that the SeaPrAS programme shall providethe planners with more reliable seasonal climate prediction information and guid-ance on who could be the potential beneficiaries of the predictions. Associated62 S.C. Karapplication systems shall alsobe developed for energy demand, water resourcemanagement, agriculture- drought prediction, crop yield. Work is in Progresstowards this end.The global model used for this study is the Indian global model (In-GLM1). Thisglobal climate model is the climate version of the NCMRWF’s medium-rangeweather forecast model (older version). Several modifications were made to uti-lize the model for seasonal prediction and to study the seasonal to interannualvariability of the climate system. More details of the model may be found atKanamitsu et al. (1991), Kar (2002, 2007). The model was initialized with theNCEP-II reanalysis data (ftp.nomad3.ncep.noaa.gov) valid for April 15, 16, 17, 18,19, 20, May 01, May 05, May 10 and May 15 of each year and seasonal simulationswere carried out. All other surface characteristics at the initial stage of the modelwere set to climatology except for sea surface temperature (SST). The land surfaceproperties such as wetness, temperature, snow, ice cover etc. evolve along withmodel simulations. For studying the monsoon variability, the model is forced withweekly optimum interpolation (OI) sea surface temperature (SST) data, Reynoldsand Smith (1994). For prediction purpose, forecasted SSTs are downloaded fromother Centres. These are described later in the report.4.3 Results and DiscussionThe model climatology for seasonal rainfall has been compared against the CMAPrainfall climatology for the corresponding period (1982–2004). The model’s rain-fall climate is reasonably good and all the essential features of rainfall pattern overthe Indian region are simulated well by the model (figure not shown). The over-estimation of rainfall by the model over the Arabian Sea and under-estimation ofrainfall over the Bay of Bengal and the Gangetic plain is evident. Maximumunderestimation in rainfall simulations is seen over the central Bay of Bengal asa zonal stretch around 10 N extending up to Western Pacific Ocean. Year to yearvariation of rainfall during the monsoon season over the Indian region (areaaveraged) from model simulations with observed SST and observation has beenexamined. An examination of the simulated and observed rainfall for these yearsover the Indian region suggests that there are some years, where simulations arenot correct. However, some years such as 1988, 1987, 1994, 1997, 2002 the modelhindcast simulations agree well with observation. Examination of temporal corre-lation coefficients of observed rainfall with the ensemble mean simulated rainfallwith observed SST forcing shows that over India the correlation values are verylow, only a small pocket has the value greater than 03 or 0.4. More about climato-logical patterns and interannual variability are provided in Kar (2007).In order to examine the systematic error structure of the model simulated fields,the seasonal ensemble mean rainfall, zonal wind (U) and meridional wind (V) havebeen compared with CMAP rainfall and NCEP-II reanalysis datasets. The system-atic error variance of these fields for the 23 monsoon seasons are shown in Fig. 4.1.4 Representation of Uncertainties in Seasonal Monsoon Predictions 63In Fig. 4.1a, error variance of rainfall shows that model has large systematic errorover the Indian Ocean to the west of Indonesia, centred around 10 S, 90E. Anotherzone of large error is off the south-west coast of India. The maximum error is alongthe Arakkan coast and north Bay of Bengal. There are also large errors in Westernpacific. In the tropics, the latent heat release due to condensation shall drives thewind circulation. Therefore, in this region, the wind fields and rainfall are veryFig. 4.1 Systematic errorvariance from modelsimulations (a) rainfall,(b) zonal wind (u) and(c) meridional wind64 S.C. Karclosely related. In the zonal wind systematic error variance plot [Fig. 4.1b], it isseen that there is a zonal band of zonal wind error along 12 N extending from theArabian Sea to the Bay of Bengal and dipping southward extending to the WestPacific. In the equatorial Indian Ocean too, there is a large zonal wind error. Thesystematic error variance for the meridional winds are shown in Fig. 4.1c. A band ofwind error extends from the equatorial Indian Ocean northward to the peninsularIndia.Following Kang et al. (2004), the externally forced interannual variance forseasonal mean rainfall have been computed, as shown in Fig. 4.2a. It is seen that inthe Indian region, a large part of the rainfall variability in interannual timescalesforced by SST occur over the Indian seas. Two bands of maxima in the variabilityoccur where there are rainfall maxima. Over the Pacific, especially central andeastern Pacific, the SST forced variability are also large (not shown in figure). DueFig. 4.2 (a) Externally forced interannual variance for model simulated rainfall, and (b) inter-member rainfall variance4 Representation of Uncertainties in Seasonal Monsoon Predictions 65to local effect of interannually varying SST, there are some variability of rainfall inthe Indian Ocean south of equator. As already mentioned, for this study, 10 memberensemble runs were made for each monsoon season. The inter-member variance isused by many authors (e.g. Kang et al. 2004) to describe the variability caused byinternal dynamics. The inter-member variance in the present model simulations forrainfall are shown in Fig. 4.2b. It is seen that the variance is quite large in the Indianregion, with magnitudes almost comparable to the SST forced variance. However,there are major differences in shape and orientation of the inter-member variance.The pattern of inter-member variance is almost similar to the systematic errorvariance. Therefore, it may be said that the model systematic errors define theinter-member spread.In Fig. 4.3a, the ensemble spread over the Indian monsoon region is plottedagainst the model simulated ensemble mean rainfall climatology. Ideally, if theensemble spread is caused by uncertainties in dealing with physical processes inthe model, the spread should reflect this aspect. In the tropical region, convection isthe most important physical process apart from radiation. Convective heating andmoistening define the tropical circulation. There are several deficiencies in themanner in which the convection processes are included in a global model. In thefigure it is seen that as the rainfall amount increases, the spread also increases.When the rainfall amount is low, the spread is also less. This shows that there shallbe large uncertainty in predictions of flood monsoon years as compared to droughtmonsoon years. To examine if the inter-member spread is modulated by the SST, ithas been correlated with the observed SST used in the model simulations. It is seenthat indeed, the inter-member spread is correlated to large scale interannualvariability of SSTs, and at many regions, the correlation coefficients are statisticallysignificant. Therefore, it is concluded that the inter-member spread is not entirelydue to internal dynamics, rather it gets modulated by interannually varying SSTforcing. To estimate the internal variability generated by this model, the model wasintegrated for 23 monsoon seasons with climatological SST. The internal variancecomputed from runs with climatological SST runs is shown in Fig. 4.3b. It is seenthat the internal variance is quite large in the Indian region and is similar tothe systematic error variance for rainfall. Over large part of the country, rainfallvariability in interannual timescale is due to internal dynamics in these modelsimulations.At this point, it is worthwhile to examine how in the model ensemble spreadincreases as forecast length increases. For this purpose, ensemble forecasts up to6-days have been carried out for each day of a monsoon season. Ensemble spreadhas been computed for each dayof R&D inatmospheric, oceanic and allied sciences, sponsored research in meteorology,publication of its biennial Journal – Vayu Mandal (since 1970), IMS News, Newsletters, scientific books etc. It also organizes annual series of national conferencenamed “TROPMET” since 1992 supplemented with international conference called“INTROMET” every 4 year. Awareness programmes about meteorology and alliedsciences are regularly organized in the country. It is co-founder of InternationalForum of Meteorological Societies. It also felicitates the outstanding scientists byconferring on them the Fellowships and has constituted five national and oneinternational awards in the field of meteorology and atmospheric sciences.The details of the Society are available on http://www.indianmetsoc.com.National Council of IMS (2009–2011)PresidentDr L S RathoreTel: 91-11-24619844lrathore@gmail.comVice-PresidentsDr. R. K. DattaTel : 09811213169rkdatta_in@yahoo.comDr. S. D. AttriTel: 91-11- 24620701sdattri@gmail.comSecretaryShri. D. K. MalikTel.: 09810618585dk_malik@hotmail.comJt. SecretaryDr. D. R. PattanaikTel:09868554029pattanaik_dr@vyahoo.co.in(continued)Treasurer:Sh. Virendra SinghTel: 09899213832vsvsingh69@gmail.comImmediate Past PresidentSh. R. C. Bhatiarcbhatia1912@gmail.comCouncil Members:Dr V U M RaoDr (Mrs) Parvinder MainiDr K K SinghDr. D. P. DubeySh. B. P. YadavSh N. NigamDr. M. RavichandranSh. Utpal Bhattacherjeex PrefaceContents1 Modernization of Observation and Forecasting System in IMDin Support of Agromet Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Ajit Tyagi2 Monthly and Seasonal Indian Summer Monsoon Simulatedby RegCM3 at High Resolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13S.K. Dash, Savita Rai, U.C. Mohanty, and S.K. Panda3 Simulation of Heavy Rainfall in Association with ExtremeWeather Events: Impact on Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35U.C. Mohanty, S. Pattanayak, A.J. Litta, A. Routray,and O. Krishna Kishore4 Representation of Uncertainties in Seasonal Monsoon PredictionsUsing a Global Climate Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61Sarat C. Kar5 Intra Seasonal Variability of Rainfall in India on Regional Basis . . . 73Manish K. Joshi, K.C. Tripathi, Avinash C. Pandey, and I.M.L. Das6 Assimilation of Surface Observations in a High ResolutionWRF Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83Dipak K. Sahu and S.K. Dash7 An Evaluation of the Simulation of Monthly to Seasonal SummerMonsoon Rainfall over India with a Coupled Ocean AtmosphereGeneral Circulation Model (GloSea) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101D.R. Pattanaik, Ajit Tyagi, U.C. Mohanty, and Anca Brookshawxi8 Prediction of Monsoon Variability and Subsequent AgriculturalProduction During El Niño/La Niña Periods . . . . . . . . . . . . . . . . . . . . . . . . . 123M.V. Subrahmanyam, T. Satyanarayana, and K.P.R. Vittal Murthy9 Improved Seasonal Predictability Skill of the DEMETERModels for Central Indian Summer Monsoon Rainfall . . . . . . . . . . . . . . 139Ravi P. Shukla, K.C. Tripathi, Sandipan Mukherjee,Avinash C. Pandey, and I.M.L. Das10 Simulation of Indian Summer Monsoon Circulation with RegionalClimate Model for ENSO and Drought Years over India . . . . . . . . . . . 149Sandipan Mukherjee, Ravi P. Shukla, and Avinash C. Pandey11 Changes in Surface Temperature and Snow over the WesternHimalaya Under Doubling of Carbon Dioxide (CO2) . . . . . . . . . . . . . . . . 163P. Parth Sarthi, S.K. Dash, and Ashu Mamgain12 Simulation of Tornadoes over India Using WRF-NMM Model . . . . . 173A.J. Litta, U.C. Mohanty, S.C. Bhan, and M. Mohapatra13 A Pilot Study on the Energetics Aspects of Stagnationin the Advance of Southwest Monsoon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187Somenath Dutta and Lt. Vishwarajashree14 Integrated Agrometeorological Advisory Services in India . . . . . . . . . 195L.S. Rathore, S.K. Roy Bhowmik, and N. Chattopadhyay15 South-West Monsoon Variability and Its Impact on DrylandProductivity in Drought Affected Districts of Amravati Divisionin Maharashtra State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207G.U. Satpute and S.S. Vanjari16 Simulation of Growth and Yield Attributes of Wheat GenotypesUnder Changing Climate in Recent Years in India . . . . . . . . . . . . . . . . . . 221S.D. Attri, K.K. Singh, and R.K. Mall17 Strategies for Minimizing Crop Loss due to Pest and DiseaseIncidences by Adoption of Weather-Based PlantProtection Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235N. Chattopadhyay, R.P. Samui, and L.S. Rathore18 Climate-Based Decision Support Tools for Agriculture . . . . . . . . . . . . . 245Mark S. Brooks, Aaron P. Sims, Ashley N. Frazier, Ryan P. Boyles,Ameenulla Syed, and Sethu Ramanxii Contents19 Challenges in District Level Weather Forecastingfor Tribal Region of Chhattisgarh State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257J.L. Chaudhary20 Agromet Information System for Farm Management . . . . . . . . . . . . . . . 263M.C. Varshneya, N. Kale, V.B. Vaidya, Vyas Pandey, and B.I. Karande21 Advanced INSAT Data Utilization for Meteorological Forecastingand Agrometeorological Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273P.C. Joshi, B. Simon, and B.K. Bhattacharya22 Data Mining: A Tool in Support of Analysis of Rainfall on Spatialand Temporal Scale Associated with Low Pressure SystemMovement over Indian Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287Kavita Pabreja23 Information Systems as a Tool in Operational Agrometeorology:Applications to Irrigation Water Management in EmiliaRomagna-Italy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299Federica Rossi24 Impact of Climate Change on Crop Water Requirementsand Adaptation Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311V.U.M. Rao, A.V.M. S. Rao, G.G.S.N. Rao, T. Satyanarayana,N. Manikandan, and B. Venkateshwarlu25 Climate Change and Its Impact on Wheat and Maize Yieldin Gujarat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321Vyas Pandey and H.R. Patel26 Climate Change Adaptation and Mitigation for Drought-ProneAreas in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335R.P. Samui and M.V. Kamble27 Climate Change in Relation with Productivity of Rice and Wheatin Tarai Region of Uttarakhand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355H.S. Kushwaha28 Estimation of Wheat Productivity Under Changing Climatein Plains Zones of Chhattisgarh Using Crop Simulation Model . . . . 369S.R. Patel, S. Tabasum, A.S. Nain, R. Singh, and A.S.R.A.S. SastriContents xiii29 Impact of Climate Change on the Grape Productivityin the Southern Coast of the Crimea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385S. Korsakova30 The Impact of Extreme Weather Events on Agriculturein the United States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397Raymond P. Motha31 Inter-Annual Variation of Fog, Mist, Haze and Smoke at Amritsarand Its Impact on Agriculturalof forecast as the root mean square of deviationsof all the ensemble members from the ensemble mean of the forecasted day.Figure 4.4 shows the ensemble spread (mm/day) for Day-1, Day-2, Day-4 andDay-6 forecasts. It is interesting to note that the distribution and amount of spreadis different for different length of forecasts. In Day-1 forecasts, the spread overmost of the Indian region is between 2 and 4 mm/day. Over the west coast, andsurrounding oceanic region, the spread is about 4 mm/day. Over parts of Gujaratand Rajasthan including parts of Uttar Pradesh, the spread is less than 2 mm/day.66 S.C. KarSpread is quite large (8–12 mm/day) over Thailand and surrounding countries.In Day-2 and Day-4 forecasts, the spread over the Indian region has increased.The west coast, the eastern region and the Bay of Bengal have now a spread ofFig. 4.3 (a and b) Scatter diagram of ensemble mean climatology of rainfall and ensemble spread,(b) internal variance of simulated rainfall obtained from runs with climatological SST4 Representation of Uncertainties in Seasonal Monsoon Predictions 676 mm/day. Near the equatorial Indian Ocean region, the spread has reduced to lessthan 2 mm/day. By Day-6, the spread over most parts of the Indian region has furtherincreased to 6 mm/day or more. There are pockets where spread is 8 mm/day andincreasing to 10 mm/day. Over the equatorial Indian Ocean, the spread is furtherreduced. The model has a systematic bias and the ensemble spread is clearlydemonstrating this bias. As the forecast length is further increased up to seasonaltimescales, the ensemble spread start to display variations within a season. It is seenthat the spread is more or less large over the Indian region, if the monsoon is active.The magnitude of spread is less if rainfall activity is subdued over the Indian region.4.3.1 Real-Time Monsoon Prediction SystemThe main problem for real time prediction with an atmospheric model is to have theprediction of boundary conditions like SST, which influences the monthly andseasonal mean anomalies. Since monsoon 2007, a set of forecasts of SST anomaliesFig. 4.4 Ensemble spread (mm/day) for Day-1, Day-2, Day-4 and Day-6 forecasts. Contour levelsare 2, 4, 6, 8, 10 mm/day. Regions with spread more than 2 mm/day are shaded68 S.C. Karobtained from IRI, USA are being used in the seasonal prediction system. In thisreport, we describe real-time predictions made for 2008 monsoon. During pre-monsoon months of 2008, moderate La Nina conditions were prevailing over theequatorial Pacific Ocean. Several ocean models and coupled ocean atmospheremodels were indicating that during monsoon months, La Nina conditions shallweaken further and ENSO neutral conditions shall prevail. However, there waslarge variations in the predictions of SST by different models. It was consideredappropriate to represent this uncertainty in the SST predictions into the globalmodel used for seasonal predictions. Three scenarios of SST for summer monthsfor 2008 were received from IRI. The seasonal (JJAS) mean anomalies of SST areshown in Fig. 4.5a. On to this anomaly, a perturbation was added and subtracted togenerate other two SST scenarios. This perturbation represents the uncertainties inSST predictions (Fig. 4.5b).Using these SST anomalies, the IN-GLM1 model was integrated for preparingmonsoon rainfall forecasts for 2008. For each SST scenario, 6 member ensembleruns were made using observed initial conditions of April 6–11, 2008. Therefore,there were 18 members in total. The NCMRWF real-time global analysis data wereused as initial conditions for the IN-GLM1 model. Rainfall anomalies (mm/day)from each member run are shown in Fig. 4.6. It I seen that the model responds tothese three SST scenarios, and over the Indian monsoon region, the model providesdifferent scenarios of rainfall activity. However, there are some robust signal whichdo not change much with different SST forcing nor with different initial conditions.Ensemble mean of these three sets of runs are shown in Fig. 4.7. Based on theseresults, a forecast was prepared for rainfall on April 13, 2008. Considering the biasFig. 4.5 (a and b) Seasonal mean (JJAS) anomalies of predicted SST used in model simulationsand (b) perturbation used to represent uncertainties in ST predictions4 Representation of Uncertainties in Seasonal Monsoon Predictions 69Fig. 4.6 Predicted rainfall anomalies from each ensemble member with three scenarios ofpredicted SST for 2008 monsoon seasonFig. 4.7 Ensemble mean of predicted rainfall anomalies from each of the three scenarios ofpredicted SST for 2007 monsoon season70 S.C. Karin the model as well as other strength and weaknesses, the forecast given was thatIndia shall experience below normal to normal monsoon rainfall during monsoonseason of 2008. In order to update the forecast, similar exercise was made in May2008 with model initial conditions of May and updated SST predictions. At the endof the season, IMD noted that for the country as a whole, the seasonal rainfall from1st June to 30th September was 98% of its long period average (LPA). There wereseveral deficiencies in the regional features of predicted seasonal anomalies ofrainfall. It may however be noted that skill of dynamic models in predicting seasonalmean anomalies of monsoon rainfall is low. Therefore, a probabilistic seasonalprediction system has been developed. However, to ensure reliability of predictions,more test runs are being carried out. The probabilistic scheme is being calibrated.4.4 ConclusionsConsidering the limitations of statistical methodologies utilized by the IMD forseasonal monsoon prediction, a seasonal prediction system based on globaldynamic model is being developed at NCMRWF. This development is beingmade under the Seasonal Prediction and Application to Society (SeaPrAS)programme. The Indian global Climate model (In-GLM1) has been integratedfor 23 monsoon seasons with observed, climatological and persisted sea surfacetemperature forcing. For each season, simulations have been carried out usingten ensemble members. It is found that the model simulated climatology is rea-sonably good, however, inter-member ensemble spread of rainfall is quite largeover the Indian monsoon region. Examination revealed that the interannuallyvarying SST modulates the inter-member variance and therefore, this variancemay not be purely due to internal dynamics. For making seasonal predictions,these uncertainties and systematic errors have to be represented. This observationprovides a basis for a probabilistic prediction scheme based on a probabilitydistribution function from the ensemble members. There are also large uncer-tainties in the SST predictions from various models and a method has beendeveloped to incorporate these uncertainties while making seasonal predictionsusing prescribed SSTs. It was found that the model responds to local and remoteSST variations reasonably well, after establishing the usability of the modelfor seasonal predictions, the prediction system is being tested for real timepredictions for monsoon seasons since 2006. A probabilistic prediction systemhas also been developed and being calibrated. The prediction system has to betested for large number of years in real-time to have enough confidence inpredictions.4 Representation of Uncertainties in Seasonal Monsoon Predictions 71ReferencesGoswami BN (1998) Interannual variation of Indian summer monsoon in a GCM: externalconditions versus internal feedbacks. J Climate 11:501–522Goswami BN, Xavier PK (2005) Dynamics of ‘Internal’ interannual variability of Indian summermonsoon in a GCM. J Geophys Res 110:D24104. doi:10.1029/ 2005JD006042Kanamitsu M, Alpert JC, Campana KA, Caplan PM, Deaven DG, Iredell M, Katz B, Pan HL, SelaJ, White GH (1991) Recent changes implemented into theglobal forecast system at NMC.Weather Forecast 6:425–435Kang IS, Lee J, Park CK (2004) Potential predictability of summer mean precipitation in adynamical seasonal prediction system with systematic error correction. J Climate 17:834–844Kar SC (2002) Description of a high-resolution global model (T170/L28) developed at NCMRWF.Research report 1/2002, NCMRWF, 28 ppKar SC (2007) Global model simulations of interannual variability of the Indian summer monsoonusing observed SST variability, NCMRWF research report, NMRF/RR/3/2007, 40 ppKar SC, Masato Sugi, Nobuo Sato (2001) Interannual variability of the Indian summermonsoon and internal variability in the JMA global model simulations. J Meteorol Soc Jpn79(2):607–623Reynolds RW, Smith TM (1994) Improved global sea surface temperature analyses. J Climate7:929–948Shukla J (1987) Interannual variability of monsoon. In: Fein JS, Stephens PL (eds) Monsoons.Wiley, New York, pp 399–464Sikka DR (1980) Some aspects of the large-scale fluctuations of summer monsoon rainfall overIndia in relation to fluctuations in the planetary and regional scale circulation parameters. ProcIndian Acad Sci Earth Planet Sci 89:179–19572 S.C. KarChapter 5Intra Seasonal Variability of Rainfallin India on Regional BasisManish K. Joshi, K.C. Tripathi, Avinash C. Pandey, and I.M.L. DasAbstract ISMR with its annual, seasonal and daily variability affects most ofsocial, economic and human activities throughout the Indian subcontinent. Inparticular, the drought and floods not only affect all types of agricultural productsbut is also responsible for loss of human lives and property. Late or prolongedmonsoon break can lead to catastrophic effects. The knowledge of the intra seasonaland inter annual variability of daily rainfall climatology can be used profitably forcrop production. In the present study, the 60 years long NCEP reanalysis dataof precipitation rate has been spectrally analyzed on decadal basis to find out theperiodicity in All-India, Southwest (SW) and Southeast (SE) precipitation patterns.There is not much change observed in decadal spectra over the last six decades solong as the low frequency components are concerned but the higher frequencycomponents have been found to be a gradually increasing factor in the temporalvariation on decadal scale during the last three decades. The rain pattern in the SWregion does not exhibit much variation on decadal scale during last six decadesimplying that the SW region is nearly non-evolving and follows the characteristicpattern of the region. In the SE region the strength of the signals corresponding tohigher frequency cycles keep changing from decade to decade. It is concluded thatSE region is more stable in temporal variability than the SW region.M.K. Joshi (*) • K.C. TripathiK. Banerjee Centre of Atmospheric & Ocean Studies, Institute of Interdisciplinary Studies,University of Allahabad, Allahabad 211 002, Indiae-mail: manishkumarjoshi@gmail.com; kctripathi@gmail.comA.C. PandeyDepartment of Physics, University of Allahabad, Allahabad 211 002, Indiae-mail: avinashcpandey@rediffmail.comI.M.L. DasM. N. Saha Centre of Space Studies, University of Allahabad, Allahabad 211 002, Indiae-mail: profimldas@yahoo.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_5, # Springer-Verlag Berlin Heidelberg 2011735.1 IntroductionRainfall is the end product of a number of complex atmospheric processes which varyboth in space and time (Luk et al. 2001). Knowledge of the space-time variability ofrainfall is important for meteorology, hydrology, agriculture, telecommunications,and climate research. Rainfall has been long analyzed by means of standard statisticssuch as average value, variance, coefficient of variation and percentiles (Laughlinet al. 2003).Krishnamurti and Bhalme (1976) reported the presence of the spectral peaks at10–20 day in pressure and other data. While Krishnamurti and Ardanuy (1980) usedlonger surface pressure data and observed 10–20, 20–30 and 30–40 day variability.Murakami reported the existence of 5 day and 15 day peak while viewing thespectral analysis of Indian monsoon (Murakami 1977). The spectral analysis of a70 year record of daily precipitation was performed to search for periodicities onsubseasonal time scales during the summer monsoon and reported the presence of40–50 day spectral peak corresponding to Madden Julian oscillation over mostportion of India south (Hartmann and Michelsen 1989). The seasonal variability ofspectral peaks in the 40–50 day range for winds and precipitation in the tropicalPacific and Indian Ocean region was studied (Hartmann and Gross 1988). Further,it was reported that the HIM time series is simple in structure with only the annualoscillation and its first two harmonics accounting for almost the entire variability(Rangarajan 1994). The Homogeneous Indian Monsoon (HIM) region rainfall wasanalysed for the epoch 1871–1990 using Singular Spectral Analysis (Rangarajan1994).Singular Spectrum Analysis (SSA) was applied to the Indian Summer MonsoonRainfall (ISMR) series for extracting the statistically significant oscillations withperiods 2.8 and 2.3 years from the white noise of the ISMR series (Vijayakumar andKulkarni 1995).In recent years, (Peters et al. 2002) has presented a power law behavior in thedistribution of rainfall over at least four decades. Scargle periodogram and wavelettransform methods were used, to study the periodicity of Indian Summer MonsoonRainfall (ISMR) changes between 1871 and 2004 and review the possible influenceof solar activity on the rainfall (Lihua et al. 2007). The seasonal monsoon rainfall isfound to consist of two dominant intraseasonal oscillations with periods of 45 and20 days and three seasonally persisting components, by using Multichannel SingularSpectrum Analysis (MSSA) of daily rainfall anomaly (Krishnamurthy and Shukla2007). The variability and long-term trends of extreme rainfall events over centralIndia have been examined and it was reported that inter-annual, inter-decadal andlong-term trends of extreme rainfall events are modulated by the SST variationsover the tropical Indian Ocean (Rajeevan et al. 2008).The purpose of this paper is to investigate the regular variations in rainfallover All-India and the two regions namely Southwest (SW) and Southeast (SE).A central part of the study is the spectral analysis of a 60 year record of daily rainfalldata from. The goal of this analysis is to evaluate the changes that are taking place74 M.K. Joshi et al.in the pattern in All-India as well as the two regions defined above on the decadalbasis and to observe the spectra of Daily Rainfall Climatology (DRC) of All-Indiaand the two regions to investigate the broad features in the regions.5.2 Data Used and MethodologyThe 60 years (1948–2007) NCEP reanalysis data of precipitation rate has beenspectrally analyzed on decadal scale to find out the periodicity in rainfall of All-India (66�E to 90�E and 5�N to 35�N), SW (73�E to 76�E and 11�N to 20�N) and SE(77�E to 80�E and 8�N to16�N) regions. The decadal spectra have been used toevaluate the changes that are taking place in the rain pattern and the daily rainfallclimatology has been used to investigate the broad features of the rain pattern inthese regions.If R(m,n) be the precipitation for the nth day of the mth year of the 60 yearsNCEP data, then the DRC of the nth day, Rc(n), is defined as:RcðnÞ ¼ 160X60m¼1R m; nð ÞPrior to perform spectral analysis, the high frequency fluctuations in the DRChave been removed by applying a 5-day running mean to obtain more accurateperiodicities of rainfall.5.3 Results and DiscussionFigure 5.1 shows the decadal spectra of precipitation over All-India for the period(a) 1948–1957, (b) 1958–1967,(c) 1968–1977, (d) 1978–1987, (e) 1988–1997 and(f) 1998–2007 respectively. There is not much change observed in decadal spectraof All-India over the last six decades so long as the low frequency components areconcerned. In contrast, the higher frequency components, the 80th or other nearbyharmonics corresponding to a cycles of 40–50 days (usually called the MJ oscilla-tion), has been found to be a gradually increasing factor in the temporal variationon decadal scale during the last three decades. The gradually increasing energy ofthese signals can be seen from Fig. 5.1d–f.Figure 5.2 shows the decadal spectra of precipitation over SE for the period(a) 1948–1957, (b) 1958–1967, (c) 1968–1977, (d) 1978–1987, (e) 1988–1997 and(f) 1998–2007 respectively. Apart from the annual and other components presentin the All-India rainfall pattern, the harmonics from 150 to 270 are also found tobe important in SE region. The dominant cycles between the harmonics 150–270are variable on inter decadal basis. Some signals appear in some decades in an5 Intra Seasonal Variability of Rainfall in India on Regional Basis 7550 100 150 200 250 300 35000.511.522.53Harmonic no.Spectral Density or Power365 day182.3 day122 day47 day49 day50 100 150 200 250 300 35000.511.522.53Harmonic no.Spectral Density or Power365 day182.3 day122 day46 daya b50 100 150 200 250 300 35000.511.522.53Harmonic no.Spectral Density or Power365 day182.3 day122 day49 day44.5 day40.1 day50 100 150 200 250 300 35000.511.522.53Harmonic no.Spectral Density or Power365 day182.3 day122 day45.6 day43.4 day47.4 dayc d50 100 150 200 250 300 3500.511.522.53Harmonic no.Spectral Density or Power365 day182.3 day122 day41 day42 day49.3 day50 100 150 200 250 300 3500.511.522.5No. of HarmonicsSpectral Density or Power365 day182.3 day122 day41 day45.6 day49.9 daye fFig. 5.1 Discrete Fourier transform of precipitation for all-India region over the period (a)1948–1957, (b) 1958–1967, (c) 1968–1977, (d) 1978–1987, (e) 1988–1997, (f) 1998–200776 M.K. Joshi et al.unpredictable way like the 277th harmonic in the 1988–1997 periods which is notpresent in any other decade. The second last and the last decade contains animportant harmonic of 356 and 314 as shown in Fig. 5.2e, f respectively, indicatingthat the pattern of rainfall in the SE region is ‘evolving’ on a decadal basis.Figure 5.3 shows the decadal spectra of precipitation over SW for the period(a) 1948–1957, (b) 1958–1967, (c) 1968–1977, (d) 1978–1987, (e) 1988–1997 and(f) 1998–2007 respectively. As far as low frequency components are considered,the rainfall pattern in the SW region does not exhibit much variation on decadalscale during last six decades. The spectrum of all decades for SW region containsthe annual cycle of 365 day, semiannual cycle of 182.3 day, terannual cycle of122 day and the quadannual cycle of 91 day. In addition to this, the other prominentcycles present in the first decade, are quite similar to those present in other decades.Thus it can be concluded that the SW region is nearly non-evolving and follows thecharacteristic pattern of the region.50 100 150 200 250 300 35000.511.52Harmonic no.Spectral Density or Power159th 203rd 220th193rd50 100 150 200 250 300 3500.20.40.60.811.21.41.61.8Harmonic no.Spectral Density or Power163rd 211th 246th220tha b50 100 150 200 250 300 3500.20.40.60.811.21.41.61.822.2Harmonic no.Spectral Density or Power166th171th174th205th244th253rd50 100 150 200 250 300 3500.20.40.60.811.21.41.61.822.2Harmonic no.Spectral Density or Power162nd202nd212th262nd239thc d50 100 150 200 250 300 3500.20.40.60.811.21.41.61.82Harmonic no.Spectral Density or Power356th harmonicor10.2 day152nd 162th181th199th219th277th250th0 50 100 150 200 250 300 3500.20.40.60.811.21.41.61.82Harmonic no.Spectral Density or Power314th harmonicor11.6 day149th156th175th180th 218th241th247th257the fFig. 5.2 Discrete Fourier transform of precipitation for SE region over the period (a) 1948–1957,(b) 1958–1967, (c) 1968–1977, (d) 1978–1987, (e) 1988–1997, (f) 1998–20075 Intra Seasonal Variability of Rainfall in India on Regional Basis 77Figures 5.4–5.6 show the DFT of DRC for All-India, SE, and SW regionsrespectively, after excluding the more dominant cycles viz. annual, semiannualand terannual to observe the relative strength of other cycles. Besides, the annual,semiannual and terannual the other dominating cycles present in the spectrum ofDRC for All-India rainfall are the cycles of 91 day, 45 day, 36 day, 18 day, 14 dayand 10 day. The quadannual cycle and the 36 day cycle were also observed in thespectra of SE and SW regions. The cycle of 45 day that corresponds to MaddenJulian Oscillations also referred as MJO and the 18 day cycle was observed in thespectra of SW and All-India region. The spectrum of DRC of SW region containstwo important cycles of 60 day and 52 day which are clearly distinct as prominent50 100 150 200 250 300 3500123456Harmonic no.Spectral Density or Power365 day182.3 day122 day91 day63 day49 day31 day27 day50 100 150 200 250 300 3500123456Harmonic no.Spectral Density or Power365 day182.3 day122 day91 day63 day52 day 37 day 27 daya b50 100 150 200 250 300 3500123456Harmonic no.Spectral Density or Power365 day182.3 day122 day91 day49 day 31 day 27 day37 day57 day50 100 150 200 250 300 350012345Harmonic no.Spectral Density or Power365 day182.3 day122 day91 day52 day46 day31 day 23 day61 day27 day48 dayc d50 100 150 200 250 300 3500.511.522.533.544.555.5Harmonic no.Spectral Density or Power365 day182.3 day122 day91 day64 day48 day46 day37 day27 day50 100 150 200 250 300 35000.511.522.533.544.55Harmonic no.Spectral Density or Power365 day182.3 day122 day91 day57 day52 day45 day37 day36 day 26 daye fFig. 5.3 Discrete Fourier transform of precipitation for SW region over the period (a) 1948–1957,(b) 1958–1967, (c) 1968–1977, (d) 1978–1987, (e) 1988–1997, (f) 1998–200778 M.K. Joshi et al.peaks in the Fig. 5.6. These two prominent peaks were neither observed in thespectra of All-India nor in the SE region. The high frequency cycles i.e. the cycle of6 day, 4 day, 3 day and 2.5 day were also observed in the spectra of all regions. Inaddition to this, the cycle of 5 day was observed in SE and SW region.The basic difference in the spectra of SE region when compared with the spectraof All-India lies in the ratio of strengths of annual and other cycles representingintraseasonal and other smaller variations. While in the All-India pattern, the annualcycle singly dominates the rainfall; in the SE regions, other variations have signifi-cant strengths. This is obvious in the sense that All-India rainfall is nonetheless, anaverage over the smaller regions. However this fact leads to the conclusion thatthere are other regions which exhibit relatively less variations on small time scales.0 20 40 60 80 100 120 140 160 180 20000.020.040.060.080.10.120.14Harmonic no.Spectral Density or Power91day45 day36 day 18 day14 day10 day6 day4.67 day4day2.5 day 2 day3 day2.7 dayFig. 5.4 DFT of DRC excluding annual, semiannual and terannual cycles for All-India region0 20 40 60 80 100 120 140 160 180 20000.050.10.150.20.250.30.350.40.45Harmonic no.Spectral Density or Power36 day16 day91 day9 day11 day23 day6 day 5 day 4 day 3 day 2.5 day2 dayFig. 5.5 DFT of DRC excluding annual, semiannual and terannual cycles for SE region5 Intra Seasonal Variability of Rainfall in India on Regional Basis 79Indeed SW region is one such region. It is observed that that the annual cycle ismore dominant as is seen from the spectra. Thus, the SW region is much less respon-sible to local temporal rain as compared to the SE region. This conclusion cannotbe reached on the basis of the analysis of simple variance as, in that case, overallvariability will be reflected and no the relative strengths of local and global andtemporal signals.Analysis of spectra of climatology of rain over All-India, SE and SW regionsconfirms that this pattern is a characteristic pattern of the Indian rainfall. This can beseen in the relatively strong signals of 91 day cycle for the SW region as comparedto SE and All-India regions.5.4 ConclusionTemporal pattern of the precipitation over India is analysed on regional basis. Thespectra of the regions are investigated on decadal basis and the relative strengths ofthe low and high order harmonics are compared. It is observed that temporally localpattern of rain have more profound effect on the overall feature in the SE regionthan in the SW region. This feature is also reflected in the All-India rainfall pattern.Further, it is seen that the high frequency cycles, the harmonics around the MJoscillations, have shown an increase in strength in the last three decades for the allIndia and SE rain patterns. However, the SW pattern remains comparatively morestable. It may be concluded that the pattern of the SE precipitation is evolving ondecadal basis and an increase in the strength of MJ oscillation may be evident in thenext couple of decades. Study of spectra on finer resolutions and their inter-comparison may further inform the society of the local rain patterns.0 20 40 60 80 100 120 140 160 180 20000.10.20.30.40.50.60.70.8Harmonic no.Spectral Density or Power91 day21 day52 day60 day45 day16 day8 day 5 day6 day 4 day 3 day 2.7 day 2.5 day 2.4 day36 day 18 dayFig. 5.6 DFT of DRC excluding annual, semiannual and terannual cycles for SW region80 M.K. Joshi et al.ReferencesHartmann DL, Gross JR (1988) Seasonal variability of the 40–50 day oscillation in wind andrainfall in the tropics. J Atmos Sci 45(19):2680–2702Hartmann DL, Michelsen ML (1989) Intraseasonal periodicities in Indian rainfall. J Atmos Sci 46(18):2838–2862Krishnamurti TN, Ardanuy P (1980) The 10 to 20-day westward propagating mode and “breaks inthe monsoons”. Tellus 32:15–26Krishnamurti TN, Bhalme HN (1976) Oscillations of a monsoon system. Part I. Observationalaspects. J Atmos Sci 33:1937–1954Krishnamurthy V, Shukla J (2007) Intraseasonal and seasonally persisting patterns of Indianmonsoon rainfall. J Climate 20:3–20Laughlin GP, Zuo H, Walcott J, Bugg AL (2003) The rainfall reliability wizard-a new tool torapidly analyse spatial rainfall reliability with examples. Environ Model Softw 18:49–57Lihua M, Yanben H, Zhiqiang Y (2007) The possible influence of solar activity on Indian summermonsoon rainfall. Appl Geophys 4(3):231–237Luk KC, Ball JE, Sharma A (2001) An application of artificial neural networks for rainfallforecasting. Math Comput Modell 33:883–699Murakami M (1977) Spectrum analysis relevant to Indian monsoon. Pure Appl Geophys 115(5–6):1145–1166Peters O, Hertlien C, Christensen K (2002) A complexity view of rainfall. Phys Rev Lett 88(1):1–4Rajeevan M, Bhate J, Jaswal AK (2008) Analysis of variability and trends of extreme rain-fall events over India using 104 years of gridded daily rainfall data. Geophys Res Lett.doi:10.1029/2008GL035143Rangarajan GK (1994) Singular spectral analysis of homogeneous Indian monsoon (HIM) rainfall.J Earth Syst Sci 103(4):439–448Vijayakumar R, Kulkarni JR (1995) The variability of the interannual oscillations of the Indiansummer monsoon rainfall. Adv Atmos Sci 12(1):95–1025 Intra Seasonal Variability of Rainfall in India on Regional Basis 81.Chapter 6Assimilation of Surface Observations in a HighResolution WRF ModelDipak K. Sahu and S.K. DashAbstract In this study surface parameters such as dry-bulb temperature, dew-pointtemperature and wind speed at 8:30 IST (0300 UTC) from 32 Government InterColleges (GICs) in Uttarakhand are assimilated into a high-resolution WeatherResearch and Forecasting (WRF) version 2.2 model by using Three-DimensionalVariational (3DVAR) surface data assimilation technique. These surface weatherobservations are made in a programme entitled Participation of Youth in Real-time/field Observations to Benefit the Education (PROBE), which was initiated by theDepartment of Science and Technology (DST), India in the state of Uttarakhand in2003. The WRF model has been integrated over the Uttarakhand domain at ahorizontal resolution of 5 km. The assimilation of surface weather parametershave resulted in a noticeable impact on the horizontal and vertical distributions oftemperature and wind fields. Also the comparison of simulated fields obtained withand without surface data assimilation in the model indicates comparatively lesserror after assimilation.6.1 IntroductionLand surface is an important interface that exchanges energy between the earth and theatmosphere. Also surface parameters are crucial to the simulation of weatherparameters by numerical methods. Because of lack of high quality spatial distributionsof surface weather parameters those are not assimilated into mesoscale models andhence the difficulty in reasonable mesoscale predictions. To improve local weathersimulations, it is necessary to collect and assimilate high resolution surface obser-vations into a mesoscale model. In the past, several studies (Ruggiero et al. 1996;D.K. Sahu (*) • S.K. DashCentre for Atmospheric Sciences, Indian Institute of Technology Delhi, Hauz Khas, New Delhi110 016, Indiae-mail: dipakmath@gmail.com; skdash@cas.iitd.ac.inS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_6, # Springer-Verlag Berlin Heidelberg 201183Alapaty et al. 2001; Alapaty et al. 2008; Stauffer et al. 1991; Vinodkumar et al.2008a, b) have been conducted to improve simulations of weather parameters byusing direct or model analyzed surface observations. The study conducted byRuggiero et al. (1996) described a system for the frequent intermittent assimilationof surface observations into a mesoscale model. Results from a case study indicatedthat the frequent intermittent assimilation of surface data can provide a superiormesoscale forecast. Alapaty et al. (2008) and Vinodkumar et al. (2008a) haveinvestigated the effect of indirect surface data assimilation using Flux-AdjustingSurface Data Assimilation System (FASDAS). Their results indicate that FASDASconsistently improved the accuracy of the model simulations. Further Vinodkumaret al. (2008b) investigated the impact of land surface processes and data assimilationon mesoscale convection and precipitation for a heavy rain event associated withoffshore troughs over the Indian monsoon region. Their results showed that surfacedata assimilation simulated the strongest vertical wind velocity fields as well as theassociated potential vorticity fields as compared with the model simulation fromNCEP reanalysis initial and boundary conditions (2.5� � 2.5�). In an earlier studyStauffer et al. (1991) presented a direct continuous assimilation of standardresolu-tion rawinsonde and meso-alpha surface observations throughout the model integra-tion rather than at the initial time. Within the planetary boundary layer they alsoinvestigated the effect of data assimilation by using four-dimensional data assimila-tion technique based on Newtonian nudging. The main objective of their study wasto effectively utilize the combined strength of these two simple data systems whileavoiding their individual weaknesses. Assimilation of surface wind and moisturedata throughout the model planetary boundary layer generally showed a positiveimpact on the simulated precipitation. Xavier et al. (2008) assimilated the satelliteand conventional meteorological data by using analysis nudging to study the effectof increased vertical and horizontal resolution as well as convective parameteriza-tion. The significant result from their study indicated that the improvements in thesimulation using nudging run are better than the improvements in the simulation dueto high-resolution and cumulus parameterization sensitivity.Uttarakhand is located in the northern part of India and has a total geographicarea of 51,125 km2. Approximately 93% of Uttarakhand is covered by HimalayanMountains and about 64% of the mountains are forest. This state has sparse weatherdata available for use in numerical weather prediction (NWP). Under a programentitled the Participation of Youth in Real-time/field Observations to Benefit theEducation (PROBE) initiated by the Department of Science and Technology (DST),Met Labs are successfully installed in several Government Inter Colleges (GICs) inUttarakhand. One of the objectives of this initiative is to use the surface weatherobservations in a high-resolution Weather Research and Forecasting (WRF) modelto examine their impact on the local weather prediction. Initiated from Uttarakhandin 2003, this PROBE program has now spread over different states in Indiaincluding NCR-Delhi, Orissa and Tamilnadu. Currently under this program, a goodset of surface meteorological parameters has been collected from Uttarakhand in thelast 3 years. In the present study these data are utilized in the data assimilation schemeof the WRF model.84 D.K. Sahu and S.K. DashIn the present study surface observations such as dry-bulb temperature, windspeed and dew point temperature have been assimilated into Three-DimensionalVariational (3DVAR) data assimilation technique of the WRFV2.2 model in orderto examine the impact of additional data in the local weather simulations. Thesurface observations are used at a minimum distance of about 5 km in the horizon-tal. Section 6.2 in this paper describes the WRF model in brief and the experimentaldesign. Section 6.3 gives the details of observations used in the 3DVAR surfacedata assimilation (SDA) model simulations. Section 6.4 analyzes the statisticalerrors in terms of relative errors (REs) and root mean square differences(RMSDs). Sections 6.5 and 6.6 discuss the horizontal and vertical structures ofthe parameters before and after assimilation. Finally conclusions are given inSect. 6.7.6.2 Model Description and Experimental DesignThe state-of-the-art WRF is the result of collaborative effort among the NCARMesoscale and Microscale Meteorology (MMM) Division, the National Oceanicand Atmospheric Administration’s (NOAA), National Centers for EnvironmentalPrediction (NCEP), Forecast System Laboratory (FSL), the Department ofDefense’s Air Force Weather Agency (AFWA), Naval Research Laboratory(NRL), the Centre for Analysis and Prediction of Storms (CAPS) at the Universityof Oklahoma, and the Federal Aviation Administration (FAA) and a number ofother university scientists. WRF model is designed to replace existing U.S. researchmodels (e.g., MM5) and operational models (e.g., Eta) under a common softwarearchitecture. The performance of WRF has been demonstrated to be comparable oreven better than MM5 and other existing mesoscale models in weather forecasting(Gallus et al. 2005; Grams et al. 2006). A detail description of the model equations,physics and dynamics are available in Dudhia (2004). WRF is a limited area, non-hydrostatic, with terrain following eta-coordinate mesoscale modeling systemdesigned to serve both operational forecasting and atmospheric research needs.There are a wide range of physical parameterization schemes available in WRF.After conducting a number of sensitivity studies we have selected the followingphysical parameterization schemes in this study. The schemes used include Ferrier(new Eta) microphysical parameterization, the ensemble Grell-Devenyi cumulusparameterization (Grell and Devenyi 2002), Dudhia shortwave radiation (Dudhia1989) and Rapid Radiative Transfer Model (RRTM) longwave radiation (Mlaweret al. 1997), the Yonsei University (YSU) planetary boundary layer (PBL) scheme(Noh et al. 2003), the thermal diffusion land surface model and Monin-Obukhovsurface layer scheme.Advanced Research WRF (ARW) dynamic core version 2.2, is used in thisstudy. Following two numerical experiments have been conducted in this study.In the single domain experiment the model is centered at 30.2�N and 79.4�E with76 � 78 horizontal grid points in the longitudinal and latitudinal directions at6 Assimilation of Surface Observations in a High Resolution WRF Model 85horizontal resolution of 5 km as shown in Fig. 6.1. The Mercator map projection isused as the model horizontal coordinates. The model has been integrated verticallyup to 100 hPa with 27 terrain-following sigma levels.This single domain run referred to as control run (CTRL) in the text wasconducted by using the NCEP-NCAR Final Analysis (FNL) at 1� � 1� grid resolu-tion to prepare initial and lateral boundary conditions. The second experiment,henceforth referred to as surface data assimilation run (SDA) was conducted withmodified initial and lateral boundary conditions by assimilating additional surfaceobservations at 32 stations with the help of 3DVAR data assimilation schemeavailable with WRFV2.2. Barker et al. (2004) and Lorenc et al. (2000) provide a detaildescription of the 3DVAR system used in this study. In SDA case three differentexperiments, the first from 0000 UTC 02 Dec to 0000 UTC 03 Dec 2006, the 2ndfrom 0000 UTC, 11 Dec to 0000 UTC, 12 Dec 2006 and the 3rd from 0000 UTC, 27Dec to 0000 UTC, 28 Dec 2006 each for 24 h were conducted. These threeexperiments will be referred to as EXP1, EXP2 and EXP3 subsequently.6.3 Observational Data UsedIn this study we have used the surface measurements from U-PROBE (DST, 2007),since no other reliable data are available over the region of interest. U-PROBEprogramme initiated by DST aims at making climate and weather science educationinteresting and useful where school children and teachers participate actively.Under this programme about 97 Met Labs are functioning in Uttarakhand statetill today. However, we have selected data from 32 stations shown in Fig. 6.2a inFig. 6.1 Domain used for WRF integration covering Uttarakhand state86 D.K. Sahu and S.K. Dashthis study. Under U-PROBE all the surface measurements of dry-bulb temperature,wind speed and dew point temperature were made at 08.30 IST as required by themodel configuration. The domain selected in this study has very complex terrainwhich includes some parts of the Himalaya. The terrain heights used in the modelare shown in Fig. 6.2b, so that the relative heights of stations shown in Fig. 6.2a canbe conveniently assessed. Thus we assimilated surface weather parameters in 5 kmWRF-ARW model at 0300 UTC (08.30 IST) to prepare the updated initial andlateral boundary conditions for 24 h model integration.6.4 Relative Errors and RMSDsA comparative study has been conducted between the CTRL and SDA at threedifferent stations Rishikesh, Newtehri and Thal in case of the three numericalexperimentsEXP1, EXP2 and EXP3. To estimate errors between the two modelsimulations and observed values, relative errors (REs) are calculated at each stationfor each parameter between CTRL and observations on one hand and SDA andobservations on the other.Relative error between CTRL simulations and observations, RE-ctrl ¼ C�Oð ÞO��� ���Relative error between SDA simulations and observations, RE-sda ¼ D�Oð ÞO��� ���Here C represents the CTRL model output at 0300 UTC, D represents the SDAmodel output and O represents the observational value at 0300 UTC (08.30 IST).Lat 30.2°N/Lon 79.4°Ea bFig. 6.2 (a) Location of 32 Met Labs from which surface data are assimilated (b) Terrain heightsused in WRF-ARW model6 Assimilation of Surface Observations in a High Resolution WRF Model 87Also the respective difference between the CTRL and SDA simulationsrepresented by the root-mean-square difference (RMSD) is given byRMSD ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1NPNi¼1Ci � Dið Þ2sHere Ci and Di represent the CTRL and SDA predicted values respectively at aparticular time of the 3 h output of 24 h model integration. N is the number of modeloutput time in the 24 h model integration, in this case N ¼ 9. It is well known thatRMSD is a good estimate of the average difference and RE is a relative differencemeasure (Bonnardot and Cautenet 2009). Table 6.1 shows the RE-ctrl and RE-sda fordry-bulb temperature, dew point temperature and relative humidity at Thal, Rishikeshand Newtehri. The additional variable wind speed is available only at Newtehri andhence has been considered here. Wind speeds are not available at Thal and Rishikesh.Comparison shows that at all the three stations RE-sda is comparatively less than RE-ctrl, which implies the advantages of surface data assimilation in the SDA experiment.Generally a good agreement between observation and forecast leads to a smaller valueof RMSE. But here smaller value of RMSD indicates the additional surface datahas less impact on respective weather parameters and larger value indicates moreimpact. Table 6.2 represents the RMSD between the CTRL and SDA at all the threestudied stations. It shows that in case of temperature the RMSD is less compare toother parameters and this result again leads to the inference that assimilation ofsurface observations from additional 32 Met Labs has improved the simulations.However, the impact is less in case of temperature compared to that of relativehumidity, dew point temperature and wind speed.6.5 Spatial Errors in Meteorological Parameters6.5.1 Analysis at 850 hPaThe spatial distribution of difference in wind speed and dew point temperature at850 hPa are shown in Figs. 6.3 and 6.4 respectively. The difference is calculated bysimply subtracting the SDA model simulations from CTRL model simulation ateach 3 h model output. The spatial differences in dry-bulb temperature and relativehumidity are also analyzed at 850 hPa but not shown in figures to save space. Level850 hPa is chosen because in more than half of the domain the terrain heights aregreater than 1 km [Fig. 6.2b]. It is found that the maximum difference in dry-bulbtemperature is up to 1�C and the minimum difference is 0.5�C. In case of EXP1 themodel simulates higher temperature in SDA in most part of the domain. However,EXP2 and EXP3 simulated higher temperature in case of CTRL. Figure 6.3 showsthat maximum and minimum differences in wind speed between CTRL and SDA at850 hPa are 2.4 m/s and 0.9 m/s respectively. It is seen from Fig. 6.3 that in mostparts of the domain captured differences are up to 0.8–0.9 m/s and certain parts up88 D.K. Sahu and S.K. DashTable6.1RelativeErrors(REs)betweenthemeasuredparametersat03UTCandcorrespondingcontrol(CTRL)anddataassimilation(SDA)experimentsAnalysistimeThal(29.45� N,80.16� E)Rishikesh(30.04� N,78.15� E)Newtehri(30.22� N,78.26� E)TemperatureDewpointtemperatureRelativehumidityTemperatureDewpointtemperatureRelativehumidityTemperatureDewpointtemperatureRelativehumidityWindspeedCTRLSDACTRLSDACTRLSDACTRLSDACTRLSDACTRLSDACTRLSDACTRLSDACTRLSDACTRLSDA2-Dec-060.8280.8284.0482.2170.8330.7290.1180.0892.3691.1360.5850.4010.0880.0716.359.240.7690.563.4523.2703UTC11-Dec-060.5790.5570.60.3690.3370.2750.3030.3010.5860.490.0490.0190.7190.71.1620.840.4010.355.7655.103UTC27-Dec-060.7330.7123.22.1730.7220.6090.3360.3361.3291.0910.3180.2130.5660.5511.597.920.6690.572.6221.8203UTCTable6.2RootMeanSquareDifferences(RMSDs)ofday-1betweencontrol(CTRL)anddataassimilation(SDA)experimentsExperimentsThal(29.45� N,80.16� E)Rishikesh(30.04� N,78.15� E)Newtehri(30.22� N,78.26� E)TemperatureWindspeedDewpointtemperatureRelativehumidityTemperatureWindspeedDewpointtemperatureRelativehumidityTemperatureWindspeedDewpointtemperatureRelativehumidity2-Dec-060.1430.2044.3617.3460.1090.3253.5955.5110.06420.1824.8477.21911-Dec-060.1080.280.6093.2190.1030.1270.6412.3670.04550.2550.612.79927-Dec-060.1520.2932.9987.6340.0660.4521.9244.0110.05570.2642.3354.8816 Assimilation of Surface Observations in a High Resolution WRF Model 8931.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E 77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E2.42.11.81.51.20.90.60.30.30.60.91.21.5002 Dec-0300 UTC31.8N 31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E 77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E1.81.51.41.20.90.60.30.30.40.60.9002 Dec-0600 UTC31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E1.41.20.80.60.40.20.20.40.60.80177.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E11 Dec-0300 UTC 11 Dec-0600 UTC27 Dec-0300 UTC 27 Dec-0600 UTCFig. 6.3 Horizontal distributions of wind speed differences at 850 hPa between CTRL and SDAsimulations90 D.K. Sahu and S.K. Dash31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E 77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E00.51.52.53.54.55.56.501234567123456789101112131402 Dec-0300 UTC 02 Dec-0600 UTC31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E 77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E00.20.30.40.60.81.21.31.41.6111 Dec-0300 UTC 11 Dec-0600 UTC27 Dec-0300 UTC 27 Dec-0600 UTCFig. 6.4 Horizontal distributions of dew-point temperature differences at 850 hPa between CTRLand SDA simulations6 Assimilation of Surface Observations in a High Resolution WRF Model 91to 2–2.4 m/s between CTRL and SDA. Figure 6.4 shows that maximum andminimum spatial differences between CTRL and SDA at 850 hPa in dew pointtemperature are up to 14�C and 1.6�C for EXP1 and EXP2 respectively. It is seenfrom Fig. 6.4 that SDA model simulates maximum dew point temperature thanCTRL in all three experiments. Spatial difference of relative humidity at 850 hPa ismaximum up to 15% in EXP1. From all the three experiments it is seen that SDAsimulates maximum values of relative humidity than CTRL simulations at both0300 and 0600 UTC.6.5.2 Analysis at 500 hPaFigures 6.5 and 6.6 depict the differences in wind speed and dew point temperatureat 500 hPa in each experiment. Here also the difference is calculated by simplysubtracting the model simulations of SDA from those of CTRL model simulation ateach 3 h model output for each described parameters. The spatial differences in dry-bulb temperature and relative humidity are also analyzed at 500 hPa but the figuresare not shown. It is seen that the maximum and minimum differences in the dry-bulb temperature at 500 hPa are up to 1.5�C and 0.5�C in EXP2 and EXP3respectively. Similarly Fig. 6.5 shows that the differences in wind speed at500 hPa are maximum up to 6 m/s in case of EXP1 and 2 and the minimum is1.5 m/s in EXP3. It is seen from Fig. 6.5 that the differences in wind speed in eachexperiment are comparatively less at 0600 UTC than at 0300 UTC. The results arevery good at 850 hPa level. Figure 6.6 depicts that the maximum differences in dewpoint temperature are up to 30�C in EXP2 and 3 and in EXP1 it is 16�C. Also inFig. 6.6 the differences in dew point temperature is comparatively less at 0600 UTCthan at 0300 UTC. It is also found that the differences in relative humidity at500 hPa are maximum up to 40–50% in EXP1 and 2 and in EXP3 it is up to 30%.6.6 Effect of Surface Data Assimilation on Vertical Structuresof Meteorological VariablesIn this section the impact of assimilating additional surface observations on thevertical profiles of dry-bulb temperature, wind speed and dew point temperature atthree selected stations such as Thal (29.45�N/80.16�E), Rishikesh (30.04�N/78.15�E) and Newtehri (30.22�N/78.26�E) has been examined. Figure 6.7 depictsthe variations in vertical profiles of wind speed between CTRL and SDAsimulations in EXP1 and 2. It is seen that there is a maximum variation up to2 m/s at lower level of the atmosphere and the variations are distinctly visible up to700 hPa. From Fig. 6.7b, it is seen that at Newtehri the variations in vertical profilesof wind speeds are significant up to 600 hPa. In Fig. 6.8 there are noticeable92 D.K. Sahu and S.K. DashFig. 6.5 Horizontal distributions of wind speed differences at 500 hPa between CTRL and SDAsimulations6 Assimilation of Surface Observations in a High Resolution WRF Model 9302 Dec-0300 UTC11 Dec-0300 UTC27 Dec-0300 UTC 27 Dec-0600 UTC02 Dec-0600 UTC11 Dec-0600 UTC10234567891011121314151620151050035679121518212427305101520253031.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N31.8N31.5N31.2N30.9N30.6N30.3N30N29.7N29.4N29.1N28.8N28.5N77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E 77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81E77.5E 78E 78.5E 79E 79.5E 80E 80.5E 81EFig. 6.6 Horizontal distributions of dew-point temperature differences at 500 hPa between CTRLand SDA simulations94 D.K. Sahu and S.K. DashWind speed(m / sec) Vs Pressure level atthal (29.45 N/80.16 E)Wind speed(m / sec) Vs Pressure level atrishikesh (30.04 N/78.15 E)Wind speed(m / sec) Vs Pressure level atnewtehri(30.22N/78.26 E)Wind speed(m / sec) Vs Pressure level atnewtehri(30.22N/78.26 E)Wind speed(m / sec) Vs Pressure level atrishikesh (30.04 N/78.15 E)Wind speed(m / sec) Vs Pressure level atthal (29.45 N/80.16 E)5005506006507007508008509009505005506006507007508008509009505005506006507007508008509009505005506006507007508008509009505005506006507007508008509009502 4 6 8 10 12 14 16 18 1 2 3 4 5 6 7 8 9 10 11 12 13 14 152 3 4 5 6 7 8 9 10 11 12 13 14 155005506006507007508008509009500 2 4 6 8 10 12 14 16pressure(hPo)pressure(hPo)pressure(hPo)pressure(hPo)pressure(hPo)pressure(hPo)Wind speed (m/ sec)Wind speed (m/ sec)2 4 6 8 10 12 14 16 18Wind speed (m/ sec) Wind speed (m/ sec)0 3 6 9 12 15 18 21 24Wind speed (m/ sec)Wind speed (m/ sec)CTRLSDA11 Dec 0600 UTC11 Dec 0600 UTC11 Dec 0600 UTC02 Dec 0600 UTC02 Dec 0600 UTC02 Dec 0600 UTCa bFig. 6.7 Comparison of vertical profiles of wind speeds in (a) EXP1 (02 Dec–03 Dec 2006) and(b) EXP2 (11 Dec–12 Dec 2006) between CTRL and SDA simulations6 Assimilation of Surface Observations in a High Resolution WRF Model 9502 Dec 0600 UTC 27 Dec 0600 UTC500550600650700750800850900950pressure(hPo)500550600650700750800850900950pressure(hPo)–50 –45 –40 –35 –30 –25 –20 –15 –10 –5 0 –50 –45 –40 –35 –30 –25 –20 –15 –10 –5 0 5Dewpt temperature Vs pressure level atNewtehri (30.22 N/78.26E) Dewpt temperature Vs pressure level atNewtehri (30.22 N/78.26E) 27 Dec 0600 UTC500550600650700750800850900950pressure(hPo)500550600650700750800850900950pressure(hPo)–50 –45 –40 –35 –30 –25 –20 –15 –10 –5 0 –50–60 –55 –45 –40 –35 –30 –25 –20 –15 –10 –5 5002 Dec 0600 UTCDewpt temperature Vs pressure level atRishikesh (30.04 N/78.15E) Dewpt temperature Vs pressure level atRishikesh (30.04 N/78.15E) 0Dewpt temperature Vs pressure level at Thal (29.45 N/80.16 E)Dewpt temperature Vs pressure level at Thal (29.45 N/80.16 E)500CTRLSDA02 Dec 0600 UTC 27 Dec 0600 UTC500550600650700750800850900950pressure(hPo)550600650700750800850900950pressure(hPo)-40 -33 -30 -25 -20 -15 -5-10 0 –45 –40 –35 –30 –25 –20 –15 –10 –5 0 5650-30Dewpt temperature (°C)Dewpt temperature (°C)Dewpt temperature (°C)Dewpt temperature (°C)Dewpt temperature (°C)Dewpt temperature (°C)a bFig. 6.8 Comparison of vertical profiles of dew point temperatures in (a) EXP1 (02 Dec–03 Dec2006) and (b) EXP3 (27 Dec–28 Dec 2006) between CTRL and SDA simulations96 D.K. Sahu and S.K. DashRelative humidity(%) Vs pressure level at newtehri (30.22 N/78.26 E)Relative humidity(%) Vs pressure level at newtehri (30.22 N/78.26 E)0 10 20 30 40 50 60 70 80 90 1000 10 20 30 40 50 60 70 80 90 100500550600650700750800850900950pressure(hPo)500550600650700750800850900950pressure(hPo)Relative humidity(%) Relative humidity(%) 02 Dec 0600 UTCRelative humidity(%) Vs pressure level at Rishikesh (30.04 N/78.15 E)Relative humidity(%) Vs pressure level at Rishikesh (30.04 N/78.15 E)0 10 20 30 40 50 60 70 80 90 1000 10 20 30 40 50 60 70 80 90 100500550600650700750800850900950pressure(hPo)500550600650700750800850900950pressure(hPo)Relative humidity(%) Relative humidity(%) 02 Dec 0600 UTC 27 Dec 0600 UTCRelative humidity(%) Vs pressure level at thal (29.45 N/80.16 E)Relative humidity(%) Vs pressure level at thal (29.45 N/80.16 E)5005506006507007508008509009500 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100pressure(hPo)500550600650700750800850900950pressure(hPo)Relative humidity(%) Relative humidity(%)CTRLSDA 02 Dec 0600 UTC 27 Dec 0600 UTC27 Dec 0600 UTCFig. 6.9 Comparison of vertical profiles of Relative humidities in (a) EXP1 (02 Dec–03 Dec2006) and (b) EXP3 (27 Dec–28 Dec 2006) between CTRL and SDA simulations6 Assimilation of Surface Observations in a High Resolution WRF Model 97variations in the vertical profile of dew point temperatures at all the selected stationsof EXP1 and 3. It may be noted that results of all the three experiments areanalyzed, but figures of EXP2 are not shown to save space. The variations of dewpoint temperature are about 5�C at the surface in EXP1 and it is comparatively lessin EXP3. It is found that the variation in vertical profiles of dew point temperaturein EXP2 is less compared to other two experiments. In Fig. 6.9 there are alsonoticeable variations in the vertical profiles of relative humidity and the maximumvariation is up to 10% at the surface level at all three stations in EXP1 and 3 andgradually decreases towards the upper atmosphere. In case of relative humidity alsothe results of all the three experiments are examined, but figures of EXP2 are notshown. It is found from EXP2 that the difference in vertical profiles of relativehumidity between CTRL and SDA is maximum up to 5% at the surface level whichis comparatively less than the other two experiments and it also gradually decreasestowards the upper level of the atmosphere. It may be noted that there is very lessdifference in dry-bulb temperature at the surface level between CTRL and SDAsimulations. It is found that the maximum differences in dry-bulb temperatureobtained in EXP1 are up to 0.6�C and 0.5�C at Rishikesh and Newtehrirespectively.6.7 ConclusionsThis study has been conducted to examine the impact of assimilating additionalsurface observations over Uttarakhand state of India in a high resolution WRFV2.2model by using 3DVAR data assimilation technique. The impacts on the horizontaland vertical distributions of dry-bulb temperature, wind speed, dew point tempera-ture and relative humidity are examined in detail. Results indicate that there aresome noticeable differences between the simulated fields obtained from CTRL andSDA runs. Analysis of the vertical profiles of the meteorological parameters at threedifferent stations indicate significant differences in dew point temperature, relativehumidity and wind speed after 3DVAR data assimilation. But in case of dry-bulbtemperature profiles the difference is comparative less. There is a maximumimprovement of 24.6% in temperature at Rishikesh and also a maximum of30.6% in case of wind speed at Newtehri. Results also show that the improvementsare quite large in dew point temperature and relative humidity with maximum of upto 52% and 62% respectively. Based on the above results it is inferred thatadditional surface data, when assimilated into a high resolution regional modelare able to improve the simulations of meteorological parameters. The performanceof 3DVAR assimilation may further improve by continuous assimilation of surfaceobservations, not only at initial time but also at different time steps throughout theintegration.98 D.K. Sahu and S.K. DashReferencesAlapaty K, Seaman NL, Niyogi DS, Hanna AF (2001) Assimilating surface data to improve theaccuracy of atmospheric boundary layer simulations. J Appl Meteorol 40:2068–2082Alapaty K, Niyogi DS, Chen F, Pyle P, Chandrasekhar A, Seaman N (2008) Development of theflux-adjusting surface data assimilation system for mesoscale models. J Appl MeteorolClimatol 47:2331–2350Barker DM, Huang W, Guo YR, Bourgeois A, Xiao XN (2004) A three-dimensional variationaldata assimilation system for MM5: implementation and initial results. Mon Weather Rev132:897–914Bonnardot V, Cautenet S (2009) Mesoscale atmospheric modeling using a high horizontal gridresolution over a complex coastal terrain and a wine region of South Africa. J Appl MeteorolClimatol 48:330–348DST (2007) U-PROBE Annual Data Report, Department of Science and Technology, Govt. ofIndia, pp 147Dudhia J (1989) Numerical study of convection observed during the winter monsoon experimentusing a meso-scale two-dimensional model. J Atmos Sci 46(20):3077–3107Dudhia J (2004) The weather research and forecasting model (version 2.0). In: 2nd internationalworkshop on next generation NWP model. Yonsei University, Seoul, pp 19–23Gallus WA, Correia J, Jankov I (2005) The 4 June 1999 derecho event: a particularly difficultchallenge for numerical weather prediction. Weather Forecast 20:705–728Grams JS, Gallus WA, Koch SE, Wharton LS, Loughe A, Ebert EE (2006) The use of a modifiedEbert-Mcbride technique to evaluate mesoscale model QPF as a function of convective systemmorphology during IHOP 2002. Weather Forecast 21:288–306Grell GA, Devenyi D (2002) A generalized approach to parameterizing convection combiningensemble and data assimilation techniques. Geophys Res Lett 29(14), Article 1693, 4Lorenc AC, Ballard SP, Bell RS, Ingleby NB, Andrews PLF, Barker DM, Bray JR, Clayton AM,Dalby T, Li D, Payne TJ, Saunders FW (2000) The Met. Office global three-dimensionalvariational data assimilation scheme. Q J R Meteorol Soc 126:2991–3012Mlawer EJ, Taubman SJ, Brown PD, Iacono MJ, Clough SA (1997) Radiative transfer forinhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J GeophysRes 102(D14):16663–16682Noh Y, Cheon WG, Hong SY, Raasch S (2003) Improvement of the K-profile model for theplanetary boundary layer based on large eddy simulation data. Bound Layer Meteorol107:401–427Ruggiero FH, Sashegyi KD, Madala RV, Raman S (1996) The use of surface observations in four-dimensional data assimilation in a mesoscale model. Mon Weather Rev 124:1018–1033StaufferDR, SeamanNL, Binkowski FS (1991)Use of four dimensional data assimilation in a limitedarea mesoscale model Part-II: effects of data assimilation within the planetary boundary layer.Mon Weather Rev 119:734–754Vinodkumar V, Chandrasekhar A, Alapaty K, Niyogi D (2008a) The impacts of indirect soilmoisture assimilation and direct surface temperature and humidity assimilation on a mesoscalemodel simulation of an Indian monsoon depression. J Appl Meteorol Climatol 47:1393–1412Vinodkumar V, Chandrasekhar A, Niyogi D, Alapaty K (2008b) Impact of land surface represen-tation and surface data assimilation on the simulation of an off-shore trough over the Arabiansea. Glob Planet Change. doi:10.1016/gloplacha.2008.12.004Xavier VF, Chandrasekar A, Rahman H, Niyogi D, Alapaty K (2008) The effect of assimilation ofsatellite and conventional meteorological data for the prediction of a monsoon depression overIndia using a mesoscale model. Meteorol Atmos Phys. DOI:10.1007/s00703-008-0314-76 Assimilation of Surface Observations in a High Resolution WRF Model 99.Chapter 7An Evaluation of the Simulation of Monthlyto Seasonal Summer Monsoon Rainfall overIndia with a Coupled Ocean AtmosphereGeneral Circulation Model (GloSea)D.R. Pattanaik, Ajit Tyagi, U.C. Mohanty, and Anca BrookshawAbstract The performance of the UK Met Office’s coupled ocean-atmosphereGeneral Circulation Model (GCM) is evaluated in simulation of summer monsoonrainfall over Indian monsoon region. The UK Met Office’s Global Seasonal(GloSea) forecasting model is initialized at 0000 UTC of 1st May and integratedfor a period of 6 month with 15 ensemble members to generate the model forecast.These experiments have been conducted in similar approach from 1987 to 2002(16 years) to have monthly as well as seasonal forecast of individual year. Themodel simulated rainfall is compared with the verification analysis (Xie-Arkin)during the monsoon season from June to September (JJAS).The monthly forecast climatology from June to September separately and theseasonal forecast climatology (June to September; JJAS) of rainfall are wellsimulated by the model with two maxima viz., one over the west coast of Indiaand other over the head Bay of Bengal region. However, the rainfall magnitude overthe west-coast of India is less in the model simulation for monthly as well as inseasonal simulation. The model has shown good skill in simulation of seasonal(JJAS) mean rainfall over the Indian monsoon region. However, a little overesti-mation in rainfall is noted (approximately 4%) when considered the Indian mon-soon region covering the land region and surrounding oceanic regions. The patterncorrelation during JJAS shows highly significant correlation coefficients (CCs) overthe global tropics (0.91) and Indian monsoon region (0.82). Similarly the RootD.R. Pattanaik (*) • A. TyagiIndia Meteorological Department (IMD), New Delhi, Indiae-mail: drpattanaik@gmail.com; ajit.tyagi@gmail.comU.C. MohantyCentre for Atmospheric Sciences, Indian Institute of Technology Delhi, Hauz Khas, New Delhi110 016, Indiae-mail: mohanty@cas.iitd.ernet.inA. BrookshawMet Office, Exeter, United KingdomS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_7, # Springer-Verlag Berlin Heidelberg 2011101Mean Square Error (RMSE) during JJAS is found to be less (1.01) over the globaltropics than the Indian monsoon region (1.68). The interannual variability offorecast ensemble mean rainfall over the Indian monsoon region shows similarbehaviour with that of verification rainfall variability with Correlation Coefficientof about 0.43 during the 16 years period from 1987 to 2002. The AnomalyCorrelation Coefficients (ACCs) between verification and simulated rainfall during1987–2002 over the Indian monsoon region is quite significant (more than 0.6during some years). Overall, it can be stated that the performance of the UK MetOffice’s seasonal mean simulation is reasonably good.7.1 IntroductionThe extended range forecast from monthly to seasonal scale in tropics is one of themost challenging tasks in atmospheric sciences. The climate forecasting of precipi-tation in monthly to seasonal scale has significant implication in policy planningand national economy for the agro-economic country like India. The Indian summermonsoon rainfall forecasting in long-range has been initiated more than a centuryago (Blanford 1884). The methods include; statistical, empirical and dynamical. Inlast few decades, many statistical (Shukla and Mooley 1987; Gowariker et al. 1991;Sahai et al. 2003) as well as dynamical (Palmer et al. 1992; Chen and Yen 1994;Sperber and Palmer 1996; Soman and Slingo 1997; Shukla et al. 2000; Saha et al.2006; Pattanaik and Kumar 2010) models have been developed for the seasonalprediction of precipitation. The scientific basis of these dynamical seasonalforecasting is that, in tropics, the lower-boundary forcing (sea surface temperature(SST), sea-ice cover, land-surface temperature and albedo, vegetation cover andtype, soil moisture and snow cover etc.), which evolve on a slower time-scale thanthat of the weather systems themselves, can give rise to significant predictability ofstatistical characteristics of large-scale atmospheric events (Charney and Shukla1981). Several observational and modelling studies (Charney and Shukla 1981;Palmer and Anderson 1994) provide evidence that boundary forcing in the tropicscontribute significantly to the internal variability of the tropical as well as monsooncirculations. Atmospheric General Circulation Models (AGCM) and Global Cou-pled GCMs (CGCMs) are the main tools for dynamical seasonal scale prediction.Though in dynamical model, significant improvement has been made through theimprovement of the model physics and dynamics in last few years, but present dayAGCM could not able to simulate mean and interannual variability of Indiansummer monsoon very successfully (Kang et al. 2002). It is also found that theskill of the AGCM is poorer in simulating Indian monsoon; probably this is due tolack of proper representation of realistic sea surface temperature (SST). In recentyears, it is found that the forecast errors in the seasonal prediction can be reducedthrough the combinations of the ensemble members forecast (Brankovic et al. 1990;Brankovic and Palmer 1997). Therefore, the focus is now mainly on multi-modelensemble/super ensemble forecast (Krishnamurti et al. 1999; Wu et al. 2002;102 D.R. Pattanaik et al.Wang et al. 2004 and Chakraborty and Krishnamurti 2006) for the seasonal andinterannual prediction of monsoon. The methods like simple ensemble mean (Penget al. 2002; Pavan and Doblas-Reyes 2000; Doblas-Reyes et al. 2000; Stephensonand Doblas-Reyes 2000; Palmer et al. 2004), regression improved ensemble mean(Peng et al. 2002; Kharin and Zwiers 2002), bias removed ensemble mean (Kharinand Zwiers 2002) and the multi model super ensemble (Krishnamurti et al. 1999)are included in the ensemble forecast. One approach is to generate the ensemblemember forecast based on combination of forecasts obtained from model withperturbed initial conditions. Lorenz (1969) demonstrated that small perturbationsin the initial conditions could produce a very different result in the final states.Now a number of operational Numerical Weather Prediction (NWP) centres anda number of climate prediction centres around the world are making climateprediction at seasonal scale using AGCMs and global coupled GCMs. Seasonalforecasting group, Met Office, UK is one of them. It can be mentioned here that aproject was initiated for the Development of a European multi-model ensemblesystem for seasonal to inter-annual prediction (DEMETER) by generating retro-spective forecasts (or simulations) as part of the EU project DEMETER (Palmeret al. 2004). The main objective of the project was to develop a well-validatedEuropean coupled multi-model ensemble forecast system for reliable seasonal tointerannual prediction. In this paper, comparative study of the seasonal mean andinterannual variability of Indian summer monsoon rainfall obtained from UK MetOffice’s Global Seasonal (GloSea) forecasting model during the summer monsoonseason from June to September (JJAS) has been carried out by taking 15 memberensemble forecast for 16 years (1987–2002). The rainfall for verification analysis isobtained from the global monthly precipitation using gauge observations, satelliteestimates and numerical model outputs (Xie and Arkin 1996).7.2 Details of Model and Model ProductsThe model hindcast is based on GloSea model, which is similar to the HadCM3climate version of the Met Office Unified Model, with a number of improvementsfor seasonal forecasting purposes. Details of the model physics and discussion ofthe performance of HadCM3 can be found in Gordon et al. (2000). The atmosphericcomponent is version HadAM3 (Pope et al. 2000), with a horizontal resolution of3.75� east-west and 2.5� north-south and 19 vertical levels. The oceanic componenthas 40 vertical levels (compared to 20 in HadCM3) with zonal grid spacing at 1.25�and meridional grid spacing of 0.3� near the equator increasing to 1.25� pole wardof the mid-latitudes (compared to 1.25� resolution east-west and north-south inHadCM3). A coastal tilting scheme has been included, to enable specifications ofthe land-sea mask at the ocean resolution. Like HadCM3, the GloSea coupled GCMcontains no flux corrections or relaxations to climatology. The model is initializedat 0000 UTC of 1st May and integrated for a period of 6 months in each yearseparately during the 16 years from 1987 to 2002. These experiments have been7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 103carried out in similar approach from 1987 to 2002 (16 years) to obtain monthlyas well as seasonal forecast of individual year. Each year simulation consists of15-member ensemble. The GloSea model simulated precipitation is validated withthe verification analysis rainfall obtained from the Xie-Arkin (1996) in thecorresponding month/season of same years. The different verification methodsused here are the correlation coefficient (CC), anomaly correlation co-efficient(ACC) and root mean square error (RMSE).Initial ocean conditions for the GloSea simulations are obtained by forcing theocean component with momentum, heat and fresh water fluxes from the ECMWFERA40 reanalysis. The method described by Palmer et al. (2004) has been used togenerate perturbed initial conditions for the ensemble members. Sets of wind stressand SST perturbations, designed to represent observed uncertainties in theseparameters, are pre-defined using historical differences between quasi-independentanalysis datasets. A series of wind stress perturbations, sampled from the pre-defined set, is applied with positive and negative signs to the ERA40 momentumfluxes and used in the ocean assimilation. In each simulation, instantaneous SSTperturbations from the predefined set are added to each of the assimilation seq-uences to obtain a 15-member ensemble (Graham et al. 2005). The SST pertur-bations are added to the temperature field in the top 40 m of the ocean. Thus,perturbations are applied to the ocean initial state only, with unperturbed ERA40analyses being used to initialize the model atmosphere and land components. Theperturbations are devised such that all the resulting members (perturbed or not) areequally likely descriptions of the system.7.3 Results and DiscussionThe performance of numerical models can be evaluated by comparing model-simulated parameters with the observations. Therefore, it is very important andmust to have observations for the verifications of the model performance. IndiaMeteorological Department (IMD) has a good observational network of rainfallobservations mainly spaced over land. On the other hand, numerical modelssimulate rainfall over land as well as over water body (over the whole domainof interest). Rajeevan et al. (2006) has recently generated a high resolution(1� � 1� lat./long.) gridded daily rainfall dataset for the Indian land region basedon IMD observations. Xie and Arkin (1996) had made an attempt to merge andreproduce rainfall distributions over land and ocean with the use of surfaceobservations, satellite data, buoys data and outputs from numerical models(GCM). Initially the observed gridded rainfall obtained from IMD is comparedwith the Xie-Arkin rainfall over Indian landmass to test the reliability of theXie-Arkin rainfall to be used for evaluation of the performance of the GloSeamodel in simulation of rainfall during summer monsoon over India.104 D.R. Pattanaik et al.7.3.1 Comparison of Xie-Arkin Rainfall with IMD (Observed)RainfallFigures 7.1 and 7.2 show the monthly climatology (during 1987–2002) of rainfallfor June, July, August and September obtained from IMD and Xie-Arkin respec-tively. The two maxima viz., one over west coast of India and other over the northeast India as reported in observed rainfall in June (Fig. 7.1a) is matching well overthe land region in Xie-Arkin rainfall (Fig. 7.2a). Similarly the active monsoonmonths of July and August with more rainfall (more than 7 mm/day) over thecentral India as reported in observed rainfall Fig. 7.1b, c is also well captured inXie-Arkin rainfall (Fig. 7.2b, c). The retreat phase of monsoon in September withJun rainfall clim (Observation)a cdb Jul rainfall clim (Observation)40N35N30N25N 447777104477 7447101020N15N10N5NEQ5S10S40N35N30N25N20N15N10N5NEQ5S10S50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E 50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110EAug rainfall clim (Observation)Sep rainfall clim (Observation)7Fig. 7.1 Spatial monthly climatological observed rainfall distribution (mm/day) obtained fromIMD for the 16 years period from 1987 to 2002 valid for (a) June (b) July (c) August and(d) September. Rainfall with more than 7 mm/day is shaded7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 105less rainfall over the central part of India and more over northeast India as reportedin the observation (Fig. 7.1d) is also almost identical in case of Xie-Arkin rainfall(Fig. 7.2d).Like the monthly rainfall the seasonal observed rainfall during June to Septem-ber (JJAS) as shown in Fig. 7.3a is also similar in distribution with that of meanJJAS Xie-Arkin rainfall (shown in Fig. 7.5a). Both the figures show identicalmaximum rainfall belt (more than 7 mm/day) over west coast of India and overthe northeastern and adjoining Indo-Gangetic belt. Thus the representation of Xie-Arkin rainfall is almost identical with the observed rainfall over Indian region.Therefore, Xie-Arkin rainfall can be used for the verification purpose over theIndian monsoon region covering the Indian land mass and the surrounding areas.Hence in following sub-sections the model forecast is compared with the Xie-ArkinJun rainfall clim (Xie–Arkin)a cdb40N7 441777777777777771010101010101544441110777777410101010101010101015151520777774444 117710101010101074777777 44441410 1010415441015101735N30N25N20N15N10N5NEQ5S10S40N35N30N25N20N15N10N5NEQ5S10S50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E 50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110EAug rainfall clim (Xie–Arkin)Sep rainfall clim (Xie–Arkin)Jul rainfall clim (Xie–Arkin)Fig. 7.2 Spatial monthly climatological observed rainfall distribution (mm/day) obtained fromthe verification analysis (Xie-Arkin) for the 16 years period from 1987 to 2002 valid for (a) June(b) July (c) August and (d) September. Rainfall with more than 7 mm/day is shaded106 D.R. Pattanaik et al.rainfall analysis (Xie-Arkin) (hereafter referred as verification analysis). Over theIndian land region, the interannual variability of seasonal rainfall obtained fromIMD observations showed many extreme years (Fig. 7.3b). The spatial rainfall-25-20-15-10-5051015202550E10S5SEQ5N10N15N20N25N30N35N40N60E 65E 70E 80E 85E 90E 95E 100E 105E 110E 115E55E1981198319851987198919911993199519971999200120032005YearSeasonal Rainfall Departure (%) AISMRSeasonal IMD Observed Rainfall(1987–2002)(mm/day)baFig. 7.3 (a) Spatial monthly climatological observed rainfall distribution (mm/day) obtainedfrom IMD for the 16 years period from 1987 to 2002 with rainfall > 7 mm/day is shaded.(b) The standardised anomalies JJAS rainfall from IMD7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 107anomaly during twoProduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 409Jagadish Singh and R.K. Giri32 Impact of Drought and Flood on Indian Food Grain Production . . . . 421Ajay Singh, Vinayak S. Phadke, and Anand Patwardhan33 Chinese Extreme Climate Events and AgriculturalMeteorological Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435Chen Huailiang, Zhang Hongwei, and Xue Changying34 Comparison of Sensible Heat Flux as Measured by Surface LayerScintillometer and Eddy Covariance Methods Under DifferentAtmospheric Stability Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461G.O. Odhiambo and M.J. Savage35 Crop Water Satisfaction Analysis for Maize Trial Sitesin Makhado During the 2007/2008 Season . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485M.E. Moeletsi, N.S. Mpandeli, and E.A.R. Mellaart36 Prediction of Mungbean Phenology of Various Genotypes UnderVarying Dates of Sowing Using Different Thermal Indices . . . . . . . . . 491K.K. Gill, Guriqbal Singh, G.S. Bains, and Ritu37 Effect of Thermal Regimes on Crop Growth, Developmentand Seed Yield of Chickpea (Cicer Arietinum L.) . . . . . . . . . . . . . . . . . . . . 499K.K. Agrawal, U.P.S. Bhadauria, and Sanjay Jain38 Stomatal Adaptation and Leaf Marker Accumulation Patternfrom Altered Light Availability Regimes: A Field Study . . . . . . . . . . . . 505K. Ramesh, S. Raj Kumar, and Virendra Singh39 Comparative Study of Diurnal Rate of Photosynthesis at VariousLevels of Carbon Dioxide Concentration for Different Crops . . . . . . 511A. Kashyapi, Archana P. Hage, and R.P. Samuixiv Contents40 Effect of Weather Variability on Growth Characteristicsof Brassica Crop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519Ananta Vashisth, N.V.K. Chakravarty, Goutam Bhagavati,and P. K. Sharma41 Agronomic Impacts of Climate Variability on Rice Productionwith Special Emphasis on Precipitation in South Western Plainsof Uttarakhand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529S.K. Tripathi and B. Chintamanie42 Selection of Suitable Planting Method and Nutrient ManagementTechniques for Reducing Methane Flux from Rice Fields . . . . . . . . . . 539Venkatesh Bharadwaj, A.K. Mishra, S.K. Singh, S.P. Pachauri,and P.P. Singh43 Operational Agrometeorological Strategies in Different Regionsof the World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551M.V.K. Sivakumar44 Overview of the World Agrometeorological Information Service(WAMIS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573Robert Stefanski45 Analysis of Rainfall Variability and Characteristics of RainfedRice Condition in Eastern India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579P.K. Singh, L.S. Rathore, K.K. Singh, and B. AthiyamanIndex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 597Contents xv.ContributorsK. K. Agrawal Department of Physics and Agrometeorology, College of Agricul-tural Engineering, JNKVV, Jabalpur, MP 482004, India, kkagrawal59@yahoo.co.inB. Athiyaman National Center for Medium Range Weather Forecasting, NOIDA201 307, India, athiya@ncmrwf.gov.inS. D. Attri India Meteorological Department, New Delhi-110003, India,sdattri@gmail.comG. S. Bains Department of Agricultural Meteorology, Punjab Agricultural Univer-sity, Ludhiana, IndiaU. P. S. Bhadauria Department of Physics and Agrometeorology, College of Agri-cultural Engineering, JNKVV, Jabalpur, MP 482004, India, upsb2007@rediffmail.comGoutam Bhagavati Division of Agricultural Physics, Indian AgriculturalResearch Institute, New Delhi 110012, India, goutombhagawati@gmail.comS. C. Bhan India Meteorological Department, Lodi Road, New Delhi 110 003,India, scbhan@gmail.comVenkatesh Bharadwaj Department of Agrometeorology, College of Agriculture,G.B. Pant University of Agriculture and Technology, U.S. Nagar, Pantnagar,Uttarakhand 263145, India, dr.venkatbh@rediffmail.comB. K. Bhattacharya Space Applications Centre (ISRO), Ahmedabad 380 015,India, bkbhattacharya@sac.isro.gov.inS.K. Roy Bhowmik India Meteorological Department, New Delhi 110 003, India,skrb.imd@gmail.comxviiRyan P. Boyles State Climate Office of North Carolina, NC State University,Box 7236, Raleigh, NC 27695-7236, USA, ryan_boyles@ncsu.eduMark S. Brooks State Climate Office of North Carolina, NC State University,Box 7236, Raleigh, NC 27695-7236, USA, mark_brooks@ncsu.eduAnca Brookshaw Met Office, Exeter, United Kingdom, anca.brookshaw@metoffice.gov.ukN.V.K. Chakravarty Division of Agricultural Physics, Indian AgriculturalResearch Institute, New Delhi 110012, India, nvkchak@iari.res.inXue Changying Henan Institute of Meteorological Sciences, Zhengzhou 450003,China, xuecy9@163.comN. Chattopadhyay Agricultural Meteorology Division, India MeteorologicalDepartment, Pune, India, n.chattopadhyay@imd.gov.inJ.L. Chaudhary S.G. College of Agriculture and Research Station, Kumhrawand,Jagdalpur, Chhattisgarh 494 005, India, zars_igau@rediffmail.comB. Chintamanie Department of Water Resources Development and Management,Indian Institute of Technology Roorkee, Roorkee, India, bcrchintamanie@gmail.comI.M. L. Das M. N. Saha Centre of Space Studies, University of Allahabad,Allahabad 211 002, India, profimldas@yahoo.comS.K. Dash Centre for Atmospheric Sciences, Indian Institute of Technology Delhi,Hauz Khas, New Delhi-110016, India, skdash@cas.iitd.ac.inSomenath Dutta India Meteorological Department, Pune, India, dutta.drsomenath@gmail.comAshley N. Frazier State Climate Office of North Carolina, NC State University,Box 7236, Raleigh, NC 27695-7236, USA, anfrazier@gmail.comK.K. Gill Department of Agricultural Meteorology, Punjab Agricultural University,Ludhiana, India, kgill2002@gmail.comR.K. Giri India Meteorological Department, Lodi Road, New Delhi 110003, India,rkgiri_ccs@rediffmail.comxviii ContributorsArchana P. Hage Agricultural Meteorology Division, India MeteorologicalDepartment, Pune, IndiaZhang Hongwei Henan Institute of Meteorological Sciences, Zhengzhou 450003,China, xxqxjzhw1966@163.comChen Huailiang Henan Institute of Meteorological Sciences, 110 Jinshuilu Rd,Zhengzhou, Henan, PR 450003, China, chenhl@cam.gov.chSanjay Jain Department of Physics and Agrometeorology, College of AgriculturalEngineering, JNKVV, Jabalpur, MP 482004, India, genomics_san@hotmail.comManish K. Joshi K. Banerjee Centre of Atmospheric & Ocean Studies, Institute ofInterdisciplinary Studies, University of Allahabad, Allahabad 211 002, India,manishkumarjoshi@gmail.comP.C. Joshi Space Applications Centre (ISRO), Ahmedabad 380 015, India,pcjoshi35@hotmail.comN. Kale Anand Agricultural University, Anand, Gujarat 388110, India, yogirajve-dashram@gmail.comM.V. Kamble Agricultural Meteorology Division, India Meteorological Depart-ment, Pune, India, mdhr_kmbl@yahoo.comSarat C. Kar National Centre for Medium Range Weather Forecasting, Ministryof Earth Sciences, A-50, Sector-62, NOIDA, UP, India, sckar@ncmrwf.gov.inB.I. Karande Anand Agricultural University, Anand, Gujarat 388110, India,babankarande@yahoo.co.inA. Kashyapi Agricultural Meteorology Division, India Meteorological Depart-ment, Pune, India, kashyapi_a@yahoo.co.inS. Korsakova Center for Hydrometeorology of the Autonomous Republic of theCrimea, Agrometeorological Station of the Ukraine, State Committee for Hydro-meteorology, Nikitskij Sad, Yalta, Ukraine, korsakova@i.uaO. Krishnacontrasting year (1987 is deficit year and 1988 is excess year)and also the recent deficient year of 2002 as seen in the observed rainfall is verymuch identical with the Xie-Arkin rainfall (Fig. not shown).7.3.2 Simulation of Mean MonsoonThe model climatology is represented here by retrospective forecasts (or “modelsimulations”), made with a 15-member ensemble, over the 16-year period from1987 to 2002. Therefore, for each new forecast, there is a reference set of 240(15 � 16) simulations. As shown in Fig. 7.2 for the Xie-Arkin rainfall thecorresponding spatial map of monthly means forecast rainfall from 240 membersduring June to September is shown in Fig. 7.4. The rainfall map plotted in Figs. 7.2and 7.4 is in mm/day and a value more than 7 mm/day is shaded. It is seen fromFigs. 7.2 and 7.4 that the forecast climatology during the month of June (Fig. 7.4a)is almost identical with verification (Fig. 7.2a) climatology with two maxima, onenear the west coast of India and other over the head Bay of Bengal. However, thewest coast maximum is slightly under estimated and it is confined only over smallerarea at northern portion of the coast in forecast climatology (Fig. 7.4a). During themonth of July the forecast climatology (Fig. 7.4b) shows similar patterns withcorresponding verification climatology (Fig. 7.2b). The rainfall maxima over thehead Bay of Bengal region and west coast of India and the rain shadow regions ofTamilnadu and north western part of India in the month of July are well captured inforecast climatology (Fig. 7.4b). However, the model simulation shows that thewest coast maximum is under estimated and it is confined only to northern portionin forecast climatology (Fig. 7.4b) similar to the pattern of June forecast climatol-ogy (Fig. 7.4a). It is noticed that model simulated rainfall for the month of August(Fig. 7.4c) is in good agreement with the verification rainfall climatology (Fig. 7.2c)with two maxima, one over head Bay of Bengal and other over west coast of India.The two rain shadow regions such as Tamilnadu and north western India are alsowell simulated by the model in the month of August (Fig. 7.4c). The rainfall patternduring August is almost similar with that of July rainfall patterns, however, slightreduction in rainfall over the main land of India is noticed both in forecast(Fig. 7.4c) as well as verification climatology (Fig. 7.2c). From Fig. 7.2d it isnoticed that the rainfall distribution indicate the withdrawal features of monsoonduring September. During the withdrawal phase of monsoon in September theforecast climatology (Fig. 7.4d) is in close resemblance with the verificationclimatology (Fig. 7.2d). However, rainfall simulated with GloSea model isoverestimated in the north central India over the Gangetic and Brahmaputra valleystretching from the Bay of Bengal region particularly during September and duringthe peak monsoon months of July, August (Fig. 7.4b–d). On the other hand, modelsimulated rainfall is underestimated over the west coast of India during July,August and September. Like the monthly pattern, the seasonal climatology (Juneto September; JJAS) from model forecast (Fig. 7.5b) shows almost similar rainfall108 D.R. Pattanaik et al.distribution pattern with rainfall climatology obtained from the verification analysis(Fig. 7.5a) and is in close resemblance in terms of magnitude as well as spatialdistribution. The coupled model could capture both the rainfall maxima over thewest-coast of India and north Bay of Bengal region which is in good agreement withthe verification analysis. Like in the monthly forecast rainfall during July toSeptember as shown in Fig. 7.4b–d, the seasonal forecast rainfall also showsoverestimation in the north central India over Gangetic and Brahmaputra valleystretching from the Bay of Bengal region. These features are shown in Fig. 7.5c bytaking difference between verification and forecast rainfall over that region, whichindicates negative difference over the north central India over Gangetic andBrahmaputra valley stretching from the Bay of Bengal region.In order to study the pattern correlation of verification and model forecast outputover the different regions, the seasonal (JJAS) climatology of verification analysis40NJun rainfall climatology(UKMO)Aug rainfall climatology(UKMO)Jul rainfall climatology(UKMO)Sep rainfall climatology(UKMO)7 777777777777177101010101010101015 15154444435N30N25N20N15N10N5NEQ5S10S40N35N30N25N20N15N10N5NEQ5S10S40N35N30N25N20N15N10N5NEQ5S10S40N35N30N25N20N15N10N5NEQ5S10S50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E105E110E 50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E105E110E50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E105E110E50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E105E110E710acbd1510104Fig. 7.4 Spatial monthly climatological UKMet office’s model forecast rainfall distribution (mm/day) for the 16 years period from 1987 to 2002 valid for (a) June (b) July (c) August and (d)September. Rainfall with more than 7 mm/day is shaded7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 109is correlated with seasonal climatology of model forecast and the correspondingCorrelation Coefficients (CCs) are given in Table 7.1. The CCs for 3 month forecast(June to August; JJA) is also included in Table 7.1. It is seen from Table 7.1 that the10S5SEQ5N10N15N20N25N30N35N40NJJAS (Verification Analysis – Forecast 50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E0–3–3–6–3–3–33 333 300000 00010S5SEQ5N10N15N20N25N30N35N40NJJAS Xie–Arkin rainfall climatology50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E110101010101044477777777a10S5SEQ5N10N15N20N25N30N35N40NJJAS forecast rainfall climatology50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E1101010104447777777bcFig. 7.5 Spatial seasonal (June–September) climatological rainfall distribution (mm/day) for the16 years period from 1987 to 2002 valid for (a) from (Xie-Arkin) (b) UK Met Office’s modelforecast and (c) the difference (a)–(b). Rainfall more than 7 mm/day is shaded in ‘(a)’ and ‘(b)’and the negative values are shaded in ‘(c)’110 D.R. Pattanaik et al.mean patterns show significant CC over the global tropics (0.91) followed by Indianmonsoon region (0.79) and even also for the smaller Indian region (0.72). Though,the mean seasonal rainfall is slightly more in model forecast than verification, theseasonal climatology of model simulation is similar with that of verificationclimatology over the entire tropical belt including the Indian monsoon region.7.3.3 Simulation of Interannual VariabilityThe General Circulation Model (GCM) has an “average” behaviour or climatologysimilar to real atmosphere in large spatial and temporal scale. As model integrationproceeds, there is a tendency for results to increasingly resemble the model clima-tology, introducing a systematic bias into the forecasts. To remove this bias,forecasts are expressed in terms of deviations from the GCM’s own climatology -a process referred to as calibration. In any current prediction, the deviation of themodel from its own climatology provides the forecast of how the real atmosphere isexpected to deviate from real climatology.The year-to-year variation of Indian rainfall from 1987 to 2002 indicates manyextreme years and also the normal years. The standardised anomalies of Indiansummer monsoon rainfall obtained from the India Meteorological Department(IMD) as shown in Fig. 7.3b indicates many extreme years during these 16 yearsmentioned above. Based on the criteria of �1 Standard Deviation, 1987and 2002are considered to be deficit year and 1988 and 1994 are considered to be excessyears during this period. In order to study the interannual variability of modelsimulation the year-to-year variation of JJAS total rainfall for 15-member ensembleover the entire Indian monsoon region bounded by 50�E–110�E and 10S�N–35�Nfrom 1987 to 2002 for the verification analysis and the model forecast is illustratedin Fig. 7.6. The Indian monsoon region is considered as the study area, which is thesame as chosen by Krishnamurti et al. (2004) for verification of DEMETER results.In case of model forecast the rainfall from 1987 to 2002 plotted in Fig. 7.6 is thesimple ensemble mean of 15 members in each year. It is seen that the mean is morein case of model’s forecast (711.7 mm) against the observation (683.4 mm). Thecorrelation between observed and forecast rainfall plotted in Fig. 7.6 is found to be0.43. However, over the land only region of India the seasonal mean forecastrainfall during JJAS is about 20% less compared to the observed mean JJAS rainfallTable 7.1 The Correlation Coefficient (CC) and Root Mean Square Error (RMSE) betweenobserved rainfall climatology and model hindcast climatology during the 16 years period from1987 to 2002 for June to August (JJA) and June–September (JJAS)RegionsCorrelation coefficient RMSE (mm/day)JJA JJAS JJA JJASIndian region (70�E–95�E, 5�N–35�N) 0.72 0.76 2.56 2.17Indian monsoon region (50�E–110�E, 10�S–35�N) 0.79 0.83 1.81 1.68Global tropics (0–360�E, 30�S–30�N) 0.91 0.91 1.00 1.017 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 111of 889 mm obtained from IMD observation. Similarly the CC between IMD rainfalland the forecast rainfall over the land only region of India is also poor. It is also seenfrom Fig. 7.6 that the monsoon rainfall obtained from the model during few yearssuch as during 1987, 1995, 1996, 1997, 1998 are very close to verification analysis.However, it is noticed from Fig. 7.6 that significant difference occurs betweenverification and forecast rainfall during some years (1988, 1991, 1999 and 2001).It may be mentioned here that although many earlier studies (Sperber et al. 2001;Ji and Vernekar 2000) have noted poor performance of forecasts in prediction ofAsian monsoon a season in advance, the skill of the forecast shown here is hopefulsince the seasonal as well as monthly forecast is well simulated in the model.In order to study the skill of seasonal climate forecasts for individual seasons, theverification and forecast anomaly for a few recent years (1994, 1997, 1998 and2002) have been plotted and are shown in Figs.7.7–7.10 respectively. Out of these4 years, 1994 is an excess monsoon rainfall year with JJAS rainfall departure ofþ10%, 2002 is a deficient monsoon year with JJAS rainfall departure of �19% and1997 & 1998 are normal years with JJAS rainfall departure of þ2% and þ5%respectively. While plotting these anomalies a local linear bias correction is applied(a posteriori) by expressing the forecasts relative to the model climatology, asdefined from 225 realizations (15 � 15 members). In all cases, the cross-validationtechnique is used, in which the (15-year) reference climatology for each simulationis constructed without the year of the simulation itself. Cross-validation simulatestypical real-time operational practice of using a fixed period for the reference model5005506006507007508008501987198819891990199119921993199419951996199719981999200020012002YearTotal rainfall (mm)Xie-Arkin modelmean=683.4 mm mean=711.7 mmFig. 7.6 Year-to-year variation of mean rainfall (mm) from observation (Xie-Arkin) and MetOffice’s model simulations for the Indian monsoon region (50�E–110�E, 10�S–35�N) during the16 years period from 1987 to 2002112 D.R. Pattanaik et al.climatology, which does not extend to the current forecast year, and thereforeprovides a more faithful representation of real-time skill. The rainfall receivedover India during the monsoon season (JJAS) in 1994 is excess which is alsoindicated by the verification JJAS rainfall anomaly (Fig. 7.7a). The rainfall obtainedfrom the model forecast also shows large positive anomalies over most parts of thecountry during monsoon season in 1994 (Fig. 7.7b). It is found that many El Ninoyears are associated with negative rainfall departure over India but one strong ElNino year of 1997 is associated with positive departure of rainfall (Fig. 7.3b).Slingo and Annamalai (2000) while examining the response of Indian summer10S5SEQ5N10N15N20N25N30N35N40N2222–2–40000 00000 000000JJAS forecast rainfall anomaly (1994)10S5SEQ5N10N15N20N25N30N35N40N000 2222–2–6–6–4 –2–2–224444200000000000JJAS Xie–Arkin rainfall anomaly (1994)50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55EabFig. 7.7 Seasonal (June–September; JJAS) rainfall anomalies (positive anomalies are shaded) for1994 in mm/day valid for (a) observation (Xie-Arkin) (b) UK Met Office’s model forecast7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 113monsoon to the major El Nino, they found that sea surface temperature (SST)anomalies associated with strong El Nino do not always affect the Indian monsoonrainfall in the same manner like the year 1997 when the monsoon rainfall was abovenormal over many parts of India (Fig. 7.8a). However, the forecast anomaly for theyear 1997 shows negative anomalies over most parts of the country (Fig. 7.8b).Followed by the El Nino year of 1997, the year 1998 was moderate La Nina yearand the monsoon rainfall over India was with positive departure of rainfall(Fig. 7.3b). The verification rainfall during 1998 shows positive departure overnortheast India, northern India, west coast of India and southern peninsular India10S5SEQ5N10N15N20N25N30N35N40N10S5SEQ5N10N15N20N25N30N35N40N50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55EJJAS Xie–Arkin rainfall anomaly (1997)JJAS forecast rainfall anomaly (1997)0000022–2–2–2–2–2–2–2–2–4–6–80 00000000000000000000000abFig. 7.8 Seasonal (June–September; JJAS) rainfall anomalies (positive anomalies are shaded) for1997 in mm/day valid for (a) observation (Xie-Arkin) (b) UK Met Office’s model forecast114 D.R. Pattanaik et al.(Fig. 7.9a). In case of forecast departure (Fig. 7.9b) the patterns are in wellagreement with that of observed anomaly (Fig. 7.9a) over most parts of Indiaexcept some parts of northeast India. During the recent deficient year of 2002, theverification rainfall departure is negative over the entire country except some partsof north east India (Fig. 7.10a). It may be mentioned here that none of the globalmodel could able to predict this large deficiency of rainfall over India during 2002.The model forecast here could able to show negative departure over most partsof the country (Fig. 7.10b). However, a belt of excessive rains in the interior ofIndia over the Gangetic and Brahmaputra valley persists in forecast departure10S5SEQ5N10N15N20N25N30N35N40N10S5SEQ5N10N15N20N25N30N35N40N50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E0000000000000 000000000000064200000222 2–2–2–6–426844JJAS Xie–Arkin rainfall anomaly (1998)JJAS forecast rainfall anomaly (1998)abFig. 7.9 Seasonal (June–September; JJAS) rainfallanomalies (positive anomalies are shaded) for1998 in mm/day valid for (a) observation (Xie-Arkin) (b) UK Met Office’s model forecast7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 115(Fig. 7.10b) though it is not seen in verification anomaly. As shown earlier similarpatterns of excessive rainfall belts over the Gangetic and Brahmaputra valley alsopersists in the model forecast during July to September on monthly scale (Fig. 7.2)and during the season as a whole (Fig. 7.5b). In another contrasting El niño and Laniña years of 1987 and 1988 (Fig. not shown), the rainfall anomalies obtained fromthe model forecast is in close resemblance with the verification anomalies over10S5SEQ5N10N15N20N25N30N35N40N10S5SEQ5N10N15N20N25N30N35N40N50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E50E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E55E000000200000000002–2–4–2–2–2–2–22–20000000 0JJAS Xie–Arkin rainfall anomaly (2002)JJAS forecast rainfall anomaly (2002)abFig. 7.10 Seasonal (June–September; JJAS) rainfall anomalies (positive anomalies are shaded)for 2002 in mm/day valid for (a) observation (Xie-Arkin) (b) UK Met Office’s model forecast116 D.R. Pattanaik et al.India in El Nino year of 1987 whereas, the excess monsoon rainfall of 1988 is notcaptured well in the model.7.3.4 Precipitation Forecast SkillThe skill of seasonal anomaly forecast has been evaluated on year-to-year from1987 to 2002 based on the Anomaly Correlation Coefficient (ACC) and Root MeanSquare Error (RMSE) between verification and forecast rainfall over the Indianmonsoon region and global tropics. The ACC and RMSE on year-to-year basis havebeen calculated and are shown in Figs. 7.11 and 7.12 respectively. In addition, ACCcalculated for June to September (JJAS) and June to August (JJA) are also given inFigs. 7.11 and 7.12. For calculating ACCs between verification and forecastrainfall, the JJAS forecast rainfall is interpolated to 2.50 � 2.50 lat-long gridscoincide with the verification rainfall grids obtained from Xie-Arkin during all16 years. It is seen from Fig. 7.11 that the ACCs vary from year to year both overthe Indian monsoon region (Fig. 7.11a) as well as the global tropics (Fig. 7.11b)bounded by 0–360�E, 30�S–30�N. It is found that over the global tropics ACCsvalues are comparatively higher than that of the corresponding ACCs over theIndian monsoon region during most of the years. Over the Indian monsoon region,though most of the years indicate positive ACCs (in some years ACCs > 0.6),slight negative ACCs are also noticed during few years as indicated in Fig. 7.11a.The negative ACCs could be due to mismatching of spatial distribution of modelsimulated anomaly and verification anomaly over the Indian monsoon region. TheACCs value during JJAS is the maximum with positive sign for the year 1998followed by 1994, 1995, 1992, 1991 and 2002. Thus, the higher values of ACCs canprovide some useful guidance on whether to expect above or below normalmonsoon rainfall a season in advance. Though it is also seen from Fig. 7.11a, bthat ACCs over global tropics are almost same during JJA and JJAS, but a littledifferences are present over the Indian monsoon region (Fig. 7.11a) particularly in1987, 1990, 1991. It is observed that the ACCs is more over the Indian monsoonregion during JJAS compared to JJA for the years 1987, 1989, 1991, 1994, 1997,1998 and less in the rest of the years of the 16 years period. Similarly, as shown inFig. 7.12, the RMSE value is also little higher over the Indian monsoon region(more than 3.0 mm/day) during some years compared to RMSE values overthe tropics (less than 1.5 mm/day). Unlike ACCs, the RMSE value is higher overthe Indian monsoon region in the season JJAS compared to that of JJA during all theyears. Comparing the ACCs and RMSE values in each individual years, it is seenfrom Figs. 7.11 and 7.12 that in some individual years like 2002 & 2001, the RMSEis comparatively lower than other years with the magnitude of corresponding ACCis very small. Thus, for the superior skill of the seasonal forecast smaller RMSE isneeded and a slight improvement for the ACC for seasonal forecasts of the monsoonis required and more efforts could be made to achieve the same.7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 117-0.4-0.200.20.40.60.81987198819891990199119921993199419951996199719981999200020012002YearAnomaly Correlation Coefficient (ACC)JJAJJASb-0.4-0.200.20.40.60.81987198819891990199119921993199419951996199719981999200020012002YearAnomaly Correlation Coefficient (ACC)JJAJJASaFig. 7.11 Seasonal (June–August; JJA and June to September; JJAS) Anomaly CorrelationCoefficient (ACC) for the 16 years period from 1987 to 2002 valid for (a) the Indian monsoonregion (50�E–110�E, 10�S–35�N) and (b) global tropics118 D.R. Pattanaik et al.00.511.522.533.51987198819891990199119921993199419951996199719981999200020012002YearRoot Mean Square Error (RMSE)JJAJJAS00.511.522.533.51987198819891990199119921993199419951996199719981999200020012002YearRoot Mean Square Error (RMSE)JJAJJASabFig. 7.12 Seasonal (June–August; JJA and June–September; JJAS) Root Mean Square error(RMSE) for the 16 years period from 1987 to 2002 valid for (a) the Indian monsoon region(50�E–110�E, 10�S–35�N) and (b) global tropics7 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 119It can be stated here that the present analysis of UKMO model (GloSea3)simulation shows reasonable skill of rainfall (in global tropics as well as in Indianmonsoon rainfall region), although the skill is poor over the land-only region ofIndia continent. The real time forecast can be used in refined manner (probabilisticforecast) by using suitable statistical techniques over the Indian monsoon region.The recently upgraded UKMO seasonal forecast model (GloSea4), which hasHadGEM3 as its atmospheric component with some additional improvement isexpected to do better over the India monsoon region.7.4 Summary and DiscussionThe UK Met office monthly (from June to September) as well as seasonal (JJAS)forecasted rainfall from 240 members during 16 years period (1987–2002) iscompared with the rainfall obtained from the Xie-Arkin (verification analysis)during the same period. From the above results, following broad conclusions canbe drawn:The UK Met office’s model is well efficient in simulation of rainfall in themonthly forecast climatology from June to September as well as in the seasonalforecast climatology (JJAS) with two maxima one over the west coast of Indiaregion and other over the head Bay of Bengal. In model simulation, the west-coastrainfall maximum is slightly underestimated in Met Office’s simulation and con-fined to only northern part of west-coast of India with smaller area. This may be dueto model performance where model could not able to simulate locally organizedconvection properly over the Western Ghats mountain region during monsoonseason. Though the model able to capture seasonal climatology very well, a beltof excessive rain in the interior of India over the Gangetic and Brahmaputra valleystretching from north Bay of Bengal persists in case of seasonal climatology. It isseen that the climatology mean rainfall is more in case of model’s simulation(711.7 mm) against the verificationanalysis (683.4 mm) over the entire Indianmonsoon region bounded by 50�E–110�E and 10�S–35�N. However, over the landonly region of India the seasonal mean model forecast rainfall is about 20% lessthan the corresponding observed rainfall. The correlation coefficient betweenverification and forecast climatology during JJAS and JJA are in close resemblanceover the Global tropics and Indian monsoon region with significant CCs. This resultalso supports earlier study that the predictability and the interannual variability ofAsian summer monsoon rainfall in a model are dependent on the model simulationof the climatological mean monsoon (Sperber and Palmer 1996).The year-to-year variations of forecast ensemble mean of JJAS total rainfall overthe Indian monsoon region bounded by 50�E–110�E, 10�S–35�N show similarbehaviour in some years with a CC of 0.43 during the 16 years period from 1987to 2002. However, the correlation is poor when land only of India is considered. Inthis study, the Anomaly Correlation Coefficients (ACCs) between verification andforecast rainfall over Indian monsoon region during 1987–2002 indicate large120 D.R. Pattanaik et al.positive value (more than 0.6) during many years. Similarly, the Root Mean SquareError (RMSE) also varies from 1.5 to about 3.3 mm/day over the Indian monsoonregion during the period from 1987 to 2002. For the superior skill of the seasonalforecast, further improvement is needed to have smaller RMSE and a slightimprovement for the ACC for seasonal forecasts of the monsoon.Acknowledgements The authors are thankful to the Director General of Meteorology forencouragement and for providing all facility to carry out this study. Thanks to UK Met officefor providing the hindcast from GloSea model. Thanks are also due to NASA for making availablethe TRMM rainfall data used in the present study.ReferencesBlanford HF (1884) On the connection of the Himalayan snowfall with dry winds and seasons ofdraughts in India. Proc R Soc Lond 37:3–22Brankovic C, Palmer TN (1997) Atmospheric seasonal predictability and estimates of ensemblesize. Mon Weather Rev 125:859–874Brankovic C, Palmer TN, Molenti F, Tibaldi S, Cubasch U (1990) Extended range predictions withECMWF models: time lagged ensemble forecasting. Q J R Meteorol Soc 116:867–912Chakraborty A, Krishnamurti TN (2006) Improved seasonal climate forecasts of the South Asiansummer monsoon using a suite of 13 coupled ocean-atmosphere models. Mon Weather Rev134(6):1697–1721Charney JG, Shukla J (1981) Predictability of monsoons. In: Lighthill J, Perce RP (eds) Monsoondynamics. Cambridge University Press, Cambridge, pp 99–109Chen TC, Yen MC (1994) Interannual variation of the Indian monsoon simulated by the NCARCommunity Climate Model: effect of the tropical Pacific SST. J Clim 7:1403–1415Doblas-Reyes FJ, Déqué M, Piedelievre JP (2000) Multimodel spread and probabilistic forecastsin PROVOST. Q J R Meteorol Soc 126:2069–2087Gordon C, Cooper C, Senior CA, Banks H, Gregory JM, Johns TC, Mitchell JFB, Wood RA(2000) The simulation of SST, sea ice extents and ocean heat transports in a version of theHadley Centre coupled model without flux adjustments. Clim Dyn 16:147–168Gowariker V, Thapliyal V, Kulshrestha SM, Mandal GS, Sen Roy N, Sikka DR (1991) A powerregression model for long-range forecast of southwest monsoon rainfall over India. Mausam42:125–130Graham RJ, Gordon M, Mclean PJ, Ineson S, Huddleston MR, Davey MK, Brookshaw A, BarnesR (2005) A performance comparison of coupled and uncoupled versions of the Met Officeseasonal prediction General Circulation Model. Tellus 57A:320–339Ji Y, Vernekar AD (2000) Simulations of the Asian summer monsoons of 1987 and 1988 with aregional model nested in a global GCM. J Clim 10(8):1965–1979Kang IS, Jin K, Wang B, Lau KM, Shukla J, Krishnamurthy V, Schubert SD, Waliser DE, SternWF, Kitoh A, Meehl GA, Kanamitsu M, Galin VY, Satyan V, Park CK, Liu Y (2002)Intercomparison of the climatological variations of Asian summer monsoon precipitationsimulated by 10 GCMs. Clim Dyn 19(5–6):383–395Kharin VV, Zwiers FW (2002) Climate predictions with multimodel ensembles. J Clim 15:793–799Krishnamurti TN, Kishtawal CM, LaRow TE, Bachiochi DR, Zhang Z, Williford CE, Gadgil S,Surendran S (1999) Improved weather and seasonal climate forecasts from multimodelsuperensemble. Science 285:1548–15507 An Evaluation of the Simulation of Monthly to Seasonal Summer Monsoon Rainfall 121Krishnamurti TN, Mitra AK, Yun WT, Vijaya Kumar TSV, Dewar WK (2004) Seasonal climateforecasting of the Asian monsoon using multiple coupled models. FSU Report No. 04–05, June2004Lorenz EN (1969) Atmospheric predictability as revealed by naturally occurring analogues.J Atmos Sci 26:636–646Palmer TN, Anderson DLT (1994) The prospects for seasonal forecasting: a review paper. Q J RMeteorol Soc 126:2013–2033Palmer TN, Brankowic C, Viterbo P, Miller MJ (1992) Modelling interannual variations ofsummer monsoons. J Clim 5:399–417Palmer TN, Alessandri A, Andersen U, Cantelaube P, Davey M, Délécluse P, Déqué M, Dı́ez E,Doblas-Reyes FJ, Feddersen H, Graham R, Gualdi S, Guérémy JF, Hagedorn R, Hoshen M,Keenlyside N, Latif M, Lazar A, Maisonnave E, Marletto V, Morse AP, Orfila B, Rogel P,Terres JM, ThomsonMC (2004) Development of a European multi-model ensemble system forseasonal to inter-annual prediction (DEMETER). Bull Am Meteorol Soc 85:853–872Pattanaik DR, Kumar A (2010) Prediction of summer monsoon rainfall over India using the NCEPclimate forecast system. Clim Dyn 34:557–572Pavan V, Doblas-Reyes J (2000) Multimodel seasonal hindcasts over the Euro-Atlantic: skillscores and dynamic features. Clim Dyn 16:611–625Peng P, Kumar A, Van den Dool H, Barnston AG (2002) An analysis of multimodel ensemblepredictions for seasonal climate anomalies. J Geophys Res 107(D23):4710. doi:10.1029/2002JD002712Pope VD, Gallani ML, Rowntree PR, Stratton RA (2000) The impact of new physical parameter-isations in the Hadley centre climate model: HadAM3. Clim Dyn 16:123–146Rajeevan M, Bhate J, Kale JD, Lal B (2006) High resolution daily gridded rainfall data for theIndian region: analysis of break and active monsoon spells. Curr Sci 91:296–306Saha S, Nadiga S, Thiaw C, Wang J, Wang W, Zhang Q, Van den Dool HM, Pan HL, Moorthi S,Behringer D, Stokes D, Pena M, Lord S, White G, Ebisuzki W, Peng P, Xie P (2006) TheNCEP climate forecast system. J Clim 19:3483–3517Sahai AK, Pattanaik DR, Satyan V, Grimm AM, Alice M (2003) Teleconnections in recent timeand prediction of Indian summer monsoon rainfall. Meteorol Atmos Phys 83:217–227Shukla J, Mooley DA (1987) Emperical prediction of the summer monsoon rainfall over India.Mon Weather Rev 115:695–703Shukla J, Anderson J, Baumhefner D, Brankovic C, Chang Y, Kalnay E, Marx L, Palmer TN,Paolino DA, Ploshay J, Schubert S, Straus DM, Suarez M, Tribbia J (2000) Dynamicalseasonal prediction. Bull Am Meteorol Soc 81:2593–2606Slingo JM, Annamalai H (2000) The El Nino of the century and the response of the Indian summermonsoon. Mon Weather Rev 128:1778–1797Soman MK, Slingo J (1997) Sensitivity of Asian summer monsoon to aspects of sea surfacetemperature anomalies in the tropical Pacific Ocean. Q J R Meteorol Soc 123:309–336Sperber KR, Palmer T (1996) Interannual tropical rainfall variability in general circulation modelsimulations associated with the Atmospheric Model Intercomparison Project. J Clim 9:2727–2750Sperber KR, Brankovic C, Deque M, Frederiksen CS, Graham R, Kitoh A, Kobayashi C, Palmer T,Puri K, Tennant W, Volodin E (2001) Dynamical seasonal predictability of the Asian summermonsoon. Mon Weather Rev 129(9):2226–2248Stephenson DB, Doblas-Reyes FJ (2000) Statistical methods for interpreting Monte Carlo ensem-ble forecasts. Tellus 52A:300–322Wang B, Kang IS, Lee JY (2004) Ensemble simulations of Asian-Australianmonsoon variabilityby 11 AGCMS. J Clim 17(4):803–818Wu MLC, Schubert S, Kang IS, Waliser D (2002) Forced and free intraseasonal variability overthe south Asian monsoon region simulated by 10 AGCMs. J Clim 15(20):2862–2880Xie-Arkin (1996) Analyses of global monthly precipitation using gauge observations, satelliteestimates, and numerical model predictions. Journal of Climate 9:840–858122 D.R. Pattanaik et al.Chapter 8Prediction of Monsoon Variabilityand Subsequent Agricultural ProductionDuring El Niño/La Niña PeriodsM.V. Subrahmanyam, T. Satyanarayana, and K.P.R. Vittal MurthyAbstract It is well know that the south west monsoon or popularly known summermonsoon dictates the economy of the sub-continent. A good monsoon year resultsin good rainfall increased production and a boom in the economy on the contrarya bad monsoon year with deficit rainfall results in decrease in the yield and asubsequent reduction in agricultural production and economy of the country. IndianGovernment and Indian scientists did a wonderful job in increasing the agriculturalproduction to cater to the needs of increased population. However, in years of ElNiño the monsoon activity and the monsoon rainfall is below normal and results ina decrease in production. In order to get a sustainable development, this is the areawhere serious scientific measures are to be implemented to get sustained develop-ment. It is possible to achieve this difficult task, because El Niño sends a forewarn-ing signal in December with an increased abnormal SST of the coast of the Brazil inPacific. The aim and objective of the present article is to focus about the phenomenaof El Niño and La Niña, the mechanism the manifestation and the intensity. Thereduction in monsoon circulation in fluxes and rainfall. This was amply describedby estimating the oceanic fluxes, which will send clear signals as to the intensityof an El Niño and La Niña. In this article the yearly all India grain productionfrom 1953 to 2007 was also examined. The increasing trend can be interpretedas the proper agro technical measures and use of complex fertilizers (intensiveM.V. Subrahmanyam (*)South China Sea Institute of Oceanology, Chinese Academy of Science, Beijing, Chinae-mail: mvsm.au@gmail.comT. SatyanarayanaDepartment of Environmental Sciences, Acharya Nagarjuna University, Guntur, AP, IndiaandCentral Research Institute for Dryland Agriculture, Hyderabad 500 059, Indiae-mail: satya_1006@yahoo.co.inK.P.R.V. MurthyDepartment of Meteorology and Oceanography, Andhra University, Visakhapatnam, AP, Indiae-mail: kprvm@yahoo.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_8, # Springer-Verlag Berlin Heidelberg 2011123cultivation) besides increase in the land under cultivation (bringing waste lands intoirrigated fields – extensive cultivation. From the sequential March of grain produc-tion it is seen that the El Niño years resulted in deficit in production. It is suggestedthat proper agro management techniques can at least decrease the deficit in produc-tion. Various management programs in this direction were also suggested.8.1 Introduction8.1.1 MonsoonMonsoon winds blow from cold to warm regions because cold air takes up morespace than warm air. Monsoons blow from the land toward the sea in winter andfrom the sea toward land in the summer. India’s winters are hot and dry. Themonsoon winds blow from the northeast and carry little moisture. The temperatureis high because the Himalayas form a barrier that prevents cold air from passingonto the subcontinent. The summer monsoons enter to the subcontinent from thesouthwest. The winds carry moisture from the Indian Ocean and bring heavy rainsfrom June to September. The summer monsoons are welcomed in India becausefarmers depend on the rains to irrigate their land.The Asian monsoon is characterized by a seasonal reversal of surface winds anda distinct seasonality of precipitation. The fundamental driving mechanisms ofthe monsoon cycle are the cross-equatorial pressure gradients resulting from differ-ential heating of land and ocean, modified by the rotation of the earth and theexchange of moisture between the ocean, atmosphere and land (Webster 1987).Strong seasonality of wind and rainfall patterns, the monsoon regions also experi-ence a high degree of variability on intrapersonal, interannual, and interdecadaltimescales (Clark et al. 2000). The variability of the Asian monsoon is linked withthe El Niño–Southern Oscillation (ENSO). The occurrence of El Niño is generallyassociated with a weak monsoon, and La Niña is associated with a strong monsoon(Webster and Yang 1992). The connection between the Asian monsoon and ENSOappears to be statistically nonstationary (Troup 1965).Several studies have been carried out on the role of air-sea interaction processesover the Indian Ocean (Cadet and Diehl 1984; Shukla and Mishra 1977; Weare1979; Rao and Goswami 1988; Mohanty and Ramesh 1993). These studies aremainly concentrated on the relationship between sea surface temperature anomaliesover Arabian Sea and summer monsoon rainfall over India. Shukla (1987) foundthat heavy (deficit) rainfall is followed by negative (positive) SST anomalies, butthe magnitude of the SST anomalies for the premonsoon months is small andinsignificant. Joseph and Pillai (1984) obtained the same results. On the contrary,the studies of Rao and Goswami (1988) suggest that SST in the southeast ArabianSea during the premonsoon season is significantly correlated with ISM rainfall. Thestudy of Ramesh Kumar et al. (1986) indicated that large negative anomalies off theSomali and Arabian coasts are associated with good monsoon rainfall over India.124 M.V. Subrahmanyam et al.The relationship between the SST in the Eastern Equatorial Indian Ocean (EEIO)and monsoon rainfall was studied by Sadhuram (1997). Their study indicated thatSST anomaly during October and November in the EEIO is useful for monsoonrainfall prediction.The origin and amount of moisture being transported to the Indian subcontinentduring the southwest monsoon season was also studied by many investigators.Pisharoty (1965) utilizing the data collected during IIOE, examined the moisturebudget and found that evaporation from Arabian Sea to be the main contributor forsummer monsoon rainfall. Saha and Bavadekar (1973) using additional upper airdata concluded that 70% of moisture flux from the south Indian ocean accountsfor the bulk of moisture needed for summer monsoon rainfall. The intraseasonalvariations of moisture budget have also been examined by several scientists(Ramamurthy et al. 1976; Ghosh et al. 1978; Howland and Sikdar 1983; Cadetand Reverdin 1981a, Sadhuram and Ramesh Kumar 1988). While the importance ofevaporation over the Arabian Sea is suggested by Ghosh et al. (1978), Murakamiet al. (1984), the role of cross equatorial (Cadet and Reverdin 1981a, b; Howlandand Sikdar 1983; Sadhuram and Ramesh Kumar 1988; Ramesh Kumar andSadhuram 1989; Ramesh Kumar et al. 1999). Most of these studies suggest thatthe cross equatorial moisture flux provides an important source of moisture forthe ISM rainfall though the evaporation from the Arabian Sea is quite significant.Most of the above mentioned studies have used only limited data sets as no longterm observation were available. Now with the availability of daily and twice dailyreanalysis results from NCEP/NCAR project various aspects of monsoon can bestudied in detail which can throw more light on the different phases of monsoonactivity. In the present study, an attempt is made to understand the role of low levelflow on the ISM rainfall. In the present study, the possible linkages between theIndian summer monsoon rainfall and surface meteorological fields (latent heat)were investigatedduring the El Niño years and La Niña years.India is an agricultural country, 80% of the population depends on agriculturalproduction. Indian Government has initiated number of steps to increase agricultureproduction, which is known as green revolution. On the agronomy side bettervarieties of agricultural seeds have been produced, which are high yielding, droughtresistant and reduce the crop life period. More hydrological products are initiated,so that the river waters are utilized in a better fashion, by constructing dams acrossthe rivers and providing water to more access of land, so that the agricultural land isincrease, which is known as extensive cultivation. Research was also on for betterutilization of water, so that water is utilized fully by the agricultural crop, but notwasted by percolation or de-percolation or by excessive runoff. Drip irrigation is anexample in this direction. Better agro technical measures and proper utilization ofspecified fertilizers (natural and Bio fertilizer is better than chemical fertilizer) andproper and accurate utilization of insecticides and pesticides will also ensureincrease yield per acre. So by the intensive and extensive cultivation of agriculturalproduction can be increased.In spite of the technological development mentioned above the agriculturalproduction in the country very much depends on the timely rainfall received from8 Prediction of Monsoon Variability and Subsequent Agricultural Production 125monsoons in normal quantities (extensive rainfall leads to floods deficit rainfallleads to drought). Both the disasters, there is substantial crop damage and loss ofproduction. El Niño and La Niña are related to SST off the coast of Brazil in Pacific.El Niña in Spanish means child Christ because this phenomena manifests thegreater half of December around Christmas La Niña (surprisingly means girl Christ)is opposite of El Niño. During El Niño, SSTs are more than normal during La Niña,they are less than normal respectively. Both the phenomena related to summermonsoon (ref). El Niño results in deficit rainfall and La Niña manifests excessiverainfall. As there is a gap of 6 months there is sufficient time to plan for the comingevent of El Niño and La Niña. One can change the cropping pattern and applyproper agro technical measures aimed at achieving sustainable development withthe production loses at a minimum level.8.2 Data MethodologyIn the present study, to find the possible linkages between the Indian summermonsoon rainfall and surface meteorological fields (latent heat) were investigatedduring the El Niño years and La Niña years. One of the reliable global data set isavailable from NCEP (National Centre for Environmental Prediction).NCEPReanalysis data set has been used for the present study. The data is having resolu-tion of 2.5� � 2.5� grid. We have used the latent heat flux of every month averageover Arabian Sea for the monthly analysis and mean of JJAS for seasonal analysis.El Niño and La Niña period chosen for this study is from the ENSO Years based onOceanic Niño Index (ONI) http://ggweather.com/enso/oni.htm. We have usedthe monthly and seasonal for all India and sub divisional wise rainfall data, whichare available at http://www.tropmet.res.in. All India food grain production from1953–2006 was utilized to study effect Indian monsoon rainfall on food grainproduction. The detrend food grain production data the relationship with rainfallwere analyzed.8.3 Results and Discussion8.3.1 Variation of Oceanic Latent Heat FluxThe southwest monsoon rain fall is having the impact on the Indian agriculture. Inthis article we studied how the Latent Heat flux is influencing the rainfall in theIndia. The mean variation of LHF anomaly during the monsoon season (JJAS) overArabian Sea has been given in the Fig. 8.1. Gradual decrease of red colour indicatesthe El Niño episodes and blue indicates the La Niña episodes. Gradual decrease ofcolor changes from Strong to weak episodes and the green colour indicates the126 M.V. Subrahmanyam et al.normal monsoon. There is a much variations observed since 1950 to early 1990s.After that the variation of anomalies are shows the positive. It indicates the LHFis increasing by its values after early 1990s. The trend for the LHF anomalies ispositive, it clearly indicates that the LHF is increasing may be due to globalwarming. The sea surface temperature (SST) in the Indian Ocean shows a warmingtrend in recent decades (Swadhin et al. 2007) can be related in increasing of LHF,which depends on, temperature, and wind. The North Indian Ocean shows SSTwarming of about 0.4�C during 1958–2000 (Bijoy Thompson et al. 2008).8.3.2 Variation of All India RainfallFigure 8.2 depicts the mean monsoon rainfall anomaly over the period 1950–2006and is showing the decreasing trend. The red colour (Strong El Niño in the dark red20151050–5–10–15–20Anomaly1948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005YearFig. 8.1 Mean variation of latent heat flux anomaly over the Arabian Sea during Southwestmonsoon season (JJAS)6040200–20–40–60Rainfall Anomaly 195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006YearFig. 8.2 Variation of all India Rainfall anomalies during Southwest Monsoon season8 Prediction of Monsoon Variability and Subsequent Agricultural Production 127colour and weak El Niño in the light red colour) depicts the El Niño episode andlight colour depicts the weak El Niño. The blue colour (strong La Niña in dark blueand weak La Niña in light blue colour) depicts the La Niña episodes. The maximumnegative anomaly is observed on 1972, which is a strong El Niño episode, similarlyin the moderate El Niño episode in the year 2002 showing the negative anomaly,during these episodes the rainfall is less than the normal monsoon episode. It isobserved positive anomaly is also observed during the El Niño episodes during themoderate El Niño episode during 1994. The maximum positive anomaly is found inthe weak La Niña episode during the year 1961. During the Strong La Niña episodesand moderate episodes showing only positive anomaly, where as moderate La Niñahaving negative anomaly also. Significant positive anomalies observed in the LaNiña episodes and negative anomalies during the El Niño episodes.8.3.3 Monthly and Season Wise Variation in Latent Heat FluxFigure 8.3 shows the variation of latent heat flux during the monsoon months andthe seasonal mean over Arabian Sea. LHF decreases from June to September. Thesame feature can be observed in the all the episodes except during strong El Niño,strong La Niña and weak La Niña episodes. During Strong El Niño episode the LHFis higher and continued in the July where as in the strong La Niña LHF is higher inJuly than June and in the weak La Niña June and July are having the same LHF.The departure of LHF in relation to normal monsoon over Arabian Sea is studiedin particular to the monsoon months (JJAS) during El Niño and La Niña episodesare given in the Fig. 8.3. During the Strong El Niño event the LHF is higherthan normal monsoon during all theJune, July, August and September months.170 JuneAugustSWMJulySeptember160150140130120110100NormalMonsoonModerateEl NinoModerateLa NinaWeakEl NinoWeakLa NinaStrongLa NinaStrongEl NinoLatent Heat Flux (W/m2 )Fig. 8.3 Variation of latent heat flux during the normal monsoon, El Niño and La Niña episodes inthe months of June, July, August, September and southwest monsoon (SWM) season over ArabianSea128 M.V. Subrahmanyam et al.In the moderate El Niño episodes it is observed that during June, and September arehigher, where as in the months July and August it is lower than the normalmonsoon. In the weak El Niño episodes June month shows the higher and the restof the months it is lower. It was observed that during the El Niño episodes the LHFis higher during the June. In the moderate El Niño episode LHF is higher thannormal during June and September, where as lower during July and August. Duringthe weak El Niño June month LHF is higher where as in other months LHF is lower.During strong and moderate La Niña episodes, LHF is lower in June andSeptember and higher in July and August. In the contrary during the weak LaNiña episodes the LHF is higher in all the months. From these observations, it canbe concluded that higher LHF during July is an important for the strong El Niñoepisodes. The weak La Niña episodes LHF is higher during all the months.8.3.4 Monthly and Season Wise Variation in all India RainfallFigure 8.4 depict the mean variation of variation of all India rainfall during normalmonsoon, El Niño and La Niña episodes. From the figure it is observed that duringthe El Niño episodes the rainfall is below normal monsoon, expect in the weak ElNiño episode. During La Niña episodes the all India rainfall is more than the normalmonsoon rainfall. The June month rainfall is necessary for the crops, during strongEl Niño episode the rainfall is less when compared with all other episodes. Duringthe strong El Niño episode the rain fall is far below the normal monsoon rainfall, iteffects the crop growth which leads to lower production of crop. In the month of Julythe variation of rainfall is increasing from the strong to weak episode of El Niño. Thesame feature can be seen in the month of September. In August in the moderate ElNiño episode the rainfall is higher than the strong and weak El Niño episodes. Therainfall is higher in the months of June, August and September during the strong LaNiña episode, whereas during the month of July the rainfall is higher in the moderate350300250200150100500Rainfall (cm)NormalMonsoonModerateEl NinoModerateLa NinaWeakEl NinoWeakLa NinaStrongLa NinaStrongEl NinoJuneAugustSWMJulySeptemberFig. 8.4 Mean variation of all India mean rainfall during El Niño, La Niña and normal monsoonepisodes in June, July, August, September and Southwest monsoon season8 Prediction of Monsoon Variability and Subsequent Agricultural Production 129La Niña episode. The clear picture of El Niño can be seen in the months of July andSeptember as the rainfall increasing from Strong to weak episodes and the La Niñafeature can be seen in the month of September decreasing rainfall from the strongto weak episodes. The mean picture of the monsoon month shows the clearly thatduring El Niño episodes the rainfall is decreasing from strong to weak episodes andincreasing during the La Niña episodes from strong to weak episodes.8.3.5 Variations in Rainfall Departures DuringEl Niño and La Niña PeriodFigure 8.5 depict the mean variation of departures fromEl Niño and La Niña episodeswith the normal monsoon. During the strong El Niño episode, the departures arenegative it shows during the strong El Niño episode the rainfall is less when comparedwith the mean normal monsoon period and is positive during the La Niña episodes,there may be some exceptions like during moderate El Niño episode during in themonth of August the rainfall is higher than the other El Niño episodes and higherrainfall can be seen in the month of July in the weak El Niño episode. During themoderate La Niña the rainfall is lower when compared with the other La Niñaepisodes.8.3.6 Variation of Rainfall at Subdivision Level8.3.6.1 Over Coastal Andhra PradeshMean variation of rainfall in the coastal Andhra Pradesh is given in the Fig. 8.6aFrom the figure shows a slight decreasing trend in the rainfall over coastal50.040.030.020.010.00.0–10.0–20.0–30.0–40.0–50.0Departure from normal monsoonModerateEl NinoModerateLa NinaWeakEl NinoWeakLa NinaStrongLa NinaStrongEl NinoJuneAugustSWMJulySeptemberFig. 8.5 Mean variations of departures from El Niño and La Niña episodes with the normalmonsoon130 M.V. Subrahmanyam et al.Andhra Pradesh. The El Niño, La Niña and normal monsoon episodes can be seenin the different colours. El Niño episodes in the red and La Niña episodes in bluecolour. The strong colour indicates the strong episode and light colour indicates theweak episodes for El Niño and La Niña. Significant higher positive anomalies canbe observed during La Niña episodes and normal monsoon and lower anomalies inthe El Niño episodes. Negative anomalies can be observed during the strong andmoderate La Niña episodes also. The higher positive rainfall anomaly of around 60observed during the normal monsoon and higher negative anomaly of 53 observedduring moderate El Niño episode.Figure 8.6b depicts departure of rainfall from the El Niño and La Niña episodeswith the normal monsoon. The departure is negative in the El Niño episodes. Thedeparture is higher in the months of September during moderate and weak El Niñoepisodes and reasonable negative departures in the month of July. Significantpositive departures can be seen in August in Strong La Niña, in September duringmoderate La Niña and in June and September months in weak La Niña episodes.Negative departures can be seen in the months of June and July during strong LaNiña episode. in June and August months during moderate La Niña and in July andSeptember months of weak La Niña episodes.Coastal Andhra Pradesh mean rainfall anomaly during SW Monsoon 80604020–200–40–60Rainfall anomaly (mm)19501949194819511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006Departure from Normal MonsoonStongEl NinoModerateEl NinoWeakEl NinoStrongLa NinaModerateLa NinaWeakLa Nina40200–20–40–60–80–100June JulySeptemberAugustSWMFig. 8.6 (a) Mean variation of rainfall in the coastal Andhra Pradesh region. (b) Departure ofcoastal Andhra Pradesh rainfall from the El Niño and La Niña episodes with the normal monsoon8 Prediction of Monsoon Variability and Subsequent Agricultural Production 1318.3.6.2 Variation of Rainfall Over TelanganaFigure 8.7a shows the mean variation of rainfall anomaly in the Telangana region.In the telangana region the mean seasonal rainfall shows decreasing trend. Thecolour codes are same as earlier. During 1988 significant higher positive anomalycan be observed. Significant higher positive rainfall can be observed during La Niñaepisodes and normal monsoon. During the strong La Niña episode the higherpositive anomaly was observed in the year 1988. And the higher negative anomalyobserved during strong El Niño episodes in the year 1972.Departure of Telangana rainfall from the normalmonsoon during El Niño andLa Niña episodes are given in the Fig. 8.7b. Negative departures can be seen duringthe El Niño episodes. Significant higher departure observed in the July monthduring weak El Niño episode. Positive departure can be seen in the La NiñaTelangana mean rainfall anomaly during SW MonsoonRainfall anomaly (mm)15019481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006100500–50–100100.0JuneDeparture from normal monsoonJulySeptemberAugustSWM80.060.040.020.00.0–20.0–40.0–60.0–80.0–100.0StrongEl NinoStrongLa NinaModernEl NinoModernLa NinaWeakEl NinoWeakLa NinaFig. 8.7 (a) Mean rainfall anomaly in the Telangana region. (b) Departure of Telangana regionrainfall during El Niño and La Niña episodes with normal monsoon132 M.V. Subrahmanyam et al.episodes. Significant negative departure can be seen in the June and August monthsof moderate La Niña episode and in the September month of weak La Niña episode.8.3.6.3 Variation of Rainfall Over RayalseemaMean southwest monsoon rainfall variation in the Rayalseema region is shown inthe Fig. 8.8a. The seasonal rainfall is near to normal rainfall. In all El Niño episodesthe rainfall is lesser than the normal rainfall, there are some exceptions also. In theLa Niña episodes the rainfall is higher than the normal. The higher positive rain fallanomaly of 85 in the year 1996 during normal monsoon and higher negativeanomaly of 45 also observed during normal monsoon in the year 1952.Departures of rainfall during El Niño and La Niña episodes with normal mon-soon are given in the Fig. 8.8b. During the strong El Niño episode positive departureRayalseema mean rainfall anomaly during SW Monsoon100.0Rainfall anomaly (mm)80.060.040.020.0 0.0–20.0–40.0–60.019481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006806040200–20–40–60–80StrongEl NinoStrongLa NinaModerateEl NinoModerateLa NinaWeakEl NinoWeakLa NinaJune JulySeptemberAugustSWMDeparture from normal monsoonFig. 8.8 (a) Mean variation of rainfall in the Rayalseema region in the southwest monsoonseason. (b) Departures of rainfall during El Niño and La Niña episodes with normal monsoon8 Prediction of Monsoon Variability and Subsequent Agricultural Production 133can be seen in the month of June and during the strong La Niña episode negativedeparture can be seen. The maximum negative departure can be seen in the weakEl Niño episode in the month of September. And maximum positive departure canbe seen in the month of August during strong La Niña episode. During the moderateLa Niña the maximum positive departure can be seen in the month July.8.3.7 Spatial Distribution Ocean Latent Heat FluxesComparison During El Niño and La Niña PeriodsLarge seasonal and inter annual variability in SST exists in the western andsoutheast tropical Indian Ocean. The processes, viz; latent heat flux, radiativeflux, advection and entrainment play major roles in SST. Inter annual variabilityof heat flux is found to have greater influence on SST in most parts of the IndianOcean (Behra et al. 2000). They have simulated the SST anomalies over differentregions in Indian Ocean and found that latent heat dominates in all the regions.Advection and net heat flux are found to be dominant in the equatorial IndianOcean. Warming observed in 1982–1983 (El-Niño) was due to the high latent heatflux. Interannual variability of heat fluxes could not be compared as there were noearlier studies for the same years. However, latent heat flux which is found to playa dominant role in SST variability, the distribution of latent heat flux under summerand winter monsoon seasons (Rao and Sriram 2005). The role of the latent heat fluxplays vital role to explain the advent and performance of monsoon systemsas revealed by some radiation and heat flux parameters (Subramaniyam 2006).Figure 8.9a, b shows the latent heat flux during El Niño and La Niña periods. Itshows wide contrast between El Niño and La Niña. The figure shown to the left isLHF during El Niño. It is evident that LHF is intense and covering large area. Eventhough the areal extent is same, the intensity is less in La Niña. These fluxes areestimated in the month of June. It is observed that the contrast is more in June. Wecan use it is an indicator for performance of the monsoon rainfall.8.3.8 Agriculture Production and Its Relationship with RainfallThe increasing trend of agriculture production can be interpreted as the proper agrotechnical measures and use of complex fertilizers (intensive cultivation) besidesincrease in the land under cultivation bringing waste lands into irrigated fields. Thedetrended yields find removing the increased trend pattern in agriculture productioncalculated deviation in all India annual rainfall. This is graph shows detrended yieldwhich can be related to meteorological parameters like rainfall to find the relation-ship. Figure 8.10 depicts the it is showing that yield reduction coincide with rainfall134 M.V. Subrahmanyam et al.negative deviations which is coincide the El Niño periods .High yield amountcoincide with positive deviations which is coincide with La Niña periods. Someexemption also find during El Niño and La Niña periods.45 50 55 60 65 70 75 80-505101520Latitude (° N)a45 50 55 60 65 70 75 80Longitude (° E)Longitude (° E)-505101520Latitude (° N)01020304050607080901001101201301401501601701801902002102202302402502602702802903003103200102030405060708090100110120130140150160170180190200210220230240250260270280290300310320bFig. 8.9 Mean special distribution of latent heat flux during (a) El Niño and (b) La Niña episodesin the month of June8 Prediction of Monsoon Variability and Subsequent Agricultural Production 1358.4 ConclusionsThe teleconnections for Indian summer monsoon are impartment for prediction andplanning. El Niño and La Niña in December gives sufficient time (6 months) forthe prediction of Asian Summer monsoon. The intensity of El Niño and La Niña isvery much related to the Oceanic fluxes in the Indian Ocean. The Ocean Fluxesespecially Latent heat flux over Arabian Sea shows an Increasing trend. The meanmonsoon LHF is showing the positive trend and all India rainfall showing thenegative trend. The LHF and all India rainfall are showing the negative correlation .The important conclusion of this study is Latent Heat Flux differ very much duringEl Niño and La Niña episodes in June, which gives in an important clue toagricultural and ecological planning.. During the strong El Niño (La Niña) theLHF is higher (lower) and the rainfall is lower (higher). The departure of LHF andall India rainfall is also showing the same feature. Departure from the normalmonsoon can reveal how the rainfall varying during the El Niño and La Niñaepisodes. The food production increased because of proper agro technical measureslike better seed varieties, improvedcomplex fertilizers and well tested insecticidesand fungicides, but yet there is a decreasing trend during El Niño periods thesustainable development should be aimed at taking consideration of low rainfallin June and July months during El Niño. Alternate crops which require less waterare to be planed, which will ensure sustainable development.0.25Y axis Rainfall deviationsX axis - Detrended y = 0.598xR2 = 0.3840.20.150.10.050.05–0.1–0.1 0.1 0.2 0.3 0.4–0.2–0.3–0.15–0.2–0.2500Fig. 8.10 The graph between detrended yield and rainfall deviations136 M.V. Subrahmanyam et al.ReferencesBehra SK, Salvekar PS, Yamagata T (2000) Simulation of interannual SST variability in thetropical Indian ocean. J Climate 13:3487–3499Cadet DL, Diehl BC (1984) Interannual variability of surface fields over the Indian Ocean duringrecent decades. Mon Wea Rev 112:1921–1935Cadet D, Reverdin G (1981) Water vapour transport over the Indian ocean during summer 1975.Tellus 33:476–487Cadet D, Reverdin G (1981b) Water Vapour Transport over the Indian Ocean during Summer1975. Tellus 33:476–487Clark CK, Cole JE, Webster PJ (2000) Indian ocean SST and Indian summer rainfall: predictiverelationships and their decadal variability, J Climate 13:2503–2519Gangadhava Rao LV, Shree Ram P (2005) Upper ocean physical processes in the tropical Indianocean, A monograph prepared under CSIR Emeritus Scientist Scheme. National Institute ofOceanography, GoaGhosh SK, Pant MC, Dewan BN (1978) Influence of the Arabian Sea on the Indian summermonsoon. Tellus 30:117–125Howland MR, Sikdar DN (1983) The moisture budget over the N.E. Arabian Sea during pre-monsoon and monsoon onset. Mon Wea Rev 111:2255–2268Joseph PV, Pillai PV (1984) Air-sea interaction on a seasonal scale over north Indian Ocean. Part I:Interannual variations of sea surface temperature and Indian summer monsoon rainfall.Mausam 35:323–330Mohanty UC, Ramesh KJ (1993) Characteristics of certain surface meteorological parametersin relation to the interannual variability of Indian summer monsoon. Proc Indian Acad Sci(Earth Planet Sci) 102(1):73–87Murakami T, Nakazawa T et al (1984) On the 40–50 day oscillation during the 1979 northernhemisphere summer. Part. II, Heat and Moisture budget. J Meteor Soc Jpn 62:469–484Pisharoty PR (1965) Evaporation from the Arabian Sea and the Indian southwest monsoon. In:Proceedings of the symposium on meteorological results of the IIOE. India MeteorologicalDepartment, Bombay, pp 43–54Ramamurthy K, Jambhunathan R et al (1976) Moisture distribution water vapour flux over theArabian Sea during active and weak spells of southwest monsoon. Ind J Met Hydrol Geophys27(2):127–140Ramesh Kumar MR, Sadhuram Y (1989) Evaporation over the Arabian Sea during two contrastingmonsoons. Meteorol Atmos Phys 41:87–97Ramesh Kumar MR, Sathyendranath S et al (1986) Sea surface temperature variability over NorthIndian Ocean – a study of two contrasting monsoon seasons. Proc Indian Acad Sci Earth PlanetSci 95(3):435–446Ramesh Kumar MR, Shenoi SSC et al (1999) On the role of the cross equatorial flow on summermonsoon rainfall over India using NCEP/NCAR reanalysis data. Meteorol Atmos Phys70:201–213Rao KG, Goswami BN (1988) Interannual variations of sea surface temperature over the ArabianSea and the role of low level flow on the summer monsoon rainfall over the Indian subcontinentduring two contrasting monsoon years in an Monsoon: a new perspective. Mon Wea Rev116:558–568Sadhuram Y (1997) Predicting monsoon rainfall and pressure indices from sea surface tempera-ture. Curr Sci 72(3):166–167Sadhuram Y, Ramesh Kumar MR (1988) Does evaporation over the Arabian Sea play a crucialrole in moisture transport across the west coast of India during an active monsoon period? MonWea Rev 116:307–312Saha KR, Bavadekar SN (1973) Water vapour budget and precipitation over the Arabian Seaduring the northern summer. Q J R Meteor Soc 99:273–2788 Prediction of Monsoon Variability and Subsequent Agricultural Production 137Shukla J (1987) Monsoons. In: Fein JS, Stephens PI (eds) Interannual variability of Monsoons.Wiley, New York, pp 399–464Shukla J, Misra BN (1977) Relationships between sea surface temperature and wind speed over theCentral Arabia Sea and monsoon rainfall over India. Mon Wea Rev 105:998–1002Subramaniyam (2006) Role of North Indian Ocean in the advent and performance of monsoonsystems as revealed by some radiation and heat flux parameters. PhD, Andhra University, IndiaSwadhin KB et al (2007) What causes the Indian Ocean warming? Geophys Res Abstr 9:10950Thompson B, Gnanaseelan C et al (2008) North Indian Ocean warming and sea level rise in anOGCM J. Earth Syst Sci 117:169–178Troup AJ (1965) The southern oscillation. Quart J Roy Meteor Soc 91:490–506Weare BC (1979) A statistical study of the relationship between ocean surface temperature and theIndian monsoon. J Atmos Sci 26:2279–2291Webster PJ (1987) Monsoons. In: Fein JS, Stephens PI (eds) The elementary monsoon. Wiley,New York, pp 399–464Webster PJ, Yang S (1992) Monsoon and ENSO: selectively interactive systems. Q J RMeteor Soc118:877–926138 M.V. Subrahmanyam et al.Chapter 9Improved Seasonal Predictability Skillof the DEMETER Models for CentralIndian Summer Monsoon RainfallRavi P. Shukla, K.C. Tripathi, Sandipan Mukherjee, Avinash C. Pandey,and I. M. L. DasAbstract A method to improve the predictability of the seven models in theDEMETER project is proposed. This technique has been applied to investigatethe effect on the predictability of seasonal precipitation (June–July–August) in thecentral Indian region. Three best models of DEMETER have been selected basedon error and correlation analysis and standard deviation of the observed data. It isfound that the ensemble mean prediction, when using these three models is betterthan the ensemble mean prediction when using all the seven models. It is alsoobserved that the root mean square error is reduced when the ensemble mean of themodels is taken. At the same time it is also observed that the ensemble mean doesnot, in every case, give better forecast skill scores in case of dichotomous forecastswhich tell the skill of the models in forecasting rare events.9.1 IntroductionSeasonal timescale climate predictions are important for the society for a variety ofreasons (Thomson et al. 2000; Hartmann et al. 2002). Predictability is affected byperturbations in the initial conditions as well as the uncertainty in the representationof partial differential equations as a finite-dimensional set of ordinary differentialR.P. Shukla (*) • K. Tripathi • S. MukherjeeK. Banerjee Center of Atmospheric & Ocean Studies, University of Allahabad, Allahabad 211002,Indiae-mail: ravishukla72@gmail.com; kctripathi@gmail.com; mukherjee.sandipan@rediffmail.comA.C. PandeyM. N. Saha Centre of Space Studies, University of Allahabad, Allahabad 211002, Indiae-mail: avinashcpandey@rediffmail.comI.M.L. DasDepartment of Physics, University of Allahabad, Allahabad 211002, Indiae-mail: profimldas@yahoo.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_9, # Springer-Verlag Berlin Heidelberg 2011139equations in a digital computer. At present it is not possible to theoretically esti-mate a probability distribution of such model uncertainty (Palmer 2001). Anapproach to overcome uncertainties arising out of such model uncertainty is totake an ensemble of semi-independent global climate models, called the Multi-Model Ensemble (MME) (Palmer et al. 2004). It has been established that multi-model ensemble makes more reliable forecasts than individual members, or anensemble of many members of a single model, under the funded research projectssuch as Prediction Of climate Variations On Seasonal to Interannual Time-scales(PROVOST) (Palmer and Shukla 2000), Dynamical Seasonal Prediction (DSP)(Shukla et al. 2000; Palmer and Shukla 2000) and Development of EuropeanMulti-model Ensemble system for seasonal to inTER-Annual Prediction (DEME-TER) (Palmer et al. 2004) established by the European Center for Medium-RangeWeather Forecast (ECMWF). In such ensemble forecasts, the uncertainties in theinitial state are addressed through an ensemble of different ocean initial conditionsfor each model.The DEMETER project consists of seven global coupled ocean-atmospheremodels viz. CERF (European Center for Research and Advanced Training inScientific Computation, France), ECMWF, INGV (Instituto Nazionale de Geofisicae Vulcanologia, Italy), LODYC (Laboratoire d’Oceanographie Dynamique et deClimatologie, France), METF (Centre National de Recherches Meteorologiques,Meteo-France, France), UKMO (The Met Office, UK) and MAXP (Max PlanckInstitut fur Meteorologie, Germany). Each model consists of nine individualmembers.The Multi-model Ensemble technique has also been useful whenpredicting the winds at 850 hPa between the Madagaskar and Western Australia(Rai et al. 2008).Most of the Indian subcontinent receives 70–90% of its annual rain fall duringthe summer monsoon. Prediction of All India seasonal mean Rainfall (ISMR) isuseful for the country’s policy makers as it is of great importance for agriculturalplanning as well as management of droughts and floods which directly affectcountry’s agricultural production, economy and human loss (Gadgil and Rao2000). Therefore, understanding and predictability of ISMR has been a subjectof several scientific investigations since long (Blanford 1884, 1886). A recentadvancement in the prediction of ISMR is that of real time forecasting technique(Xavier and Goswami 2007). Seasonal forecasting of the Indian summer monsoonrainfall has been widely studied (Kumar et al. 1995; Krishnamurthy and Shukla2000). Predictability of the seasonal mean rainfall by the AGCM using differentinitial conditions, boundary condition and models has been recently studied (Kangand Shukla 2006; Kang et al. 2004; Shukla and Fennessy 1994).Owing to the importance of the ISMR for the Indian subcontinent, a variant ofmulti-model ensembling technique is applied for the seasonal prediction of precip-itation in the central Indian region (73–83�E, 18–27�N) for the period 1980–2001.The technique used in the present study consists of applying Multi-modelensembling to three better models of DEMETER selected on the basis of RMSerror and correlation coefficient with the observed precipitation. This model isreferred to as the Selected Multi-Model Ensembling (SMME). The predictive skills140 R.P. Shukla et al.of the DEMETER model, MME and SMME have been compared. The study maybe further undertaken by comparing the variants proposed herein with the super-ensemble method of multi-model ensembling (Krishnamurthy et al. 1999).This paper is divided into followings sections. Section 9.2 gives a brief descrip-tion of the model and data. Section 9.3 describes the forecast skill score measure.Results of mean prediction predictive skill and categorical forecast measures withrespect to observations have been discussed in Sect. 9.4 and concluding remarks arepresented in Sect. 9.5.9.2 Model and DataThe data used in the present study is 22 year hind cast of model generated data ofthe DEMETER project models. The data has been obtained from the InternationalCenter for theoretical Physics (ICTP), Trieste, Italy under joint collaborationbetween ICTP and Centre of Ocean–Land–Atmospheric (COLA) Studies, USA,through Targeted Training Activity (TTA) program. The DEMETER hindcastswere started from 1 February, 1 May, 1 August and 1 November to assess theseasonal dependence on prediction skill. Each hindcast comprises an ensemble ofnine members.The observed rainfall data used is the gridded rainfall data (1� � 1� resolution)from India Meteorological Department (IMD) for the period 1980–2001 based on1,803 station (Rajeevan et al. 2006). The time series corresponding to 1980–2001 isused for our analysis.9.3 MethodologyAn ensemble of nine ocean initial conditions is taken to address the uncertainties inthe initial state for each model. MME is done by taking the average of suchensemble forecasts. The outputs of the DEMETER models and the observedprecipitation are averaged over the central Indian region and the mean is subtracted(anomaly retained) before doing the analysis.9.3.1 Categorical Forecast Skill for Dichotomous ForecastsCategorical (dichotomous) forecast skills refer to the ability of the models to predictthe rare events (Wilks 1995). For the dichotomous forecasts, the data is divided intotwo classes: rare events and normal events. Rare event is defined as one in which thedata is outside the range of standard deviation. A normal event is one when theobserved data is under the standard deviation of the observed data.9 Improved Seasonal Predictability Skill of the DEMETER Models for Central Indian 141For evaluating the categorical forecast skills in the present study, the data isnormalized by dividing with the respective standard deviations. Since the standarddeviation of the normalized data is 1, a rare event happens when the modulus ofthe normalized observed precipitation exceeds 1. A normal event is one when themodulus of the normalized observed precipitation remains less than 1. A rare eventis called “event” in this context.The contingency table is a useful way to see what types of errors are being made.It shows the frequency of “yes” and “no” forecasts and occurrences. A “yes” in thecontingency table denotes the occurrence of the “event”. Following quantities aredefined:Hit (H) – event forecast to occur, and did occurMiss (M) – event forecast not to occur, but did occurFalse alarm (F) – event forecast to occur, but did not occurCorrect negative (N) – event forecast not to occur, and did not occurFollowing skill measures are defined for the individual models, MME andSMME:Accuracy ¼ HþNTotal Total ¼ HþMþ Nþ FBias ¼ H þ FH þMFalse alarm ratio Farð Þ ¼ FHþFProbability of false detection Pofdð Þ ¼ FNþFProbability of detection Podð Þ ¼ HHþMð ÞThreat score Tsð Þ ¼ HHþMþFð ÞHeidke skill score Hssð Þ ¼ HþNð Þ� correctrandomð Þ½ �Total� correctrandomð Þ½ �correctrandomð Þ ¼ H þMð Þ H þ Fð Þ þ N þMð Þ N þ Fð Þ½ �TotalEquitable threat score Etsð Þ ¼ H�Hrandomð ÞHþMþF�Hrandomð ÞHrandom ¼ H þMð Þ H þ Fð Þ½ �Total9.4 Results and DiscussionThe mean prediction of precipitation over the region (73–83�E, 18–27�N) has beenstudied for the period 1980–2001. In Table 9.1 RMS error and correlation coeffi-cient of DEMETER model outputs with the observed precipitation have beenshown. Standard deviation of the observed precipitation is 1.12.142 R.P. Shukla et al.Tables 9.1 and 9.2 show the RMS errors and correlation coefficients of MME,SMME and DEMETER models with the observed precipitation. It can be seen fromthese tables that the predictability skills of MAXP, METF and UKMO models arecomparable to that of MME. However, none of the models, including the MME, hasbetter predictability skills than the SMME.The issue of the possibility of a subset of the seven models being used for MMEprediction has also been dealt with. Three best models, namely MAXP, METF andUKMO, have been selected on the basis of error analysis and correlation coefficient(Table 9.1) for the SWMME models. It can be seen that the SMME gives betterresults than the MME (Table 9.2).The deviation (anomaly component) for the spatial ensemble mean of DEME-TER model outputs and observed precipitation of the central Indian region forthe 22 year period (1980–2001) for June-August (JJA) is shown in Fig. 9.1. Thedeviation (anomaly component) for the spatial average of MME and SMMEmodelsKishore Centre for Atmospheric Sciences, Indian Institute of Tech-nology, Delhi, Hauz Khas, New Delhi 110016, India, osurikishore@gmail.comS. Raj Kumar Natural Plant Products & Biodiversity Divisions, Institute ofHimalayan Bioresource Technology (CSIR), Palampur, HP 176061, India,ramechek@yahoo.co.inContributors xixDr. H.S. Kushwaha Department of Soil Science, College of Agriculture, G. B.Pant University of Agriculture & Technology, Pantnagar, Uttarakhand 263145,India, kushwahahs@yahoo.co.inA.J. Litta Centre for Atmospheric Sciences, Indian Institute of Technology, Delhi,Hauz Khas, New Delhi 110016, India, ajlitta@gmail.comR.K. Mall India Meteorological Department, National Institute of DisasterManagement, New Delhi, India, mall_raj@rediffmail.comAshu Mamgain Center for Atmospheric Sciences, IIT Delhi, Hauz Khas, NewDelhi 110016, India, ashumam@gmail.comN. Manikandan Central Research Institute for Dryland Agriculture, Hyderabad,AP 500 059, IndiaE.A.R. Mellaart EcoLink, 83, Karino 1204, South Africa, mellaart.e@soft.co.zaA.K. Mishra Department of Agrometeorology, College of Agriculture, G.B. PantUniversity of Agriculture and Technology, U.S. Nagar, Pantnagar, Uttarakhand263145, India, ashueinstein@gmail.comM.E. Moeletsi ARC-Institute for Soil, Climate and Water, Private Bag X79,Pretoria 0001, South Africa, moeletsie@arc.agric.zaU.C. Mohanty Centre for Atmospheric Sciences, Indian Institute of Technology,Delhi, Hauz Khas, New Delhi 110016, India, mohanty@cas.iitd.ernet.inM.Mohapatra India Meteorological Department, Lodi Road, New Delhi 110 003,India, mohapatra_imd@yahoo.comRaymond P. Motha U.S. Department of Agriculture, Office of the Chief Econo-mist, World Agricultural Outlook Board, 1400 Independence Avenue, Room 4441South Building, Washington, DC 20250-3812, USA, rmotha@oce.usda.govN.S. Mpandeli ARC-Institute for Soil, Climate and Water, Private Bag X79,Pretoria 0001, South Africa, SMpandeli@deat.gov.zaSandipan Mukherjee K Banerjee Center of Atmospheric & Ocean Studies,University of Allahabad, Allahabad 211002, India, mukherjee.sandipan@rediff-mail.comK.P.R. Vittal Murthy Department of Meteorology and Oceanography, AndhraUniversity, Visakhapatnam, AP, India, kprvm@yahoo.comxx ContributorsA.S. Nain Department of Agrometeorology, Indira Gandhi Krishi Vishwavidyalya,Raipur 492006 (C.G.), IndiaG.O. Odhiambo Department of Geography and Urban Planning, College ofHumanities and Social Sciences, United Arab Emirates University, 17771,Al Ain, United Arab Emirates, godhiambo@uaeu.ac.aeKavita Pabreja Research Scholar-BITS, Pilani, C-3A / 39C, DDA Flats, JanakPuri, New Delhi 110058, India, kavita_pabreja@rediffmail.comS.P. Pachauri Department of Agrometeorology, College of Agriculture, G.B. PantUniversity of Agriculture and Technology, U.S. Nagar, Pantnagar, Uttarakhand263145, IndiaS.K. Panda Centre for Atmospheric Sciences, Indian Institute of TechnologyDelhi, Hauz Khas, New Delhi 110 016, India, sampadpanda@gmail.comAvinash C. Pandey Department of Physics, University of Allahabad, Allahabad211 002, India; M. N. Saha Center of Space Studies, IIDS, University of Allahabad,Allahabad 211002, India, avinashcpandey@rediffmail.comVyas Pandey Department of Agricultural Meteorology, Anand AgriculturalUniversity, Anand, Gujarat 388110, India, pandey04@yahoo.comH.R. Patel Department ofAgriculturalMeteorology,AnandAgriculturalUniversity,Anand, Gujarat 388 110, India, hrpatel410@yahoo.comS.R. Patel Department of Agrometeorology, Indira Gandhi Krishi Vishwavidyalya,Raipur 492006 (C.G.), India, srpatelsr@yahoo.comD.R. Pattanaik India Meteorological Department, New Delhi, India, drpattanaik@gmail.comS. Pattanayak Centre for Atmospheric Sciences, Indian Institute of Technology,Delhi, Hauz Khas, New Delhi 110016, India, sujata05@gmail.comAnand Patwardhan Shailesh J. Mehta School of Management, Indian Institute ofTechnology Bombay, Powai, Mumbai, MH 400076, India, anand@iitb.ac.inVinayak S. Phadke 2/17 Dnyanayog Society, Vazira Naka, Lokmanya TilakRoad, Borivli (W), Mumbai, MH 400091, India, vinayakphadke@hotmail.comSavita Rai Centre for Atmospheric Sciences, Indian Institute of Technology Delhi,Hauz Khas, New Delhi 110 016, India, savita1559@gmail.comContributors xxiSethu Raman State Climate Office of North Carolina, NC State University,Box 7236, Raleigh, NC 27695-7236, USA, sethu_raman@ncsu.eduK. Ramesh Natural Plant Products & Biodiversity Divisions, Institute of Himala-yan Bioresource Technology (CSIR), Palampur, HP 176061, India, kramesh@iiss.ernet.inA.V.M.S. Rao Central Research Institute for Dryland Agriculture, Hyderabad, AP500 059, India, avmsrao@crida.ernet.inG.G.S.N. Rao Central Research Institute for Dryland Agriculture, Hyderabad, AP500 059, India, ggsnrao@crida.ernet.inV.U.M. Rao Central Research Institute for Dryland Agriculture, Hyderabad, AP500 059, India, vumrao54@yahoo.comL.S. Rathore India Meteorological Department, New Delhi 110 003, India,lsrathore@ncmrwf.gov.inRitu Department of Agricultural Meteorology, Punjab Agricultural University,Ludhiana, India, waliadimpy@rediffmail.comFederica Rossi Consiglio Nazionale delle Ricerche, Institute of Biometeorology,Via P. Gobetti 101, Bologna 40129, Italy, f.rossi@ibimet.cnr.itA. Routray Centre for Atmospheric Sciences, Indian Institute of Technology,Delhi, Hauz Khas, New Delhi 110016, India, ashishroutray.iitd@gmail.comDipak K. Sahu Centre for Atmospheric Sciences, Indian Institute of TechnologyDelhi, Hauz Khas, New Delhi 110 016, India, dipakmath@gmail.comR.P. Samui Agricultural Meteorology Division, India Meteorological Department,Pune, India, rsamui@yahoo.comP. Parth Sarthi Centre for Global Environmental Research, TERI, DarbariSeth Block, India Habitat Centre, Lodhi Road, New Delhi 110 003, India,drpps@hotmail.com, ppsarthi@teri.res.inA.S.R.A.S. Sastri Department of Agrometeorology, Indira Gandhi KrishiVishwavidyalya, Raipur 492006 (C.G.), India, asastri@yahoo.comG. U. Satpute SWCE, Department of Soil and Water Conservation Engineering,Dr. Panjabrao Deshmukh Krishi Vidyapeeth, PO Krishi Nagar, Akola (MS) 444104, India, gusatpute@rediffmail.comxxii ContributorsT. Satyanarayana South China Sea Institute of Oceanology, Chinese Academyof Science, Beijing, China; Central Research Institute for Dryland Agriculture,Hyderabad 500 059, India, satya_1006@yahoo.co.inM.J. Savage SPAC Research Unit, Agrometeorology discipline, School of Envi-ronmental Sciences, University of KwaZulu-Natal, P/Bag X01, Scottsville, Pieter-maritzburg 3209, Republic of South AfricaP.K. Sharma Division of Agricultural Physics, Indian Agricultural ResearchInstitute, New Delhi 110012, IndiaRavi P. Shukla Department of Physics, University of Allahabad, Allahabad211002, India, ravishukla72@gmail.comB. Simon Space Applications Centre (ISRO), Ahmedabad 380 015, India,babysimon@gmail.comAaron P. Sims State Climate Office of North Carolina, NC State University,Box 7236, Raleigh, NC 27695-7236, USA, aaron_sims@ncsu.eduAjay Singh Shailesh J. Mehta School of Management, Indian Institute of Tech-nology Bombay, Powai, Mumbai, MH 400076, India, ajayvisen@yahoo.co.inGuriqbal Singh Department of Plant Breeding and Genetics, Punjab AgriculturalUniversity, Ludhiana, India, singhguriqbal@rediffmail.comJagadish Singh Mausam Apartments, West Enclave, Pitampura, Delhi 110 034,IndiaK.K. Singh India Meteorological Department, New Delhi 110 003, India,kksingh2022@gmail.comR. Singh Department of Agrometeorology, Indira Gandhi Krishi Vishwavidyalya,Raipur 492006 (C.G.), IndiaS.K. Singh Department of Agrometeorology, College of Agriculture, G.B. PantUniversity of Agriculture and Technology, U.S. Nagar, Pantnagar, Uttarakhand263145, India, sskumar_1983@rediffmail.comP.K. Singh India Meteorological Department, Agromet Service Cell, MausamBhavan, New Delhi,and the observed precipitation is shown in Fig. 9.2.Categorical forecast verification has been performed for rainfall for the individ-ual DEMETER models, MME and SMME. Contingency table, based on thedefinition of ‘Yes’ and ‘No’ forecasts, is shown in Table 9.3. The categoricalforecast verification has been quantified by calculating dichotomous forecast skillmeasures and is given in Table 9.4.Accuracy is greater than 0.50 for the individual models and MME. The accuracyis 0.727 and 0.68 for MAXP and SMME respectively. Since the bias is greater than1, it can be argued that all the models are over predicting. Probability of detection(POD) for MAXP and UKMO is 0.50. Probability of detection (POD) for SMME is0.25. False alarm ratio (FAR) is minimum for MAXP. FAR and POD are 1 and0 respectively, both for METF and LODY, showing that these models fail to predictrare events. Probability of false detection (POFD) is small (0.22) for SMME, CERFand MAXP which indicates that the fraction of “events” which were “wrongly”forecast is small. Threat score (TS) measures accuracy when the correct negativehas been removed from the forecast and it is found 0.25 and 0.20 for MAXP andUKMO respectively. Equitable Threat Score (ETS) and Hedkie Skill Score (HSS)are also larger for the MAXP, UKMO, ECMW, CERF and SMME.Table 9.1 Root Mean Square (RMS) error and Correlation Coefficient (r) computed for totalrainfall obtained from DEMETER seasonal forecast system and observations. (Standard deviationof observed data ¼ 1.12)Model UKMO ECMW INGV LODY MAXP METF CERFRMS error 1.13 1.23 1.35 1.41 1.05 1.09 1.15r 0.32 �0.07 0.00 �0.04 0.29 0.07 0.07Table 9.2 Root Mean Square (RMS) error and Correlation Coefficient (r) computed multi-modelensemble (MME) and SMMEModel MME SMMERMS error 1.121 1.029r 0.097 0.349 Improved Seasonal Predictability Skill of the DEMETER Models for Central Indian 14302 4 6 8Time (year)10 12 14 16 18 20 2212345ObservationCERFECMWINGVLODYMETFMAXPUKMO6–1–2–3Seasonal rainfall anomaly (mm\day)Fig. 9.1 Time series of observation and individual model outputs of the DEMETER project forseasonal (JJA) precipitation anomaly averaged over central India for the period 1980–2001ObservationSeasonal rainfall anomaly (mm/day) SMMEMME-31981 1983 1985 1987 1989 1991Time (Year)1993 1995 1997 1999 2001-2-10123456Fig. 9.2 Time series of observation, multi-model ensemble (MME) and selected multi-modelensemble (SMME) of the individual model outputs of the DEMETER project for seasonal (JJA)precipitation anomaly averaged over central India for the period 1980–2001144 R.P. Shukla et al.It has been pointed out that the SMME model has better predictability skills thanall the other models including the MME when RMS error and correlationcoefficients were compared. On the basis of categorical forecast skills, however,it can be seen that individual models (Table 9.4) may outperform MME and SMMEboth. Evaluation of categorical forecast skills is important for the actual comparisonof models as it informs us of the model’s ability to detect rare events. The purposeof calculating categorical forecast skill scores is, in the present case, to strengthenthe argument that the ensemble mean forecast is not always better than the bestindividual model of DEMETER project.It is also interesting to mention that CERF and ECMWmodels are showing samedichotomous skill scores (Table 9.4). However, from Table 9.1 it is clear that RMSerror and correlation coefficients are different for the both models. Since theTable 9.3 Contingency tablefor the individual model,MME and SMME ModelObservedYes NoUKMO ForecastYes 2 6No 2 12ECMW ForecastYes 1 4No 3 14INGV ForecastYes 1 8No 3 10LODY ForecastYes 0 6No 4 12METF ForecastYes 0 5No 4 13MAXP ForecastYes 2 4No 2 14CERF ForecastYes 1 4No 3 14MME ForecastYes 1 6No 3 12SMME ForecastYes 1 4No 3 14Table 9.4 Categorical skill scores for individual models, MME and SMMEModel Accuracy Bias POD FAR POFD TS ETS HSSUKMO 0.636 2.0 0.5 0.75 0.33 0.2 0.06 0.12ECMW 0.68 1.25 0.25 0.8 0.22 0.125 0.012 0.025INGV 0.5 2.25 0.25 0.88 0.44 0.083 �0.061 �0.13LODY 0.54 1.5 0 1 0.33 0 �0.12 �0.28METF 0.59 1.25 0 1 0.277 0 �0.11 �0.25MAXP 0.727 1.5 0.5 0.66 0.22 0.25 0.13 0.22CERF 0.68 1.250 0.250 0.800 0.222 0.125 0.013 0.025MME 0.59 1.75 0.25 0.86 0.33 0.10 �0.03 �0.06SMME 0.682 1.25 0.25 0.80 0.22 0.125 0.013 0.0259 Improved Seasonal Predictability Skill of the DEMETER Models for Central Indian 145dichotomous forecast skills refer to the capability of predicting the rare events only,we conclude that both are equivalent when predicting the rare events.9.5 ConclusionThe models outputs of the DEMETER project have been analyzed for the predictionof seasonal (JJA) anomalies of the precipitation in the central Indian region for theperiod 1980–2001. Alternative technique for MME is proposed and its performanceis evaluated. The technique, selected multi-model ensembling (SMME), consists ofselecting three better individual models on the basis of error and correlationanalysis and averaging the results over the models as is done in MME. It hasbeen observed that the SMME model has better prediction skills than all theindividual models and MME. The MME does not, in every case, has betterprediction skills than the individual models when RMS error and correlationcoefficients are considered. However, the SMME has shown better skills than allthe individual models. Further, on the basis of categorical forecasts, it is seen thatMME and SMME do not always give better prediction than individual models.ReferencesBlanford HF (1884) On the connection of the Himalaya snowfall with dry winds and seasons ofdrought in India. Proc R Soc Lond 37:3–22Blanford HF (1886) Rainfall of India. Mem India Meteorol Dept 2:217–448Gadgil S, Rao PRS (2000) Famine strategies for variable climate – a challenge. Curr Sci78:1203–1215Hartmann HC, Pagano TC, Sorooshian S, Bales R (2002) Confidence builders: evaluating seasonalclimate forecast for user prospective. Bull Am Meteor Soc 83:683–698Kang IS, Shukla J (2006) Dynamical seasonal prediction and predictability. In: Wang B (ed) TheAsian Monsoon. Springer, Heidelberg, pp 585–612Kang IS, Lee JY, Park CK (2004) Potential predictability of summer mean precipitation in adynamical seasonal prediction system with systematic error correction. J Clim 17:834–844Krishnamurthy V, Shukla J (2000) Intraseasonal and interannual variability of rainfall over India.J Clim 13:4366–4377Krishnamurthy TN, Kishtawal CM, LaRow TE, Bachiochi DR, Zhang Z, Williford CE, Gadgil S,Surendran S (1999) Improved weather and seasonal climate forecasts from multimodal super-ensemble. Science 285:1548–1550Kumar KK, Soman MK, Kumar KR (1995) Seasonal forecasting of Indian summer monsoonrainfall: a review. Weather 50:449–467Palmer TN (2001) A nonlinear dynamical prospective on model error: a proposal for non localstochastic dynamic parameterization in weather and climate prediction models. Quart J RMeteor Soc 127:279–304Palmer TN, Shukla J (2000) Editorial to DSP/PROVOST. Q J R Meteorol Soc 126:1989–1990Palmer TN, Alessandri A, Andersen U, Cantelaube P, Davey M, Délécluse P, Déqué M, Diez E,Doblas-Reyes FJ, Feddersen H, Graham R, Gualdi S, Guérémy JF, Hagedorn R, Hoshen M,Keenlyside N, Latif M, Lazar A, Maisonnave E, Marletto V, Morse AP, Orfila B, Rogel P,146 R.P. Shukla et al.Terres JM, Thomson MC (2004) Development of a European multimodel ensemble system forseasonal-to-interannual prediction (DEMETER). Bull Am Meteorol Soc 85:853–872Rai S, Pandey AC, Tripathi KC, Dwivedi S (2008) Predictive skill of DEMETER models for windprediction near Madagascar. Indian J Mar Sci 37:62–69RajeevanM, Bhatt J, Kale JD, Lal B (2006) High resolution daily gridded rainfall data for theIndian region: analysis of break and active monsoon spells. Curr Sci 91:296–306Shukla J, Fennessy MJ (1994) Simulation and predictability of monsoons in Proceedings of theinternational conference on monsoon variability and prediction technology report WCRP-84,Geneva. World Climate Research Programme, pp 567–575Shukla J, Anderson J, Baumhefner D, Brankovic C, Chang Y, Kalnay E, Marx L, Palmer T,Paolino D, Ploshay J, Schubert S, Straus D, Suarez M, Tribbia J (2000) Dynamical SeasonalPrediction. Bull Am Meteorol Soc 81:2653–2664Thomson MC, Palmer TN, Morse AP, Cresswell M, Connor SJ (2000) Forecasting disease riskwith seasonal climate predictions. Lancet 355:1559–1560Wilks DS (1995) Statistical methods in atmospheric sciences. Academic, San Diego, pp 233–390Xavier PK, Goswami BN (2007) Analog method for realtime forecasting of summer monsoonsub-seasonal variability. Mon Wea Rev 135:4149–41609 Improved Seasonal Predictability Skill of the DEMETER Models for Central Indian 147.Chapter 10Simulation of Indian Summer MonsoonCirculation with Regional Climate Modelfor ENSO and Drought Years over IndiaSandipan Mukherjee, Ravi P. Shukla, and Avinash C. PandeyAbstract Indian summer monsoon circulation including the monsoon rainfall hasbeen simulated with the regional climate model (RegCM3) for two different yearsassociating an ENSO and a drought year. The horizontal resolution of the model is60 Km covering the entire Indian subcontinent. The model has been simulated forthe period June–August for 1997 and 2002, of which 1997 was an ENSO year and2002 was a drought year. The model sensitivity is examined by using two convec-tive schemes (Kuo type and Grell) and by simulating the characteristics monsoonfeatures of wind at 850 hPa and 200 hPa, temperature at 500 hPa, along with themean seasonal rainfall over central India (70–90E and 15–25N). We find that theimportant monsoon circulation features are well simulated by the model includingthe low mean seasonal rainfall of 2002 over India. We have also find that betweenthe two different parameterization schemes, the Grell runs are giving better resultsthan the other for the rainfall fields. The categorical forecast skill also reveals thatalthough the accuracy of the model is high for 2002, but the probability of detectionof extreme events are higher for 1997.S. Mukherjee (*)K. Banerjee Center of Atmospheric and Ocean Studies, University of Allahabad, Allahabad211002, Indiae-mail: mukherjee.sandipan@rediffmail.comR.P. ShuklaDepartment of Physics, University of Allahabad, Allahabad 211002, Indiae-mail: ravishukla72@gmail.comA.C. PandeyM. N. Saha Center of Space Studies, IIDS, University of Allahabad, Allahabad 211002, Indiae-mail: avinashcpandey@rediffmail.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_10, # Springer-Verlag Berlin Heidelberg 201114910.1 IntroductionFor the prediction of natural systems a variety of models have been proposed.Models have been improved from simple conceptual Lorenz equations to com-plex global climate models. General circulation models (GCMs) are yet notcapable of predicting the behaviour of local or regional climatic systems. Toovercome the deficiency, regional climate models with a higher resolution areconstructed for limited areas. Thus to capture the small-scale physical processesthat drive important local surface variables and for improved predictions of thosevariables, regional climate modeling is necessary. Again from the computationalpoint of view, it is possible to increase the resolution of regional models so as toresolve regional climatic features very well. Several regional models are in usetoday for a wide variety of weather and climatic research and applicationsincluding operational weather forecasting (Giorgi 1990; Dudhia 1993; Bhaskaranet al. 1996).Rainfall over India varies both in space and time during summer monsoonseasons and the large scale rainfall plays an important role in the agricultureplanning, disaster and water resource management of the Indian subcontinent(Gadgil and Rao 2000). The Indian summer monsoon circulation features withthe associated rainfall and its variation is still an interesting problem. There havebeen some studies to simulate monsoon features and extreme weather events overIndia by regional models. Bhaskaran et al. (1996) simulated the summer monsooncirculation with a regional model nested in a GCM found that regional modelderived precipitation is 20% larger than a GCM. Patra et al. (2000) made a com-parative study on the performances of MM5 and regional atmospheric modelingsystem in simulating the Bay of Bengal cyclones. The effect of initial conditionson the simulation of the super cyclone of Orissa was explored by Trivedi et al.(2002) using MM5 model. Dash et al. (2006) have extensively used the regionalclimate model (RegCM3) to simulate the monsoon circulation over India andhave tested the sensitivity of the model incorporating snow over the Tibetanregion.In our present study we have used the regional climate model to simulate themonsoon circulation features associated with rainfall over the Indian subcontinentfor two different years, 1997 and 2002 and compared the circulation pattern withthe observed data as the former year was an ENSO year and the later was a droughtyear. The sensitivity of the model is also explored with different physical parame-terization schemes. In Sect. 10.2, a brief description of the model is given alongwith the initial and boundary condition data. In Sect. 10.3, characteristic monsooncirculation features of wind (850 and 200 hPa), temperature at 500 hPa, and rainfallis analyzed and compared for the two different events and a conclusion has beendrawn in Sect. 10.4.150 S. Mukherjee et al.10.2 Model and Experimental DesignThe regional climate model used in the present work is the version of RegCM3developed by Giorgi et al. (1993a, b). The dynamical core of the regional climatemodel is equivalent to hydrostatic version of fifth generation NCAR, U.S., meso-scale model (MM5). The model is a compressible, grid point model with 18 verticallevels in sigma coordinate. The physics parameterization used in the simulationincludes the radiative transfer package of NCAR CCM3. A planetary boundarylayer schemes and the convective precipitation scheme of modified Kuo schemeand Grell scheme (Grell 1993) is used. These two different precipitation schemesare used to explore the sensitivity of the model. The model also includes anocean flux parameterization and pressure gradient scheme. Different land surfaceprocesses are described by Biosphere–Atmosphere Transfer scheme or BATS(Dickinson et al. 1989). BATS consists of a vegetation layer, three soil layers forsoil water content and a force restore method to calculate the temperature of surfacesoil layer and a subsurface soil layer. BATS is mainly a state-of-the-art land surfacemodel that is used by researchers for long periods.In the present study, the model is simulated over the domain and topographyas shown in Fig. 10.1. The computational domain of the model includes theentire Indian subcontinent only excluding the extreme eastern Himalayan regions.A normal Mercator projection is used with the grid cells of 60 � 60 km size. Forthe model integrations, the central longitude and central latitude is chosen at 80� Eand 20� N. Terrain heights and land use data are generated from the global data setof the United States Geographical Survey (USGS) at 10 min resolution. The initialconditions for the model used are from the NCEP reanalysis dataset-2 (2.5� � 2.5�,L17). The SST data used for the model initialization is the monthly mean andobtained from NOAA optimum interpolated(OI) SST v2, (OISST). The model hasbeen simulated for two different years 1997 and 2002, from June 1 to August 30 andthe mean monsoonal circulation feature is investigated and compared with theNCEP/NCAR reanalysis fields as well as the IMD gridded datasets.10.3 ResultsIn this section, we have explored some of the important characteristics featuresof Indian summer monsoon circulation along with rainfall in detail for the twodifferent years and are compared with corresponding fields from the NCEP/NCARreanalysis. Some of the important monsoon circulation includes westerly jet at850 hPa, the easterly jet at 200 hPa, the temperature at 500 hPa and the surfacepressure pattern. We have mainly explored the variability of the wind fields and thetemperature at 500 hPa along with the rainfall.The observed wind flow of 850 hPa for 1997 and 2002 along with the modelsimulated wind is given in Figs. 10.2–10.4. The observed mean wind field for JJA of10 Simulation of Indian Summer Monsoon Circulation 1511997 and 2002 is given in Fig. 10.2a, b. A strong westerly wind prevails for both theyears over the peninsular region of India and the observed maximum strength ofthe JJA mean westerly wind at 850 hPa is found to be 7.4 m/s for 1997 and that of6.9 m/s for 2002. Using the Grell scheme (Fig. 10.3a) we find that JJA meanwesterly wind is 8 m/s and that of 7.0 m/s for the Kuo scheme (Fig. 10.3b) for1997. When the 2002 wind field of 850 hPa is simulated using the same parameter-ization we find that the simulated mean wind is of 7.7 m/s for Grell scheme(Fig. 10.4a) but it is a fraction smaller for the Kuo scheme (Fig. 10.4b) and is6.7 m/s. Although both the schemes have reasonably well simulated the mean windFig. 10.1 The model domain along with the surface elevation. The dark spot indicates the centerof the model simulation at (80 E and 20 N)152 S. Mukherjee et al.fields of JJA of 1997 and 2002, the Grell scheme is found to simulate the mean wind13% more than actual for 2002.Similarly when the 200 hPa wind fields are simulated for 2002 using the sameparameterization schemes, we found that both the Grell and Kuo schemes areNCEP wind at 850 hPa 1997 32N30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 74E 80E 82E 84E 86E 88E 90E 92 E10222444668810101010101212NCEP wind at 850 hPa 2002 32N30N28N26N24N22N20N18N15N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E222233344566 7788899910101110222233344566 77888999101011abFig. 10.2 (a and b) Observed NCEP/NCAR reanalysis wind fields of 850 hPa for 1997 and 200210 Simulation of Indian Summer Monsoon Circulation 153Model wind at 850 hPa 1997 (Grell)32N30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 74E 80E 82E 84E 86E 88E 90E 92 EModel wind at 850 hPa 1997 (Kuo)32N30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 74E 80E 82E 84E 86E 88E 90E 92EabFig. 10.3 (a and b) Model simulated wind fields of 850 hPa for 1997 using Grell and Kuo schemes154 S. Mukherjee et al.simulating the mean wind well. The Grell scheme exactly matches the observedmean wind speed of 12.3 m/s and the Kuo scheme simulates a fraction larger with12.5 m/s. In case of 1997, the observed mean wind over India for JJA is found to be11 m/s. Simulating this wind field with two different parameterizations; we find thatGrell scheme simulates the mean wind 20% less than the actual.32NModel wind at 850 hPa 2002 (Grell) 30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E32NModel wind at 850 hPa 2002 (Kuo)30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92EabFig. 10.4 (a and b) Model simulated wind fields of 850 hPa for 2002 using Grell and Kuo schemes10 Simulation of Indian Summer Monsoon Circulation 155Figure 10.5a, b shows the observed mean temperature of JJA of 500 hPa for 1997and 2002, respectively. The observed mean temperature for 1997 is found to be269.5 K and that of 2002 is 269.9 K. The simulated temperature by Grell and Kuoschemes are shown in Fig. 10.6a, b for 1997 and in Fig. 10.7a, b for 2002. We findthat the simulated mean temperature for 2002 on Grell run is 270.0 K to and that ofthe Kuo run is 269.5 K. The Grell scheme simulates reasonable results than the Kuo.32Nab30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E268.5268.5270270.5271271.5271.5NCEP temperature at 500 hPa 199732N30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E269269.5270270270.5270271271.5272272NCEP temperature at 500 hPa 2002269.5Fig. 10.5 (a and b) Observed NCEP/NCAR reanalysis temperature fields of 500 hPa for 1997and 2002156 S. Mukherjee et al.The standardized anomaly variation of rainfall of JJA over the central India(70–90E and 15–25N) for 1997 and 2002 is shown in Figs. 10.8 and 10.9 respec-tively. The rainfall variations are compared with the IMD station datasets which arelinearly interpolated to a homogeneous grid over India. The standardized anomalyis calculated by removing the seasonal mean from the data. Although 1997 was an32NabModel simulated temperature 500 hPa 1997 (Grell)30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E267.5268268268.5 268.5269269.5270.5271271.5272272272.5273.5273.5273271.527032NModel simulated temperature 500 hPa 199730N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E268268268.5268.5269269.5270270.5271271 271270.5271.5272272272.5272.5273269267.5268268Fig. 10.6 (a and b) Model simulated temperature fields of 500 hPa for 1997 using Grell andKuo schemes10 Simulation of Indian Summer Monsoon Circulation 157Model simulated temperature 500 hPa 2002 (Grell)32N30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E67.5272.5272.5271.5271.5271.5271271 271271269.5269268.5 268.5268.5269270.5270273Model simulated temperature 500 hPa 2002 (Kuo)32N30N28N26N24N22N20N18N16N14N12N10N8N70E 72E 74E 76E 78E 80E 82E 84E 86E 88E 90E 92E272.5273273.5273274.5271.5272.5270.5271.5271271.5271270.5270269.5269268.5269268.5268.5268268269268abFig. 10.7 (a and b) Model simulated temperature fields of 500 hPa for 2002 using Grell and Kuoschemes158 S. Mukherjee et al.b2.5a21.500.5–0.5–2–1.5–11–2.5Standerdised anomaly1Jun 1Jul 1Aug 30AugTime [1 Jun-30Aug]231Jun 1Jul 1Aug 30AugStanderdised anomaly–3–2–101Time [1Jun-30Aug]c1Jun 1Jul 1Aug 30AugStanderdised anomaly–3–2–10123Time [1Jun-30Aug]Fig. 10.8 (a–c) Standardized anomaly of rainfall over central India (70–90E and 15–25 N) with(a) Grell scheme. (b) Kuo scheme. (c) IMD data for 199710 Simulation of Indian Summer Monsoon Circulation 1591Jul 1AugTime [1Jun-30Aug]Standerdised anomaly30Aug1Jun1Jul 1AugTime [1Jun-30Aug]30Aug1Jun–8–6–4–2024681Jul1Jun–3–2–10123Standerdised anomalyTime [1Jun-30Aug]1Aug 30Augabc–3–2–10123Standerdised anomalyFig. 10.9 (a–c) Standardized anomaly of rainfall over central India (70–90 E and 15–25 N) with(a) Grell scheme. (b) Kuo scheme. (c) IMD data for 2002160 S. Mukherjee et al.ENSO year, we find that the rainfalldistribution during this year was nearly normaland an average seasonal rainfall of JJA over central India was 9 mm/day whichis just little below the seasonal average of 10–15 mm/day for central region(Krishnamurthy and Shukla 2000), though we have excluded the rainfall patternfor September. The simulation of rainfall by Grell scheme for 1997 yields anaverage seasonal rainfall of 10.04 mm/day which is in good agreement with theobserved value. The Kuo scheme simulates very low average seasonal rainfallof 3.0 mm/day virtually concluding the failure of the scheme. We find that thecorrelation between the IMD gridded data and Grell run data is 0.55 and that ofIMD data and Kuo run data is 0.61 but the bias [mean(observed data) – mean(simulated data)] is found to be very high for Kuo run and is equal to 5.9 which isnot the case for Grell run. Here the bias is respectively low and equal to �1.1.For 2002, we find that the average seasonal rainfall over the central Indian regionis 7.1 mm/day which is much lower than the normal seasonal rainfall of India. Thisleads to 2002 as a drought year. When the seasonal rainfall is simulated with Grellscheme we find it 8.3 mm/day with a correlation of 0.40 with the IMD data and abias of�1.2. The Kuo run for 2002 fails drastically with a mean seasonal rainfall of2.3 mm/day with correlation of 0.36 and bias 4.8. Categorical forecast skill measurefor only the Grell scheme and for both the years is shown in Table 10.1. Theaccuracy is of the model is better for 2002, which is 0.71 but the probability ofdetection (POD) is higher for 1997 although the false alarm rate (FAR) and theprobability of false detection (POFD) of 2002 is less than 1997.10.4 ConclusionThe regional climate model (RegCM3) has previously been tested for the monsooncirculation by Dash et al. (2006) and the effect of snow cover is also tested by them.Dimri and Ganju (2007) have also explored the winter climatic behaviour ofnorthern India with RegCM3. In the present study, we have explored the effect ofchange in the sea surface temperature during ENSO period on the monsooncirculation. We have also tested the sensitivity of the climate model for a droughtyear. The change in the monsoon circulation features are then compared for thesetwo different weather events.We find that between the two different cumulus parameterization schemes, Grellscheme is more superior to the Kuo scheme in simulating rainfall fields. In case ofwind field though, the Grell scheme has simulated 13%more for 850 hPa and that ofTable 10.1 Categorical forecast skill measures for the model simulated of seasonal rainfall overcentral for 1997 and 2002. Only the Grell scheme results are compared with IMD observed dataYear Schemes Accuracy POD FAR POFD TS ETS HSS1997 Grell 0.67 0.21 0.85 0.24 0.09 0.00 �0.012002 Grell 0.73 0.06 0.39 0.12 0.04 �0.03 �0.0710 Simulation of Indian Summer Monsoon Circulation 16120% less for 200 hPa for 2002. The temperature variation of 500 hPa is wellsimulated for both the years by Grell scheme with a very little deviation from theobserved value. The mean seasonal rainfall is reasonably well predicted for thedrought year of 2002 by the Grell scheme. The comparison of the forecast measurefor 1997 and 2002 shows that the accuracy of the model is better for 2002, butthe probability of detection of extreme events is higher for 1997.ReferencesBhaskaran BR, Jones G, Murphy JM, Noguer M (1996) Simulations of the Indian summer mon-soon using a nested regional climate model: domain size experiments. Clim Dyn 12:573–578Dash SK, Shekhar MS, Singh GP (2006) Simulation of Indian summer monsoon circulations andrainfall using RegCM3. Theor Appl Climatol 86:161–172Dickinson RE, Erico RM, Giorgi F, Bates GT (1989) A regional climate model for the westernUnited states. Clim Change 15:383–422Dimri AP, Ganju A (2007) Wintertime seasonal scale simulation over western Himalaya UsingRegCM3. Pure Appl Geophys 164:1733–1746Dudhia J (1993) A nonhydrostatic version of the Penn State/NCAR mesoscale model: validationtests and simulation of an Atlantic cyclone and cold front. Mon Wea Rev 121:1493–1513Gadgil S, Rao PRS (2000) Famine strategies for a variable climate – A challenge. Curr Sci78:1203–1215Giorgi F (1990) Simulation of regional climate using a limited area model nested in a generalcirculation model. J Climate 3:941–963Giorgi F, Marinucci MR, Bates GT (1993a) Development of a second generation regional climatemodel (RegCM2). Part I: boundary-layer and radiative transfer processes. Mon Wea Rev121:2794–2813Giorgi F, Marinucci MR, Bates GT, De Canio G (1993b) Development of a second-generationregional climate model (RegCM2). Part II: convective processes and assimilation of lateralboundary conditions. Mon Wea Rev 121:2814–2832Grell GA (1993) Prognostic evaluation of assumptions used by cumulus parameterizations. MonWea Rev 121:754–787Krishnamurthy V, Shukla J (2000) Intraseasonal and interannual variability of rainfall over India.J Clim 13:4366–4377Patra KP, Santhanam MS, Potty KVJ, Tewari M, Rao PLS (2000) Simulation of tropical cyclonesusing regional weather prediction models. Curr Sci 79(1):70–78Trivedi DK, Sanjay J, Singh SS (2002) Numerical simulation of a super cyclonic storm, Orissa1999: impact of initial conditions. Meteorol Appl 9:367–376162 S. Mukherjee et al.Chapter 11Changes in surface temperature and snowover the Western Himalaya Under Doublingof Carbon Dioxide (CO2)P. Parth Sarthi, S.K. Dash, and Ashu MamgainAbstract Global warming has caused the world’s surface temperature to rise at anunprecedented rate and these changes have caused the snow and ice to melt rapidly.The variability of temperature as consequence of global warming plays a crucialrole on the snow over the western Himalaya which is a natural water reservoir byreleasing large quantity of fresh water throughout the year in the important rivers ofnorth India. The change in snow as an impact of global warming over this regionmay affect the water resources in north Indian rivers. The decline of water in snowfed rivers, with disappearance and melting of snow, glaciers and ice sheets mayhave a direct impact on the lives of millions of people who depend on these snow-fed rivers and hence there may be economic losses.The impact of the Himalayan snow on Indian Summer Monsoon Rainfall(ISMR) and Land surface Hydrology in the northern Indian during Indian SummerMonsoon (ISM) region is vital and therefore it is necessary to examine theirvariability in past, present and future time slices. Since Western Himalaya (WH)snow does have a relation with ISMR and also important holy rivers like the Gangaoriginates, therefore the domain of WH is chosen for the current study. The Modelfor Interdisciplinary Research On Climate (MIROC3.2 hires) simulated snow depthand surface temperature in experiments twentieth century simulation (20c3m) anddoubling of CO2 simulation (1pctto2x) are utilized. To know the response ofdoubling of CO2 in atmosphere on snow depth and surface temperature over WH,the model simulated snow depth and surface temperature is analyzed under thedoubling of CO2 (1pctto2x) and twentieth century (20c3m) simulation. snow depthis decreased more in 1pctto2x in compare to 20c3m simulation. It is found that theP.P. Sarthi (*)Centre for Environmental Sciences, Central University of Bihar, Patna 800014, Indiae-mail: drpss@hotmail.comS.K. Dash • A. MamgainCenter for Atmospheric Sciences, Indian Institute of Technology Delhi, Hauz Khas,New Delhi-110016, Indiae-mail: skdash@cas.iitd.ac.in; ashumam@gmail.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_11, # Springer-Verlag Berlin Heidelberg 2011163increase of surface temperature and decrease of snow depth are more during May toDecember withpositive and negative trends in surface temperature and snow depthrespectively.11.1 Introductionsnow is regarded as an important indicator of climate change because of itsinfluence on energy and moisture budget on the earth surface. The increase ordecrease in snow depth directly influences the prevailing surface temperature andleads to change in wind circulation pattern. Thus snow depth anomaly provides asink or source of the surface temperature. When snow depth is more over an area,more radiation is required to melt the snow and small amount of radiation is left towarm the earth surface. The inverse is true in the case of less snow depth overthe area. This is the classic temperature-snow feedback mechanism, which iskey component in climate models. The snow/sec-ice feedback has a significantimpact on the sensitivity of a climate model. If a marginally snow-covered areawarms, snow tends to melt, lowering the albedo, and hence leading to moresnowmelt (the ice-albedo feedback). This is the basis for predictions of enhancedwarming in seasonally snow covered regions as a result of global warming.The link between the Himalayan winter snow and ISMR has been recognized byBlanford (1884), Walker (1910), Dey and Kumar (1983), and Dickson (1984). Thelinear correlation coefficients between December-to-March snow extent in theHimalayas and June-to-Septembermonsoon rainfall were found to be approximately�0.6 by Dey and Kumar (1983) and Dickson (1984). Kripalani et al. (2003)presented the monthly climatology and variability of the Indian National Satellite(INSAT) derived snow cover estimates over the western Himalayan region. Theysuggested that the changes in observed snow cover extent and snow Depth due toglobal warming may be a possible cause for the weakening winter snow-ISMRrelationship. Using Regional Climate Model (RegCM3), a sensitivity experimentbased on snow depth anomaly over Tibet has been conducted and associated changesin circulation pattern and rainfall over India were examined by Dash et al. (2007).The snow in the Himalaya is a sensitive indicator of climate change. TheHimalayan region as a whole has warmed by about 1.8�F since the 1970s (Shresthaet al. 1999; UNEP 2002). Earlier studies show decline in snow depth/cover overmost parts of the world due to global warming (IPCC). The observational studiesalso indicated that the northern hemisphere annual snow cover extent has decreasedby about 10% since 1966. The observational studies indicated a decrease in snowextent over the past decade (Robinson and Serreze 1995; Karl et al. 1993; Groismanet al. 1994). Reduction in snow cover during the mid to late 1980s was stronglyrelated to temperature increase in snow covered area. There were highly significantinterannual and multidecadal correlations between increase in the northern hemi-sphere spring land temperature and a reduction in the northern hemisphere springsnow cover (IPCC 2001).164 P.P. Sarthi et al.The simulation of climate change projects a reduction in the extent and durationof snow cover in response to the warming condition (Kattenbergg et al. 1996).Hengchun and Mather (1997) used many general circulation models and suggestedunder double CO2 scenarios, the increase of surface temperature all over the globe.They suggested possible changes in snow accumulation due to increasing CO2.Wright et al. (2005) run the climate model based on IS92a emission scenario torepresent future climate and shown an increase of surface temperature which leadsto significant decrease in the net volume of ice.The water derived from snow melting in the Himalaya region is used fordrinking, agriculture, and power generation therefore a systemic and sustainedstudy of hydro-meteorological processes of snow regime is necessary for evaluatingchanges in the hydrology of Mountain Rivers. The variation in snow depth over theHimalaya will affect the monsoon rainfall and surface runoff especially overnorthern India.In this paper, it is aimed to study the response of the snow depth and surfacetemperature variability in twentieth century and doubling of CO2 experiments. Theintroduction along with literature survey is kept in Sect. 11.1. The Sect. 11.2describes the selection of the IPCC AR4 model and data for the current study.The results and discussions are briefly outlined in Sect. 11.3. Conclusions areplaced in Sect. 11.4.11.2 Sensitivity Experiments Using MIROC ModelImportant rivers in northern India originate from the Himalayas. The snow/glacierover the Western Himalaya is the main contributor of water in those rivers.Therefore, it is necessary to understand the characteristics and behavior of snowdepth and surface temperature over the WH. The characteristics of snow depth andsurface temperature provide an insight to develop understanding on its impact onhydrological condition. For the current study, the selected domain for WH is72�–80�E and 30�–39�N.Program for Climate Model Diagnosis and Inter-comparison (PCMDI), at theLawrence Livermore National University, USA, is volunteered to collect modelsimulated output, which are contributed by world leading modeling centers. Cli-mate model simulated output of the past, present and future climate was collectedby PCMDI mostly during the years 2005 and 2006, and this archived data is keptunder phase 3 of the Coupled Model Inter-comparison Project (CMIP3). Based onclimate models simulated result, various scientific research papers have givencontribution in Fourth Assessment Report (AR4) of the International Panel onClimate Change (IPCC).All coupled climate model, as used in CMIP3, do not simulate snow depth butthey simulate snow cover mainly. Since the aim of the current study is to analyzethe response of snow depth along with surface temperature, therefore only thosemodels are selected which simulates snow Depth. The list of climate models, which11 Changes in surface temperature and snow over the Western Himalaya 165simulated snow Depth, is given below in Table 11.1. The details of the above listedmodels are available at http://www.pcmdi.llnl.gov/ipcc/model_documentation/.The surface resolution of the above listed models is varying from 1.1� � 1.1� to3.7� � 3.7�. As WH is a small region, so higher surface resolution model willprovide more number of data points in chosen domain. Therefore, MIROC 3.2(Hires) model of Japan is found suitable for the current study. MIROC 3.2 (Hires)model simulated snow depth and surface temperature data for the period of1900–2000 and 2001–2080 in the experiments of twentieth century “20c3m and“1pctto2x” respectively is considered for the analysis to understand the role of CO2on snow depth and surface temperature over WH. The details of the twentiethcentury and Doubling of CO2 experiment are available IPCC official web site.11.3 Changes in snow Depth and surface temperatureThe annual variation in mean of November to April snow depth in 20c3m and1pctto2x experiments is shown in Fig. 11.1a, b. The values are ranging from 0.7 to1.1 m in Fig. 11.1a. The mean snow depth (Fig. 11.1a) does not show any significantchange. In Fig. 11.1b, the mean of November to April snow depth is ranging from0.4 to 1.1 m. and a gradual decreasing in mean snow depth is found when CO2 isdoubled. It is important to note that the mean of November to April snow depth isshowing a sharp decrease from 2052 onwards.In Fig. 11.2, variation during 20c3m and 1pctto2x simulation is depicted for20c3m and 1pctto2x. The pattern in mean monthly variation of snow depth in bothexperiments is well agreed to each other. It is found that the mean monthly values ofsnow depth in each month during 20c3m are excess in comparison to 1pctto2xsimulation. The minimum value of snow depth is noticed during August–Septemberin 20c3m; from there it slowly increases in the respective months and finallyattainedthe peak in the month of April and May. From the month of June, its startsdecreasing and meet the minimum values in the month of August/September.Similarly, in 1pctto2x, the mean monthly values of snow depth are noticed in themonth of September. From October onwards, it started to increase and attainsTable 11.1 Climate models with their IPCC ID, affiliated country, their resolution, key referencesand convection schemesS. No. IPCC ID Countrysurfaceresolution Key referenceConvectionscheme1 MIROC 3.2 (Hires) Japan 1.1 � 1.1 K-1 model developers (2004) AS2 CGCM3.1 Canada 3.7 � 3.7 Flato et al. (2000) MC3 CNRM-CM3 France 2.8 � 2.8 Salas-Meliá et al. (2005) MF4 BCC-CM1 China 1.9 � 1.9 Not available –5 BCCR-BCM2.0 Norway 2.8 � 2.8 Furevik et al. (2003) MFAS Arakawa-Schubert, MC Moist Convection adjustment, MF Mass Flux based166 P.P. Sarthi et al.0.00.20.40.60.81.01.2ab1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000YearsMean Snow Depth (m) Nov-April0.00.20.40.60.81.01.22001 2011 2021 2031 2041 2051 2061 2071YearsMean Snow Depth (m)Nov-AprilFig. 11.1 (a and b) Annual variation in mean of Nov-April snow depth (m) in (a) 20c3m and(b) 1pctto2x1900-20002001-20800.00.20.40.60.81.01.21.4Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecMonthsMean Snow Depth (m) Fig. 11.2 Variation of mean monthly snow depth (m) in 20c3m and 1pctto2x experiments11 Changes in surface temperature and snow over the Western Himalaya 167maximum values of snow depth in the month of April. From May onwards, thesnow depth is decreasing and attained its lowest values in the month of August/September.Figure 11.3a, b show monthly variation of surface temperature and snow depthin 20c3m and 1pctto2x experiments. In both figures, the increase in surfacetemperature is well agreed with decrease in snow depth especially during monthsof May to December. From the months of January to April, surface temperature andsnow depth does not show significant relation. It seems that the surface temperatureis influencing snow depth during May–December over WH.Figure 11.4 shows the monthly variation in trend (negative) values of snowdepth in 20c3m (in 101 years) and 1pctto2x (in 79 years). The pattern of trendvalues is well agreed to each other during April–December in both experiments.During January to March, the pattern is not same for experiments. In both-20-15-10-505101520abJan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecJan Feb Mar May Jun Jul Aug Sep Oct Nov DecMonthsTAS (deg C) & SD (m*10.)ASTSD-20-15-10-505101520MonthsTAS (deg C) & SD (m*10.)AprASTSDFig. 11.3 (a and b) Mean monthly variation of atmospheric surface temperature (AST) and snowdepth (SD) in experiments (a) 20c3m and (b) 1pctto2x168 P.P. Sarthi et al.experiments, it is found that during May–December, trend (negative) values aremore in 1pctto2x and less in 20c3m. The trend (negative) values are found maxi-mum in June and minimum in September in both experiments.The monthly variation of trend (positive) values in surface temperature anomalyin 20c3m and 1pctto2x are shown in Fig. 11.5. The values of trend (positive)are maximum in 1pctto2x and minimum in 20c3m experiment whereas pattern ofboth series shows a very good agreement in all months except month of April andJune. A sharp increase in trend (positive) values are noticed from July to December.-1.0-0.8-0.6-0.4-0.20.0Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecMonthsDecreasing trend values20c3m1pctto2xFig. 11.4 Decreasing trend in snow depth anomaly (m) in 20c3m (in 101 years) and 1pctto2x(in 79 years)0.02.04.06.08.0Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecMonthsIncreasing trend values20c3m1pctto2xFig. 11.5 Increasing trend in surface temperature anomaly (m) in 20c3m (in 101 years) and1pctto2x (in 79 years)11 Changes in surface temperature and snow over the Western Himalaya 169The trend (positive) values are not very significant during January to June in bothexperiments.11.4 ConclusionsThe important results of this study indicate that model simulated annual variation ofmean of November to April snow depths undergo a remarkable change in doublingof CO2. The snow depth may slowly attain its low value from the year 2052onwards. The mean monthly values during 1900–2000 (20c3m) are more comparedto those in the period 2001–2080 (1pctto2x). In both the experiments, the inverserelation between snow depth and surface temperature exists during May–Decemberwith stronger magnitude in 1pctto2x compared to 20c3m simulation. The monthlyvariations of trend values of negative and positive snow depths and surface temper-ature anomalies are more in doubling of CO2 (1pctto2x) case compared to those inthe twentieth century simulation (20c3m). These results show the clear impact ofCO2 on the snow depth over the Western Himalaya.ReferencesBlanford HF (1884) On the connection of the Himalayan snow with dry winds and seasons ofdrought in India. Proc R Soc Lond 37:3–22Dash SK, Sarthi PP, Shekhar MS (2007) Influence of Eurasian and Tibetan snow on IndianSummer Monsoon. In: Vinayachandran PN (ed) Understanding and forecasting of monsoons.Centre for Science and Technology of the Non-Aligned and Other Developing Countries(NAM S & T Center), New Delhi, pp 108–118Dey B, Kumar B (1983) Himalayan winter snow cover area and summer monsoon rainfall overIndia. J Geophys Res 88:5471–5474Dickson RR (1984) Eurasian snow cover versus Indian monsoon rainfall – an extension of Hahnand Shukla results. J Clim Appl Meteorol 23:171–173Flato GM, Lee WG, McFarlane NA, Ramsden D, Reader MC, Weaver AJ (2000) The CanadianCentre for Climate Modelling and Analysis Global Coupled Model and its climate. ClimateDyn 16:427–450Furevik T, Bentsen M, Drange H, Kindem IKT, Kvamstø NG, Sorteberg A (2003) Descriptionand evaluation of the Bergen climate model: ARPEGE coupled with MICOM. Climate Dyn21:27–51Groisman PYa, Karl TR, Knight RW (1994) Changes of snow cover, temperature, and radiativeheat balanceover the Northern Hemisphere. J Clim 7:184–204Hengchun Ye, Mather JR (1997) Polar snow cover changes and global warming. Int J Climatol 17(2):155–162IPCC (2001) In: Houghton JT et al (eds) Climate change 2001, the scientific basis. CambridgeUniversity Press, CambridgeKarl TR, Groisman PYa, Knight RW, Heim RR (1993) Recent variations of snow cover andsnowfall in North America and their relation to precipitation and temperature variations. J Clim6:1327–1344170 P.P. Sarthi et al.Kattenberg A, Giorgi F, Grassl H, Meehl GA, Mitchell JFB, Stouffer RJ, Tokioka T, Weaver AJ,Wigley TML (1996) Climate change 1995: the science of climate change. In: Houghton JT,Meira Filho LG, Callander BA, Harris N, Kattenberg A, Maskell K (eds) Contribution ofworking group I to the second assessment report of the intergovernmental panel on climatechange. Cambridge University Press, Cambridge, p 572Kripalani RH, Kulkarni A, Sabade SS (2003) Western Himalayan snow cover and Indian monsoonrainfall: a re-examination with INSAT and NCEP/NCAR data. Theor Appl Climatol 74:1–18K-1 Model Developers (2004) K-1 Coupled Model (MIROC) Description. K-1 Technical Report 1(edited by Hasumi H, Emori S), Center for Climate System Research, University of Tokyo,Tokyo, Japan, p 34 [http://www.ccsr.utokyo.ac.jp/kyosei/hasumi/MIROC/tech-repo.pdf]Robinson DA, Serreze MC (1995) Recent variations and regional relationships in NorthernHemisphere snow cover. Ann Glaciol 21:71–76Salas-Meliá D, Chauvin F, Déqué M, Douville H, Gueremy J, Marquet P, Planton S, Royer J,Tyteca S (2005) Description and validation of the CNRM-CM3 global coupled model.CNRMTech Rep 103, p 36Shrestha AB, Wake CP, Mayewski PA, Dibb JE (1999) Maximum temperature trends in theHimalaya and its vicinity: an analysis based on temperature records from Nepal for the period1971–94. J Clim 12:2775–2787UNEP (2002) Impact of global warming on mountain areas confirmed by UNEP-backedmountaineers. GRID-Arendel News, United Nationas Environmental Program, June 5, 2002.http://www.grida.no/inf/news/news02/news41.htmWalker GR (1910) Correlations in seasonal variations of weather. Mem India Meteorol Dept21:22–45Wright A, Wadham J, Siegert M, Luckman A, Kohler J (2005) Modelling the impact ofsuperimposed ice on the mass balance of an arctic glacier under scenario of future climatechange. Ann Glaciol 42(1):277–28311 Changes in surface temperature and snow over the Western Himalaya 171.Chapter 12Simulation of Tornadoes over IndiaUsing WRF-NMM ModelA.J. Litta, U.C. Mohanty, S.C. Bhan, and M. MohapatraAbstract A severe thunderstorm produced a tornado (F0 on the Fujita-Pearsonscale), close to Ludhiana airport (Punjab), north-west region of India (30.8� N,76.0� E) on 15 August 2007. Another severe thunderstorm produced a tornado(F3 on the Fujita-Pearson scale) which affected Rajkanika block of Kendraparadistrict of Orissa (20.7� N, 86.6� E) in the afternoon of 31 March 2009. An attemptis made to simulate this rare events using Non-hydrostatic Mesoscale Model(NMM) core of the Weather Research and Forecasting (WRF) system with a spatialresolution of 3 km for a period of 24 h. The results of the study can be utilized fornow casting of severe thunderstorms.12.1 IntroductionTornadoes are a rare weather phenomenon involving a violently rotating column ofair, which is in contact with both a cumulonimbus cloud and the surface of the earth.Spawned from powerful thunderstorms, tornadoes can cause fatalities and devastatea neighborhood in seconds. Tornadoes come in many sizes but are typically in theform of a visible condensation funnel, whose narrow end touches the earth and isoften encircled by a cloud of debris. The tornadoes can pick up motor cars and evenbuses, twist steel bridges and destroy houses and factories along their path.Tornadoes have been observed to occur in every continent except Antarctica.A.J. Litta (*) • U.C. MohantyCentre for Atmospheric Sciences, Indian Institute of Technology, Delhi, Hauz Khas, New Delhi110 016, Indiae-mail: ajlitta@gmail.com; mohanty@cas.iitd.ernet.inS.C. Bhan • M. MohapatraIndia Meteorological Department, Lodi Road, New Delhi 110 003, Indiae-mail: scbhan@gmail.com; mohapatra_imd@yahoo.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_12, # Springer-Verlag Berlin Heidelberg 2011173This dangerous phenomenon occurs mostly in the United States, but occasionallyoccurs in other parts of the world. India is also not free from occurrences of suchtornadoes.A comprehensive pre-monsoon (March–May) tornado climatology was devel-oped for the eastern parts of the Indian subcontinent particularly West Bengaland Orissa. The most comprehensive works were by Petersen and Mehta (1981),which documented 51 possible tornadoes across Bengal, 18 of which in eachcase killed 10 people or more. Twelve of these occurred from 1838 to 1963 and24 occurred after 1968. However, there might exist a tendency to report onlythe relatively significant tornadoes that leave more damage and attract moreattention. Between 1972 and 1978, 13 tornado events occurred in the area approxi-mately coinciding with Bangladesh. Considering the entire area of the country, thisgives a frequency rate of occurrence of about 1 � 10�5 year�1 km�2 (Goliger andMilford 1998). Goldar et al. (2001) documented 36 possible spring tornadoes overWest Bengal, 14 of which killed ten people or more during 1890–1900. Otherstudies such as Singh (1981) have listed a few tornadoes for India and the associatedwind speeds have been estimated to be of the order of 200–400 km h�1.Northwest India does not frequently experiences this violent weather phenome-non; but there have been a few cases over the region. In northern Delhi, 28 peoplewere killed and 700 were injured by a tornado that cut a path 5 km long andabout 50 m wide on 17 March 1978. Another tornado is reported to have killedten people near Ludhiana (Punjab) on 10 March 1975 (Kumar and Singh 1978;Kumar et al. 1979). Although only a few tornadoes occur over this part ofthe country, they have a great potential of causing damage to property and loss oflife. It is, therefore, essential to investigate the occurrence of tornadoes with thecapabilities of high resolution mesoscale models in simulating these highlylocalized events.Extensive efforts have been made for several decades to reveal the generationmechanism and structure of tornadoes. Because of the short lifetime and the smallhorizontal scales of the vortices, however, comprehensive data that reveal theirstructure and generation process have not been obtained so far. However, therecent developments of computer technology and mesoscale numerical modelshave allowed a numerical simulation to be a promising tool to reveal thedynamics of tornadoes (Noda and Niino 2005). In the present study, an attemptis made to simulate two recent tornadoes that occurred over Ludhiana on 15August 2007 and over Orissa on 31 March 2009 using a Non-hydrostatic Meso-scale Model (NMM) core of the Weather Research and Forecasting (WRF)system developed by the National Oceanic and Atmospheric Administration(NOAA)/ National Centers for Environment Prediction (NCEP). The outline ofthis paper is as follows: Sect. 12.2 gives a brief description of both tornadoevents; Sect. 12.3 presents the description of numerical model and configurations.The results and discussion are described in Sect. 12.4 and the conclusions inSect. 12.5.174 A.J. Litta et al.12.2 Case Description12.2.1 Case 1-Tornado over LudhianaA tornado was reported to affect the agricultural fields close to Ludhiana Airportin the afternoon of 15 August 2007. The phenomenon occurred at about 1100UTC and covered a distance of about 1–1.5 km in agricultural fields of UcchiMangli and Sahanewal villages. The tornado moved south-southeastwards coveringthe distance of about 0.75–1 km. It then turned, west-southwards for about100–200 m and then moved in a zigzag way. The phenomenon was so destructivethat a few big trees were uprooted, heavy branches were snapped of the trees, tinroof of a tube well room was blown away and a bund maker weighing about 80 kgwas reportedly lifted to a height of 60–70 ft and was seen rotating in the debris. Thewhole of the standing water in the paddy field was sucked in and the field was leftdry. The plants of Sorghum crop were found flattened in west southwest andnortheast direction indicating clockwise rotation in the tornado. The tornadobegan to weaken by 1115 UTC and disappear completely by 1120 UTC. Consider-ing these damages, the intensity of this tornado can be estimated to be of F0(18–32 ms�1) according to Fujita-Pearson scale (Fujita 1981).The records of the meteorological observatory situated at Ludhiana Airport,about 1–1.5 km from the site of the tornado, which records hourly observationsfrom 0000 UTC to 1200 UTC during the day have been used to describe weatherover the region. The observation report indicated that the sky was mainly sunny inthe morning of 15 August 2007 with light southerly winds. A temperature of 29�Cwith 85% relative humidity was recorded at 0300 UTC. By 0600 UTC, thetemperature rose to 32�C with 68% relative humidity and 4 okta low clouds wereobserved. Cumulus clouds developed in the afternoon and 6 okta cumulonimbusclouds were observed at the time of the event. Though, only 0.3 mm rain wasrecorded at the observatory. General weather phenomenon reportedby the obser-vatory during the day were mist and haze from 0345 to 0925 UTC, thunderstormfrom 1113 to 1120 UTC and funnel cloud from 1105 to 1120 UTC. The weatherradar situated at Patiala (about 80 km southeast of Ludhiana) reported scatteredcumulonimbus clouds at a distance of about 200 km north-northeastwards of Patialaat 0600 UTC. No radar echoes were reported from 0700 to 1000 UTC. The radarreported a cumulonimbus cloud with its top at 6 km and located about 80 kmnorthwest of Patiala (close to the area of occurrence of the tornado) at 1100 UTC.12.2.2 Case 2-Tornado over OrissaA severe thunderstorm produced a tornado, which affected Rajkanika block ofKendrapara district of Orissa in the afternoon of 31 March 2009. This tornado wasembedded in a severe squall line. The devastation caused by the tornado consumed12 Simulation of Tornadoes over India Using WRF-NMM Model 17515 lives, left several injured with huge loss of property. The origin of the tornadowas at Ostia village of Rajkanika and the tornado touched down in this villageaccording to eye witnesses at 1110 UTC of 31 March 2009. It then passed overBarada, Badatala, Asasa, Baghabuda, Gobindpur, Kantapada, Gobindapur, Ganja,Achutapur, Mukundapur, Manaidiha, Dasabhagaria, Mangalpur and Dalikaindavillages. The tornado moved from northeast to southwest from Ostia villageto Ganja and then from north-northwest to south-southeast towards Dalikaendavillage. The duration of the tornado was 10 min. The most affected area may beapproximately of 0.6 km in width and 6 km in length covering the above villages.Hence, it moved with the speed of about 36 km h�1. The Baghabuda village underRajkanika Panchayat witnessed the maximum devastation. The phenomenon wasso destructive that most of the trees either broke apart or fully uprooted, the bamboobush of about 15 ft diameter was completely uprooted and all the electricity polesincluding power transmitter were laid down to ground. The fishes weighing about1–2 kg were scattered here and there in the village which were thrown away by thetornado from a nearby pond. Some RCC roofs of non-structural concrete buildingswere blown away by 30–40 ft. Considering these damages associated with thetornado over Rajkanika can be estimated to be a strong tornado with an intensity ofF3 (70–92 ms�1) according to Fujita-Pearson scale (Fujita 1981).With the occurrence of tornado, heavy lightning, thunder and moderate to heavyhailstorm occurred which was accompanied by rain for about an hour and brokenlightning persisted till evening (1230 UTC). The hailstone diameter was confined toless than 2 cm at most places and at some places their diameters ranged from 2 to5 cm. In the approaching funnel cloud a very bright red color was observed up toa height of 50–100 m above ground, which looked peculiarly reddish at the touchdown point. The analysis of current weather observations from Chandbali (nearestmeteorological observatory), which lies at about 5 km to the north-northeast ofRajkanika, indicated that the cumulonimbus cloud developed to the south ofChandbali at about 1030 UTC. The first thunder was heard at 1035 UTC andcontinued till 1250 UTC. The rainfall started at Chandbali at 1100 UTC andcontinued till 1230 UTC. The hailstorm commenced over the station at 1110UTC and continued till 1210 UTC. Chandbali reported 41.0 mm of rainfall andhail stones with diameter of about 3 cm. The environmental temperature overChandbali gradually increased from 27.4�C at 0000 UTC and reached up to 30�Cprior to the occurrence of tornado (1100 UTC). It suddenly fell thereafter to become21.6�C at 1200 UTC. It increased again and became 25.0�C at 1500 UTC.12.3 Modeling SystemThe Non-hydrostatic Mesoscale Model (NMM) core of the Weather Researchand Forecasting (WRF) system is a next-generation mesoscale forecast modelthat is used to advance the understanding and the prediction of mesoscale intenseconvective systems. The WRF-NMM model is an efficient, state-of-art and flexible176 A.J. Litta et al.mesoscale modeling system for use across a broad range of weather forecast andidealized research applications, with an emphasis on horizontal grid sizes in therange of 1–10 km. Several studies related to the simulation of severe thunderstormevents using NMM model have been performed (Kain et al. 2006; Litta andMohanty 2008).The NMM is a fully compressible, non-hydrostatic mesoscale model witha hydrostatic option (Janjic 2003). The model uses a terrain following hybridsigma-pressure vertical coordinate. The grid staggering is the Arakawa E-grid.The model uses a forward-backward scheme for horizontally propagating fastwaves, implicit scheme for vertically propagating sound waves, Adams-Bashforthscheme for horizontal advection, and Crank-Nicholson scheme for vertical advec-tion. The same time step is used for all terms. The dynamics conserve a number offirst and second order quantities including energy and entropy.In the present simulation, the model was integrated for a period of 24 h, startingat 0000 UTC of 15 August 2007 for the first case and starting at 0000 UTC of31 March 2009 for the second case as initial values. A single domain with 3 kmhorizontal spatial resolution was configured, which is reasonable in capturing themesoscale cloud clusters. Initial conditions for the 3 km domain are derived from6-hourly FNL Global Analyses at 1.0� � 1.0� grids. Analysis fields, includingtemperature, moisture, geopotential height and wind, are interpolated to the meso-scale grids by the WRF preprocessing system (WPS). These derived fields are usedas initial conditions for the present experiments. There are 38 unequally spacedsigma (non-dimensional pressure) levels in the vertical. The physical parameter-izations used in this study are Geophysical Fluid Dynamics Laboratory (GFDL) forlong wave and short wave radiation, NMM Land surface scheme for land surface,Mellor Yamada Janjic (MYJ) scheme for planetary boundary layer, Ferrier schemefor microphysics and Janjic similarity scheme for surface layer. The cumulusparameterization used for this study is Grell-Devenyi cloud ensemble scheme(Grell and Devenyi 2002). This is a multi-closure, multi-parameter, ensemblemethod with typically 144 sub-grid members. Table 12.1 gives a brief illustrationon the model configuration of the present study.Table 12.1 NMM model configurationDynamics Non-hydrostaticHorizontal spatial resolution 3 kmIntegration time step 6 sMap projection Rotated latitude and longitudeHorizontal grid system Arakawa E-gridVertical co-ordinateTerrain-following hybrid (sigma-pressure)vertical coordinate (38 sigma levels)Radiation parameterization GFDL/GFDLSurface layer parameterization Janjic similarity schemeLand surface parameterization NMM land surface schemeCumulus parameterization Grell-Devenyi ensemble schemePBL parameterization Mellor-Yamada-JanjicMicrophysics Ferrier (new eta) scheme12 Simulation of Tornadoes over India Using WRF-NMM Model 17712.4 Results and DiscussionAsnani (2005) have described some of the favorable meteorological conditions forsevere convection in respect of tornado-genesis. They include plentiful supply ofmoisture in the lower levels, presence of convective instability in a sufficientmeasure in deep layer, the lifting mechanism to produce low-level convergenceand upper level divergence to initiate and augment the release of convectiveinstability and strong vertical wind shear of the horizontal winds. The followingsection describes the above mentioned features obtained from the WRF-NMMmodel.12.4.1 Surface and Upper air PatternThe prediction of the dominant convective mode is based on the assessment ofmagnitude of vertical motion, which is needed to initiate convection (May andRajopadhyaya 1999). Observations as well as theoretical calculations show thatmosttornadoes show a strong vertical upward motion at the centre, which is ofthe same order of magnitude as horizontal component of wind (Asnani 2005).Figure 12.1a shows the time-height cross section of model simulated pressurevertical velocity (Pas�1) over Ludhiana valid from 0000 UTC of 15 August 2007to 16 August 2007 and Fig. 12.1b over Rajkanika valid from 0000 UTC of 31 March2009 to 0000 UTC of 1 April 2009. The figure shows a high upward velocity with amagnitude of �12 Pa s�1 over Ludhiana at 1100 UTC on 15 August 2007, which isconsistent with the tornado occurrence. The Fig. 12.1b clearly shows a strongupward velocity with a magnitude of �40 Pa s�1 over Rajkanika at 0900 UTC on31 March 2009, which is the model predicted tornado hour. The model wellsimulated high magnitude pressure vertical velocity at the tornado site for boththe cases, which is an important factor related to tornado genesis.Figure 12.2a illustrates the model simulated moisture convergence at 850 hPa at1100 UTC of 15 August 2007. The maximum convergence (convergence is positiveand divergence is negative) over Ludhiana happens at 1100 UTC which is consis-tent with the weather report. Figure 12.2b illustrates the model simulated moistureconvergence at 850 hPa at 0900 UTC of 31 March 2009. This figure showsmaximum convergence over Rajkanika at 0900 UTC which is again consistentwith the tornado occurrence. In both the tornado cases, the model well simulatedstrong convergence over the tornado site during the model predicted tornado hour.Most of the severe thunderstorms produce heavy rainfall during their lifecycle of1–3 h (Vaidya 2007). The time-series plots of 3-hourly NMM model simulatedand TRMM accumulated rainfall over Ludhiana valid for 15 August 2007 at 0000UTC to 16 August 2007 at 0000 UTC is given in the Fig. 12.3a. The modelsimulated accumulated rainfall at Ludhiana is 15.96 mm which is close to theTRMM accumulated rainfall (26.0 mm). The model results in terms of intensity,178 A.J. Litta et al.time and location are very close to the observation. The inter-comparison of3-hourly model simulated rainfall and observed rainfall at Chandbali meteorologi-cal station valid from 0000 UTC of 31 March 2009 to 0000 UTC of 1 April 2009 isillustrated in Fig. 12.3b. The figure clearly shows a high rainfall (35.44 mm) overChandbali between 0900 UTC to 1100 UTC during the model simulated tornadohour, which is very close to the actual observation (41.0 mm).100ab200300400500Pressure Levels (hPa)Pressure Levels (hPa)6007008009001000100200300400500600700800900100000Z15AUG00Z16AUG03Z–12 –11 –10–40 –35 –30 –25 –20 –15 –10 –5 0–9 –8 –7 –5 –4 –3 –2 –1 006Z 09Z 12ZTime (UTC)15Z 18Z 21Z00Z31MAR00Z1APR03Z 06Z 09Z 12ZTime (UTC)15Z 18Z 21ZFig. 12.1 (a and b) Time-Height cross section of simulated pressure vertical velocity (Pas-1)(a) over Ludhiana valid from 0000 UTC of 15 August 2007 to 0000 UTC of 16 August 2007 and(b) over Rajakanika valid from 0000 UTC of 31 March 2009 to 0000 UTC of 1 April 200912 Simulation of Tornadoes over India Using WRF-NMM Model 179Studies have indicated that a meso-cyclonic flow results in the formation of asevere vortex like a tornado (Asnani 2005). The plants of Sorghum crop were foundflattened in west southwest and northeast direction indicating clockwise rotationin the tornado of Ludhiana. To the western side of the track the uprooted treeswere aligned from north to south and to the eastern side of the track the uprootedtrees were aligned from south to north in the case of tornado over Orissa.It indicated that the ground circulation associated with the tornado was cyclonicand hence this particular tornado was associated with a cyclonic vortex. However,the observations are not sufficient to draw a conclusion that there was a develop-ment of vortex or suction vortex in the vicinity of these latest tornadoes. The modelsimulated surface wind at 1100 UTC of 15 August 2007 is given in Fig. 12.4a.31.4Nabmoisture convergence at 850 hPa (11 UTC)moisture convergence at 850 hPa (09 UTC)31.2N31N30.8N30.6N30.4N30.2N30N29.8N29.6N19N84E 84.5E 85.5E 86.5E 87.5E 88.5E 89.5E89E85E 86E 87E 88E19.5N20N20.5N21N21.5N22N22.5N23N74.4E 74.7E 75E 75.3E 75.6E 75.9E 76.2E 76.5E 76.8E–90–60–300306090120150180–90–60–300306090120150180Fig. 12.2 (a and b) The model simulated moisture convergence at 850 hPa valid at (a) 1100 UTCof 15 August 2007 and (b) 0900 UTC of 31 March 2009180 A.J. Litta et al.The cyclonic rotation is visible in the figure during the tornado hour witha maximum wind speed of 12 ms�1. The model simulated surface wind at 0900UTC of 31 March 2009 is presented in Fig. 12.4b. The tornado vortex is clearlyvisible at the model predicted s between 20.5–21�N and 86–87�E with a maximumspeed of 20 ms�1. The model failed to capture the intensity of wind speed in boththe cases. However, the core of the strongest winds is shown very close to the site ofactual occurrence of the event.12.4.2 Instability Indices from the ModelAn attempt is made to examine different stability indices, which acts as indicatorsof severe convective activity. The NMM model simulated skew-t plot of Ludhianaat 1100 UTC on 15 August 2007 is illustrated in Fig. 12.5a and over Rajkanikaat 0900 UTC on 31 March 2009 in Fig. 12.5b. The skew-t plots show that in boththe cases the atmosphere was convectively unstable for an occurrence of severethunderstorms over Ludhiana and Rajkanika. Convective Available PotentialEnergy (CAPE) represents the amount of buoyant energy available to acceleratea parcel vertically and a CAPE value greater than 1500 J kg�1 is suggested by051015202530ab00Z 03Z 06Z 09Z 12Z 15Z 18Z 21Z 00ZTime (UTC)00Z 03Z 06Z 09Z 12Z 15Z 18Z 21Z 00ZTime (UTC)Accumulated rainfall (mm) OBSNMMOBSNMM051015202530354045Accumulated rainfall (mm)Fig. 12.3 (a and b) Theinter-comparison of (a)3-hourly NMM modelsimulated and TRMMaccumulated rainfall(mm) over Ludhiana validfor 15 August 2007 at 0000UTC to 16 August 2007at 0000 UTC (b) 3-hourlyNMM model simulatedand observed accumulatedrainfall (mm) over Chandbalistation valid from 0000 UTCof 31 March 2009 to 0000UTC of 1 April 200912 Simulation of Tornadoes over India Using WRF-NMM Model 18129.6N19N84E 84.5E 85E 86E 87E 88E 89E86.5E 87.5E 88.5E 89.5E85.5E19.5N20N20.5N21N21.5N22N22.5N23N74.4E 74.7E 75.3E 75.6E 75.9E 76.2E 76.5E1220surface wind (09 UTC)surface wind (11 UTC)76.8E75E29.8N30N30.2N30.4N30.6N30.8N31N31.2N31.4NabFig. 12.4 (a and b) The model simulated surface wind valid at (a) 1100 UTC of 15 August 2007(b) 0900 UTC of 31 March 2009182 A.J. Litta et al.10020030040050060070080090010001002003004005006007008009001000Mixing Ratio (g/kg)Mixing Ratio (g/kg)1 2 3 4 6 8 10 15 20 25 304035LCL4035LCL–40 –30 –20 –10 0 10Temperature (˚C)20 30 40–40 –30 –20 –10 0 10 20 30 40mb RH(%)25 50 75100mb RH(%)25 50 751001 2 3 4 6 8 10 15 20 25 30Temperature (˚C)abFig. 12.5 (a and b) The model simulated skew-t plot (a) over Ludhiana at 1100 UTC on 15August 2007 (b) over Rajkanika at 0900 UTC on 31 March 200912 Simulation of Tornadoes over India Using WRF-NMM Model 183Rasmussen and Wilhelmson (1983) as being necessary for supercells to form. Johnset al. (1993) and Korotky et al. (1993) extended the use of CAPE for tornadicenvironments to include high CAPE-low shear as well as low CAPE-high shearsituations. Table 12.2 shows the model simulated stability indices at both cases withcritical levels for severe thunderstorms.The NMM model simulated a high CAPEvalue which is 3,437 J kg�1 during the tornado hour for the first case and 3,584 J kg�1for the second tornado case (Table 12.2).The Lifted Index (LI) measures the difference between a parcel’s temperaturecompared with the environmental temperature at 500 hPa, after the parcel has beenlifted from the Lifting Condensation Level (Air Weather Service Technical Report1990). LI has proved useful for indicating the likelihood of severe thunderstorms.The chances of a severe thunderstorm are best when the LI is less than or equal to�6.This is because air rising in these situations is much warmer than its surroundingsand can accelerate rapidly and create tall and violent thunderstorms. Values lessthan �9 reflect intense instability. The NMM model is able to capture a low valuewhich is �8 at tornado hour of both cases (Table 12.2).Miller (1972) introduced the Total Totals Index (TTI) for identifying areas ofpotential thunderstorm development. It accounts for both static stability and thepresence of 850 hPa moisture. A TTI of greater than 48 indicates favorableconditions for development of severe thunderstorms; a value of 50 indicatesfavorable conditions for tornadoes (Air Weather Service Technical Report 1990).The NMM model simulated TTI value is 47 for the first case and 51 for the secondcase (Table 12.2). The CAPE and TTI show a high value and LI presented a lowvalue during tornado hour, which is a favorable condition for severe thunderstormoccurrence.12.5 ConclusionsThe model simulation studies of the tornado of 15 August 2007 over Ludhiana(Punjab) and the tornado of 31 March 2009 over Rajkanika (Orissa) lead to thefollowing broad conclusions:The model simulated meteorological parameters such as wind speed, pressurevertical velocity and moisture convergence at 850 hPa are consistent with eachTable 12.2 NMM model simulated stability indices at Tornado hour for both casesStability index Description Critical levelCase 1 Case 21100 UTC 0900 UTCLifted index T500 – Tparcel 48 47 51CAPEÐz¼LNBz¼LFCgyvpðzÞ�yvðzÞ½ �yvðzÞ dz > 1,500 3,437 3,584184 A.J. Litta et al.other and all are in good agreement with the observation in terms of the region ofoccurrence of the intense convective activity for both the tornado cases. The modelsimulated tornado hour over Ludhiana is exactly matching with the actual occur-rence. However, model simulated tornado occurred 2 h prior to the actual occur-rence of tornado over Rajkanika.The WRF-NMMmodel has well captured the vertical motion. It is found that thepressure vertical velocity is produced at proper location and time in the case oftornado over Ludhiana with a magnitude of �12 Pa s�1. The model simulated wellthe updraft with a magnitude of �40 Pa s�1 in the case of tornado over Rajkanikaduring the model simulated tornado hour. The wind speed is not in good agreementwith the observation for both the tornado cases; however, the core of the strongestwinds is shown to be very close to the site of actual occurrence of the event. Theintensity of 3-hourly rainfall rates are in good agreement with the observation forboth the tornado cases. In the first case model is able to simulate 15.96 mm against26.0 mm over Ludhiana and for the second case 35.44 mm against 41.0 mm overChandbali.The model simulated thermodynamic derivatives of stability indices such asCAPE, LI, and TTI indicate a deep instable layer around Ludhiana and Rajkanikafavorable for intense convective activity like tornado during model simulatedtornado hour.Thus the dynamic and thermo-dynamic properties of the atmosphere are wellsimulated by WRF-NMM for the occurrence of tornado over Ludhiana and alsoover Rajkanika, and agree reasonably well with the observations. The results ofthese analyses demonstrated the capability of high resolution WRF-NMMmodel insimulation of severe thunderstorm produced tornadoes.ReferencesAir weather service technical report 79/006 (1990) The use of the skew T, Log P diagram inanalysis and forecasting. Air weather service, Scott AFB, IllinoisAsnani GC (2005) Tropical meteorology (Revised Edition), vol II. Indian Institute of TropicalMeteorology, PuneFujita T (1981) Tornadoes and downbursts in the context of generalized planetary scale. J AtmosSci 38:1511–1534Goldar RN, Banerjee SK, Debnath GC (2001) Tornado in India and neighborhood and itspredictability. Regional Met. Centre, Alipore, Kolkata. Issued by Office of the AdditionalDirector General of Meteorology (Research), Meteorological Office of India MeteorologicalDeptartment, Pune, IndiaGoliger AM, Milford RV (1998) A review of worldwide occurrence of tornadoes. J Wind Eng IndAerodyn 74:111–121Grell GA, Devenyi D (2002) A generalized approach to parameterizing convection combiningensemble and data assimilation techniques. Geophys Res Lett 29(14):1693Janjic ZI (2003) A nonhydrostatic model based on a new approach. Meteorol Atmos Phys82:271–285Johns RH, Davies JM, Leftwich PW (1993) Some wind and instability parameters associated withstrong and violent tornadoes, 2. Variations in the combinations of wind and instability12 Simulation of Tornadoes over India Using WRF-NMM Model 185parameters. The Tornado: its structure, dynamics, prediction and hazards. Geophys MonogrAm Geophys Union 79:583–590Kain JS, Weiss SJ, Levit JJ, Baldwin ME, Bright DR (2006) Examination of convection-allowingconfigurations of the WRF model for the prediction of severe convective weather: The SPC/NSSL spring program 2004. Weather Forecast 21(2):167Korotky W, Przybylinski RW, Hart JA (1993) The Plainfield, Illinois, tornado of August 28, 1990:The evolution of synoptic and mesoscale environments. The Tornado: its structure, dynamics,prediction and hazards. Geophys Monogr Am Geophys Union 79:611–624Kumar S, Singh MS (1978) Satellite study of development of severe thunder/hailstorms over southPunjab on 10 March 1975. Indian J Met Hydrol Geophys 29(4):754–756Kumar S, Singh J, Raj H (1979) Severe weather outbreak associated with intersection ofsub-tropical jet stream with squall line. Vayu Mandal 9(1&2):46–47Litta AJ, Mohanty UC (2008) Simulation of a severe thunderstorm event during the field experi-ment of STORM programme 2006, using WRF-NMM model. Curr Sci 95(2):204–215May PT, Rajopadhyaya DK (1999) Vertical velocity characteristics of deep convection overDarwin. Aust Mon Weather Rev 127:1056–1071Miller RC (1972) Notes on analysis and severe storm forecasting procedures of the air force globalweather central AWS TR 200 (revised) air weather service scott air force base, IllinoisNoda AT, Niino H (2005) Genesis and structure of a major Tornado in a numerically-simulatedsupercell storm: importance of vertical vorticity in a gust front. SOLA 1:005–008Peterson RE, Mehta KC (1981) Climatology of tornadoes of India and Bangladesh. Arch MeteorolGeophys Bioklimat 29B:345–356Rasmussen RM, Wilhelmson RB (1983) Relationships between storm characteristics and1200GMT hodographs, low-level shear and stability. In: Preprints of 13th conference on severelocal storms. American Meteorological Society, Tulsa, pp 55–58Singh R (1981) On the occurrence of tornadoes in India. Mausam 32(3):307–314Vaidya SS (2007) Simulation of weather systems over Indian region using mesoscale models.Meteorol Atmos Phys 95:15–26186 A.J. Litta et al.Chapter 13A Pilot Study on the Energetics Aspectsof Stagnation in the Advance of SouthwestMonsoonSomenath Dutta and Lt. VishwarajashreeAbstract An attempt has been made to understand dynamically the stagnation inthe advancement of southwest monsoon from an energetics point of view. Recentfour cases of stagnation of duration more than 10 days during 2002–2006 have beenselected. For each case, different energy terms, their generation and conversionamongIndia, pksingh@ncmrwf.gov.inP.P. Singh Department of Agrometeorology, College of Agriculture, G.B. PantUniversity of Agriculture and Technology, U.S. Nagar, Pantnagar, Uttarakhand263145, IndiaContributors xxiiiVirendra Singh Natural Plant Products & Biodiversity Divisions, Institute ofHimalayan Bioresource Technology (CSIR), Palampur, HP 176061, India,vsgahlan@gmail.comM.V.K. Sivakumar World Meteorological Organization, Geneva, Switzerland,msivakumar@wmo.intRobert Stefanski World Meteorological Organization, Geneva, Switzerland,rstefanski@wmo.intM.V. Subrahmanyam South China Sea Institute of Oceanology, Chinese Academyof Science, Beijing, China, mvsm.au@gmail.comAmeenulla Syed State Climate Office of North Carolina, NC State University,Box 7236, Raleigh, NC 27695-7236, USA, asyed@ncsu.eduS. Tabasum Department of Agrometeorology, Indira Gandhi Krishi Vishwavidyalya,Raipur 492006 (C.G.), India, shabana.tbsm@gmail.comK.C. Tripathi K. Banerjee Centre of Atmospheric & Ocean Studies, Institute ofInterdisciplinary Studies, University of Allahabad, Allahabad 211 002, India,kctripathi@gmail.comS.K. Tripathi Department of Water Resources Development and Management,Indian Institute of Technology Roorkee, Roorkee, India, sankufwt@iitr.ernet.inAjit Tyagi India Meteorological Department, New Delhi 110 003, India, ajit.tyagi@gmail.comV.B. Vaidya Anand Agricultural University, Anand, Gujarat 388110, India,vaidya.vidyadhar@gmail.comS.S. Vanjari Department of Agronomy, Dr. Panjabrao Deshmukh Krishi Vidyapeeth,PO Krishi Nagar, Akola (MS) 444 104, India, wanjari.sanjay@rediffmail.comM.C. Varshneya Anand Agricultural University, Anand, Gujarat 388110, India,mcvarshneya@gmail.comAnanta Vashisth Division of Agricultural Physics, Indian Agricultural ResearchInstitute, New Delhi 110012, India, anantavashisth@iari.res.inB. Venkateshwarlu Central Research Institute for Dryland Agriculture, Hyderabad,AP 500 059, India, Vbandi_1953@yahoo.comLt. Vishwarajashree INS Goruda, Cochin, Indiaxxiv ContributorsChapter 1Modernization of Observation and ForecastingSystem in IMD in Support of Agromet ServicesAjit TyagiAbstract India Meteorological department has added many data and researchnetworks during the 135 years for climate-dependent sectors, such as agriculture,forestry, and hydrology, rendering a modern scientific background to atmosphericscience in India. The inclusion of the latest data from satellites and other modernobservation platforms, such as Automated Weather Stations (AWS), and ground-based remote-sensing techniques in recent years has strengthened India’s long-termstrategy of building up a self-reliant climate data bank for specific requirements,and also to fulfill international commitments of data exchange for weather fore-casting and allied research activities. It has augmented forecasting capabilities tomeet the operational requirements of day to day seamless weather forecasts invarious ranges.1.1 IntroductionAgriculture in India is the means of livelihood of almost two thirds of the workforce in the country and needs accurate weather forecasts to plan when to sowseeds, irrigate and fertilise their fields, and harvest their crops. Since most of thearea used for agricultural activity is rain fed, agricultural output is influenced byoverall seasonal rainfall as well as by intra-seasonal rainfall variations. The depen-dence of agriculture on weather information and forecasts is going to increase infuture on two counts. First is due to increase in extreme weather events and climatevariability caused by global warming and second is from increase in demand offood products from growing population having higher quality of life. In addition toA. Tyagi (*)India Meteorological Department, New Delhi 110003, Indiae-mail: ajit.tyagi@gmail.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_1, # Springer-Verlag Berlin Heidelberg 20111general forecast being provided by IMD to agriculture sector at macro scale, there isincreasing demand from horticulture, floriculture, and crop specific sectors forlocation specific forecasts on different time scales.In view of growing operational requirements from various user agencies, thereis a need for a seamless forecasting system covering short range to extended rangeand long range forecasts, particularly for the agricultural requirements. Suchforecasting system is to be based on hierarchy of Numerical Weather Prediction(NWP) models. For a tropical country like India where high impact mesoscaleconvective events are very common weather phenomena, it is necessary to havegood quality high density observations both in spatial and temporal scale to ingestinto assimilation cycle of a very high resolution non-hydrostatic mesoscalemodel. A major problem related to skill of NWP models in the tropics is due tosparse data over many parts of the country and near absence of data from oceanicregion.1.2 Augmentation of Observational SystemIn view of the importance of data in the tropical numerical weather prediction, IMDhas been in the process of implementing a massive modernization programme forupgrading and enhancing its observation system. In the first phase of modernisation550 additional AWS out of which about 125 will have extra agricultural sensors likesolar radiation, soil moisture and soil temperature and 1,350 Automatic Rain Gauge(ARG) stations will be installed in the current year. In addition to this, a network of55 Doppler Weather Radar has been planned of which 12 are to be commissioned inthe first phase. DWR with the help of algorithms can detect and diagnose weatherphenomena, which can be hazardous for agriculture, such as hail, downbursts andsquall. Normalized Difference Vegetation Index (NVDI) derived from the CCDpayload of presently available INSAT-3A satellite is useful for agriculture formonitoring the vegetation on a broad scale. NOAA/MODIS/Metop polar orbitingsatellite data receiving and processing system will be installed at New Delhi,Chennai and Guwahati. This will enable availability of real time products fromthese satellites for use in forecasting by conventional means and by assimilating inNWP models in turn improving the Agro meteorological forecasts also. A newsatellite INSAT-3D is scheduled to be launched during third quarter of the year.INSAT-3D will usher a quantum improvement in satellite derived data from multispectral high resolution imagers and vertical sounder. In addition to above, IMD isalso planning to install wind profilers and radiometer to get upper wind andtemperature data. Data from AWSs, ARGs, DWRs, INSAT-3D, NOAA/MODIS/Metop and wind profilers will available in real time for assimilation in NWPmodels. A High Power Computing (HPC) system with 300 terabyte storage isbeing installed at NWP Centre at Mausam Bhawan. It will greatly enhance ourcapability to run global and regional models and produce indigenous forecastproducts in different time scales.2 A. Tyagi1.3 Forecasting SystemCurrently, IMD runs a number of regional NWP models in the operational mode.IMD also makes use of NWP global model forecast products of other operationalcentres, like NCMRWF T-254, ECMWF, JMA, NCEP and UKMO to meet theoperational requirements of day to day weather forecasts.Recently, IMD implemented a multimodel ensemble (MME) based district level5 days quantitative forecast system as required for the Integrated Agro-advisoryService of India. The technique makes use of model outputs of state of the art globalmodels from the five leading global NWP centres (Krishnamurti et al. 1999). ThePre-assigned grid point weights on the basis of anomaly correlation coefficients(CC) between the observed values and forecast values are determined for eachconstituent model at the resolution of 0.25� � 0.25�different terms have been computed during the stagnation period and alsoduring the pre-stagnation pentad over a limited region between 65�E and 90�E, 5�Nand 30�N. These computations are based on NCEP 2.5� � 2.5� re-analysed dailycomposite data during different stagnation period and during corresponding pre-stagnation pentad.From this study it is found that:• In most of the cases there is a reduction in zonal kinetic energy (Kz) and in eddykinetic energy (KE) during stagnation period compared to pre-stagnation pentad.• In all cases, during stagnation as well as in the pre stagnation pentad, barotropiceddy energy transfer dominates over baroclinic eddy energy transfer.• In most of the cases there is a reduction in the generation of zonal availablepotential energy [G (Az)] during stagnation period compared to pre-stagnationpentad.• In most of the cases there is a reduction in the conversion from zonal availablepotential energy to zonal kinetic energy [C (Az, Kz)] during stagnation periodcompared to pre-stagnation pentad.S. Dutta (*)Meteorological Office, Pune, Indiae-mail: dutta.drsomenath@gmail.comLt. VishwarajashreeINS Garuda, Cochin, IndiaS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_13, # Springer-Verlag Berlin Heidelberg 201118713.1 IntroductionThe normal onset date of southwest monsoon over Kerala is 1st June and the date ofcovering the entire country is 15th July. But there are wide variations to this traveltime of monsoon in different years. Whereas the onset of Indian Summer monsoonover Kerala is of importance, the advance of monsoon over other parts of thecountry is also equally important. The advance of monsoon is not a continuousprocess rather it is pulsatory in nature. There are several years when stagnationoccurs and the northward or westward propagation of monsoon is arrested. Most ofthe years there are only short lulls of 7–10 days, but there are years when thisstalling of the stagnation of monsoon continues for over 2 weeks period. Suchstagnation causes concerns to the farmers and other weather dependent activities.Keshavamurty and Awade (1970) found that maintenance of mean monsoontrough against frictional dissipation is mainly due to work done by horizontalpressure gradient. Their study also indicates a loss in standing eddy kinetic energyby rising of relatively colder air and sinking motion of relatively warmer air.Rao and Rajamani (1972) studied the heat source and sinks and generationof available potential energy of the atmosphere over the Indian region duringsouthwest monsoon season. Their computation showed a net generation of APEover the region of study.Krishnamurty and Ramanathan (1982) have shown that a sharp rise in therotational kinetic energy is an interesting aspect of onset of Indian SummerMonsoon (ISM). Awade and Bawiskar (1982) have shown that bad monsoonactivity is associated with large divergence of heat at sub-tropics and large conver-gence of heat at extra tropics. According to Pasch (1983) the onset of planetaryscale monsoon is preceded by an organization of cumulus convection on theplanetary scale. Awade et al. (1985) have shown that in good monsoon yearsthere is large divergence of momentum in subtropics, while there is large conver-gence of momentum in mid latitude. They argued that this situation leads to astronger westerly in mid-latitude and stronger easterly at tropics. Krishnamurty(1985) has shown that divergent kinetic energy, must be transferred to rotationalkinetic energy, available potential energy must be transferred to divergent kineticenergy via rising motion over warm region/sinking motion over cold region. He hasalso shown that available potential energy is maintained via heating of warmer airand cooling of colder air. Rajamani (1985a) computed the diabatic heating andgeneration of APE over south Asia for typical monsoon month July 1963. The studyhas inferred positive generation of both zonal and standing eddy APE. Rajamani(1985b) made a study on available potential energy (APE) and its transformationinto kinetic energy. This study shows that differential heating between Asianlandmass and Indian ocean causes the generation of zonal APE (Az), a part ofwhich is converted into zonal kinetic energy (Kz). The study also indicates thatdiabatic heating generates standing eddy APE (AE), which is again converted intostanding eddy kinetic energy. Ramasastry et al. (1986) brought out some features ofthe process of stagnation. Krishnamurti and Surgi (1987) have shown that around188 S. Dutta and Lt. Vishwarajashreethe period of the onset of monsoon rains over India, there is a sharp rise in theconversion of zonal available potential energy (Az) to zonal Kinetic energy (Kz).Yanai et al. (1992) have shown that reversal of north-south temperature gradientin the layer between 700 and 200 hPa triggers the onset of South Asian monsoon.George and Mishra (1993) had examined the temporal variations of the zonal andeddy kinetic and available potential energy in association with the formation,growth and maintenance of vortex during southwest monsoon. Their studyindicated that barotropic eddy energy transfer dominates over baroclinic eddyenergy transfer. They have also showed that C (Kz, KE) > C (AE, KE). Biswaset al. (1998) have studied the role of the mechanical barrier of the Himalayanmassif –Tibetan plateau and the mid tropospheric sub-tropical ridge in the stagnation in theadvance of southwest monsoon.Krishnamurti et al. (1998) studied the energetics of south Asian monsoon. UsingFSU Gobal spectral model at T 170 resolution, they examined the maintenance ofthe monsoon. This study indicates that differential heating leads to the growth ofAPE, which is next passed on to the divergent motions and then finally divergentK.E. is converted to rotational K.E, which of course critically depends on theorientation of the isopleths of c and w. Results of the study by Wu and Zhang(1998) are in conformity with that of Yanai et al. (1992). These studies indicate thatduring the onset of South Asian monsoon there is a sudden increase in the zonalavailable potential energy.Raju et al. (2005) studied the onset characteristics of the southwest monsoonover India. Their study reveals that the low level kinetic energy, verticallyintegrated generation of kinetic energy and net tropospheric moisture can be usedas potential predictors to predict the onset of Southwest monsoon. Rao andMohanty (2007) have shown that the onset of the Indian southwest monsoon overthe Bay of Bengal is discernible by a gradual increase in the adiabatic generation ofkinetic energy, while over the Arabian Sea it is first noticeable by a steep and abruptincrease of generation.From the foregoing discussion it appears that hardly there is any study on theenergetic aspects of stagnation, although there are ample studies on energeticsduring onset and energetics on monsoon disturbances.The present study aims at analyzing the energetic aspects of stagnation ofsouthwest monsoon.13.2 DataThe different cases of stagnation period in the advance of SW monsoon havebeen obtained from the isochrones of monsoon advance prepared by the IndiaMeteorological Department. For the present study we have used data for u, v, o,T, rh obtained from NCEP/NCAR. We have used daily composite data of the abovefields for the stagnation period and for pre-stagnation pentad of individual casesover the limited region between 5�N and 30�N, 65�E and 90�E.13 A Pilot Study on the Energetics Aspects of Stagnation 18913.3 MethodologyFirst, from the temperature data, at each grid point, heating rate_QCphas beencomputed using first law of thermodynamics_QCp¼ dTdt � aCpo. In the computationof dTdt , tendency has not been taken care of. Then, following Krishnamurtyand the multimodel ensembleforecasts (day 1 to day 5 forecasts) are generated at the same resolution on a real-time basis. The ensemble forecast fields are then used to prepare forecasts for eachdistrict taking the average value of all grid points falling in a particular district. Aninter-comparison of spatial co-relation co-efficient (CC) between observed andforecasts rainfall on the basis of the MME technique and the member models isALL INDIA spatail CC (MONSOON 2009)0.20.250.30.350.40.45DAY-1 DAY-2 DAY-3 DAY-4 DAY-5CCGFSECMFJMAT254UKMOMMEENSMFig. 1.1 An inter-comparison of country mean spatial CC of day 1 to day 5 forecasts of rainfall byNCEP, ECMWF, JMA, NCMRWF, UKMO, mean ensemble andMME for summer monsoon 20091 Modernization of Observation and Forecasting System 3illustrated in Fig. 1.1. The results show that MME is superior to each membermodel at all the forecasts (day 1 to day 5).In order to evaluate the performance of district level forecasts, skill score –Probability of Detection (POD) is considered. POD is defined as:POD ¼ HHþMWhere H indicates hits and M for missing events for the following rainfallcategories:1. Rain or no rain2. Light Rain: 0–10 mm3. Moderate Rain: >10 mm and 65 mmState-wise performance of district level rainfall forecasts for day 1 and day 5forecasts for some selected states like Orissa, Rajasthan, Maharashtra, Gujarat andKerala, which represent east central India – the domain of monsoon low; northwestIndia – region of less monsoon rainfall; west India; region of mid-tropospherecirculation and extreme south east Peninsula are illustrated in Figs. 1.2 and 1.3.For the day 5 forecasts, results of Madhya Pradesh (central India) is also includedDAY-1 FCST : MON 200900.10.20.30.40.50.60.70.80.91NO RAIN LIGHT RAIN MOD RAIN HEAVY RAINPODORISSARAJASTHANMAHARASTRAKERALAGUJARATFig. 1.2 State-wise performance of district level day 1 forecasts for some selective states4 A. Tyagi(Roy Bhowmik et al. 2009; Roy Bhowmik and Durai 2010). The results show thatperformance skill of forecast of the district level rainfall for the rainfall amount ofmoderate range is reasonably good for all these states, where POD is more than 0.4.District-wise performance of day 1 to day 5 rainfall forecasts for the districts ofOrissa is shown in Fig. 1.4. For Orissa, POD for rain/no rain case has been above0.8 at all the districts, for light rain it is around 0.6, for moderate rainfall it isbetween 0.3 and 0.4 and for heavy rainfall it is close to 0.A dynamical statistical technique is developed and implemented for the real-time cyclone genesis and intensity prediction. Numbers of experiments are carriedout for the processing of DWR observations to use in nowcasting and mesoscaleapplications. The procedure is expected to be available in operational mode soon.Impact of INSAT CMV in the NWP models has been reported in various studies.Various multi-institutional collaborative forecast demonstration projects such as,Dedicated Weather Channel, Weather Forecast for Commonwealth Games 2010,Land falling Cyclone, Fog Prediction etc. are initiated to strengthen the forecastingcapabilities of IMD.With the availability of new observations and infrastructure from the moderni-zation programme of IMD, future Weather Forecasting System of IMD would be asbriefly given below:DAY-5 :MON-200900.10.20.30.40.50.60.70.80.91NO RAIN LIGHT RAIN MOD RAIN HEAVY RAINPODORISSARAJASTHANMAHARASTRAKERALAGUJARATMADHYA PRADESHFig. 1.3 State-wise performance of district level day 5 forecasts for some selective states1 Modernization of Observation and Forecasting System 51.3.1 Now-Casting and Mesoscale Forecasting System(Valid for Half Hour to 24 h)• Processing of Doppler Weather Radar (DWR) observations at a central location(NHAC) to generate 3 D mosaic and other graphics products for nowcastingapplications.• Enhancing mesoscale forecasting capability of local severe weather by pro-viding 3 hourly area specific rainfall and wind forecasts (up to 24 h) at theresolution of 3 km from ARPS with the assimilation (hourly intermittent cycle)of DWR, AWS, wind profilers and other conventional and non-conventionalobservations.• Implementation of dynamical Fog prediction model for visibility forecasting atthe major airports of India.• Implementation and customization of the Delhi PP model (based on the UKPPmodel of UKMO) for the Delhi Commonwealth Games, 2010. This model willgive forecasts of precipitation, fog, cloud cover and visibility at 15 min resolu-tion up to 6 h ahead.India Meteorological Department (IMD) installed four S-Band DopplerWeather Radars (DWR) manufactured by GEMATRONIK GmbH (Model:METEOR 1500S) at Chennai (2002), Kolkata (2003), Machilipatnam (2004) andVisakhapatnam (2006), replacing the old generation S – Band cyclone detectionradars at these stations. Very recently two more S-Band DWRs (manufactured byBeijing METSTAR) have been installed at Delhi (Palam) and at Hyderabad.ORISSA (POD) DAY-500.20.40.60.811.2MAYURBHANJSUNDERGARHSAMBALPURANGULJAJPURBHADRAKJAGATSINGPURKHURDANAYAGARHSONEPURBOLANGIRKALAHANDIGANJAMRAYAGADAMALKANGIRIDistrictsPODNO RAINLIGHT RAINMOD RAINHEAVY RAINFig. 1.4 District-wise performance of day 5 forecasts for the Orissa state6 A. TyagiIn addition to the current deployment of DWRs, there are plans to install more suchradars for use in severe storm forecasting and airport weather warning. Radarsprovide detailed information of reflectivity, radial velocity and spectrum width athigh spatial and temporal resolution. Recently IMD has initiated research intoprocessing of Indian DWR data for nowcasting and meso-scale modeling. TheNowcasting Application software – “Warning Decision Support System –Integrated Information (WDSS-II)” is used for real-time analyzing and visualizingremotely sensed weather radar data. The quality controlled radar data is assimilatedinto the Advanced Regional Prediction System (ARPS) model.A reflectivity field snapshot of the tropical cyclone Khaimuk at 5 km height at1,410 UTC of 14 November 2008 created from data of the three radars at Chennai,Machilipatnam and Visakhapatnam, when the cyclone was located close toTamilnadu coast is depicted in Fig. 1.5.The ARPS model (at 9 km horizontal resolution) simulated reflectivity fieldsby various experiments against the observed reflectivity valid at 0600 UTC of 27November 2008 are presented in Figs. 1.6a–c. Three main cells are detected in eachexperiment and these are marked as A, B and C respectively. An inter-comparisonof these results reveal that the cells and the spiral rain band structure is clearly seenin the experimental run (with DWR data) and the pattern is close to observedreflectivity as produced by DWR Chennai in the panel (c). The correspondingmaximum intensity of these three cells in experiment (with DWR data) is foundto be well matched with the observed one as shown in panel (c). In the experiment(with DWR data), all the three cells and rain band structure appeared similar to theobserved pattern.Fig. 1.5 Snapshot of the mosaic reflectivity field at 4.0 km during the tropical cyclone Khaimukhof at 14 November 2008, which was tracked by the three Doppler radars at Chennai,Machilipatnam and Visakhapatnam1 Modernization of Observation and Forecasting System 7Fig. 1.6 (a–c) Inter-comparison of reflectivity fields (dBZ) of various simulation experimentsagainst the observed field valid at 0600 UTC of 27 November 2008: reflectivity by DWRexperiment (using both radial wind and reflectivity), (b) Experiment without DWR observationsand (c) Observed reflectivity8 A. Tyagi1.3.2 Regional Models for Short Range Forecasting System(Valid Up to 3 days)• Seventy-two hours forecasts from WRF model with three nested domains (at theresolution of 27, 9 and 3 km). The nested model at the 3 km resolution is beingoperated at the Regional/State Met Centres at 12 h interval with 3 DVAR dataassimilation.• Seventy-two hours forecasts from MM5 model.• For Cyclone Track Prediction, 72 h forecast from Quasi Lagrangian Model(QLM) at 40 km resolution at 6 h interval; WRF (NMM) at 9 km resolutionwith assimilation package of Grid Statistical Interpolation (GSI).• For Cyclone track and intensity prediction: multimodel ensemble technique andapplication of dynamical statistical approach for 72 h forecasts; forecast wouldbe updated at 12 h interval.• Development of multimodel ensemble technique for probabilistic forecasts ofdistrict level heavy rainfall events.1.3.3 Global Model for Medium Range Forecasting(Valid Up to 7 days)• NCEP Decoding System.• Global Data Assimilation System (GDAS), 6 hourly cycles with GSI (GridStatistical Interpolation).• Global Forecast System (GFS) T-382/L64.The regional and global forecast products are routinely made available in the IMDweb site: www.imd.gov.in1.3.4 Extended Range Forecast for Rainfall and Temperature• Statistical downscaling studies using historical GCM data sets (for predictorvariables) and IMD’s observational data sets of rainfall and temperature aspredictands.• Customization of RCMs for dynamical downscaling studies.• Multi-model super ensemble approach: based on dynamical and statisticalpredictions for rainfall and temperature of monthly and seasonal time scales.• Development of probabilistic forecast models for categorical forecasts of rainfalland temperature in terms of monthly and seasonal time scales for meteorologi-cally homogeneous zones/major agro-climatic zones.• To implement a dynamical model in conjunction with the statistical model.1 Modernization of Observation and Forecasting System 91.3.5 Long Range Forecast for Summer and North-East MonsoonShort-range forecasting (WRF) strategy at the New Delhi is presented in Fig. 1.7.Mesoscale data assimilation and generation of high-resolution analysis for RSMC(Regional Specialized Meteorological Centre, Delhi as recognized by WMO)domain at the horizontal grid-spacing of 27 km and vertical resolution 50 Etalevels. Time interval would be 6 hourly cycling. Resolution of the forecast fieldswould be 27 km for RSMC and 9 km for Indian domain. Triple nested (27, 9 and3 km) model forecast would be made for specific events or expeditions (e.g.Commonwealth Game 2010). Fig. 1.8 depicts the short-range forecasting (WRF)strategy at the State Meteorological Centres.1.4 Agromet Advisory ServicesIMD started issuing district level weather forecast since 1st June 2008. Districtlevel agromet advisory meant for farmers are made based on the quantitativedistrict level (612 districts) weather forecast for seven weather parameters viz.,45N40N35N30N25N20N15N10N5NEQ5S10S15S20S40E 45E 50E 55E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105E 110E 115E 120EDeMuChKoFig. 1.7 Short-range forecasting (WRF) strategy at IMD HQ10 A. Tyagirainfall, maximum and minimum temperatures, wind speed and direction, relativehumidity and cloudiness up to 5 days. NWP products for the district level forecastgenerated from NHAC, New Delhi using multimodel ensemble is disseminated toRegional Meteorological Centres and Meteorological Centres of IMD located indifferent states. These offices undertake value addition to these products andcommunicate to 130 AgroMet Field Units (AMFUs) located at State AgricultureUniversities (SAUs), institutes of Indian Council of Agriculture Research (ICAR)etc who prepare Agro-meteorological advisories for State and Central agriculturalauthorities as well to individual farmers based on the district specific meteorologi-cal information containing observed data, reported phenomena, specially derivedweather forecasts for small domains, coupled with information on crop stage, stateand diseases. The agro- meteorological advisories prepared by the AMFUs arecommunicated back to respective Regional Meteorological Centres and Meteoro-logical Centres of IMD to prepare composite bulletin for the state. The dissemina-tion of advisories is made through various modes of mass media (including AllIndia Radio (AIR) and Doordarshan, Private TV and radio channels, news papers,Internet), extension centres of agriculture and rural development related agencies(NGOs, Kisan Call Centres/ICAR and other related Institutes/AgriculturalUniversities/KVKs etc.) and directly to different Government offices. The purposeor such advisories is to enable farmers to take remedial measures and preventweather induced losses and to mobilize resources at more organized levels tocounter detrimental conditions (Rathore and Maini 2008; Rathore et al. 2009).Fig. 1.8 Short-range forecasting (WRF) strategy at state met centre1 Modernization of Observation and Forecasting System 11ReferencesKrishnamurti TN, Kishtawal CM, Larow T, Bachiochi D, Zhang Z, Willford EC, Gadgil S,Surendran S (1999) Improved weather and seasonal climate forecasts from multimodel superensemble. Science 285:1548–1550Roy Bhowmik SK, Durai VR, Das Ananda K, Mukhopadhaya B (2009) Performance of IMDMulti-model ensemble based district level forecast system during summer monsoon 2008,IMD Met Monograph No. 8/2009Roy Bhowmik SK, Durai VR (2010) Application of multi-model ensemble technique for real-timedistrict level forecasts over Indian region in short range time scale. Meteorl Atmos Phy106:19–35Rathore LS, Parvinder Maini (2008) Economic Impact Assessment of Agro-MeteorologicalAdvisory Service of NCMRWF. Report no. NMRF/PR/01/2008, 104pp, Published byNCMRWF, Ministry of Earth Sciences, Government of India, A-50 Institutional Area, Sector-62, NOIDA, UP, INDIA 201 307Rathore LS, Singh KK, Baxla AK (2009) Agromet services for ePachayats. Information forDevelopment VII(4):14–1612 A. TyagiChapter 2Monthly and Seasonal Indian SummerMonsoon Simulated by RegCM3 at HighResolutionsS.K. Dash, Savita Rai, U.C. Mohanty, and S.K. PandaAbstract The purpose of this study is to examine the advantages of using higherresolution regional model in simulating the temperature and precipitation overIndia. The Regional Climate Model version 3 (RegCM3) has been integrated tosimulate the Indian summer monsoon rainfall for a number of years at two horizon-tal resolutions 55 and 30 km. The characteristics of interannual variations in thecontrasting monsoon years 2002 and 2003 have been examined in details at theseresolutions. Comparison shows that the model simulated area weighted averagemagnitudes and spatial distribution of rainfall over India during June to Septembermonths reasonably compare with the respective gridded rainfall values of IndiaMeteorological Department (IMD). Model simulated rainfall values with 30 kmresolution are closer to the IMD values as compared to the simulated precipitationat 55 km model resolution. Also the spatial distribution of rainfall in RegCM3 with30 km is more realistic than that of 55 km resolution. Comparison with NCEP/NCAR analysed fields shows that RegCM3 with 30 km resolution performs betterthan that with 55 km resolution in simulating the upper and lower level winds overIndia. It may be noted that both temperature and rainfall are important weatherparameters for the farmers in terms of agricultural productions. Dynamical down-scaling of the high resolution model simulated weather parameters will eventuallyhelp in agricultural risk management.S.K. Dash (*) • S. Rai • U.C. Mohanty • S.K. PandaCentre for Atmospheric Sciences, IIT Delhi, Hauz Khas, New Delhi, Indiae-mail: skdash@cas.iitd.ernet.in; savita1559@gmail.com; mohanty@cas.iitd.ernet.in;sampadpanda@gmail.comS.D. Attri et al. (eds.), Challenges and Opportunities in Agrometeorology,DOI 10.1007/978-3-642-19360-6_2, # Springer-Verlag Berlin Heidelberg 2011132.1 IntroductionThe spatial variability of monsoon rainfall is observed to be large and there wereseveral occasions when some parts of the country received heavy rainfall while atthe same time some other parts had serious rainfall deficiency. In the recent past,the monsoons of 2002 and 2003 have exhibited very contrasting characteristics sofar as rainfall over India is concerned. Large deficient rain in July 2002 over mostof the country was reported in the year 2002. On the other hand, 2003 has beenreported as a normal monsoon year. In 2002, the seasonal rainfall over thecountry as a whole was 81% of its long period average and hence it was declaredas an All India drought monsoon year (Kalsi et al. 2004). The most importantaspect of 2002 monsoon was that the rainfall in July was the lowest in the past102 years. Month wise, the rainfall was normal in June, extremely subdued inJuly, normal in August and nearly normal in September. On the other hand in theyear 2003, total seasonal monsoon rainfall over the country as a whole was 102%of its long term average. The rainfall activity in 2003 was very good during themonths of June and July, though August and September were little subdued.Rajeevan et al. (2004) using 8-parameter and 10-parameter power regressionmodels could not predict the large deficiency of rainfall of July 2002. Based onthe model forecasts of European Centre for Medium range Weather Forecasts(ECMWF), Gadgil et al. (2002) summarised that June, July and August rainfallwas deficient only over the southwestern peninsular and near normal over rest ofthe country.It has been demonstrated that for examining the weather features in greaterdetail, regional models are more suitable than the global models. Today compu-tationally it is affordable to increase the resolution of regional models so as toresolve regional climatic features reasonably well. There have been someattempts in the past to simulate monsoon features and extreme weather eventsover India by regional models. Bhaskaran et al. (1996) simulated the Indiansummer monsoon using a regional climate model with a horizontal resolutionof 50 km nested with global atmospheric GCM. Their study showed that regionalmodel derived precipitation is larger by 20% than GCM. Ji and Vernekar (1997)simulated the summer monsoons of 1987 and 1988 by using the National Centersfor Environmental Prediction (NCEP) Eta model nested in the Center for Ocean-Land-Atmosphere (COLA) GCM. Their comparative studies showed that for1987, the Eta model simulates deficient summer monsoon rainfall over northernand peninsular India and the Indonesian region and excess rainfall over southeastChina, Burma and the sub-Himalayan region compared to 1988. Azadi et al.(2001) used MM5 to simulate western disturbances during January 1997 over theIndian region and to predict precipitation associated with it. Bhaskar Rao et al.(2004) simulated many observed features of the Indian summer monsoon such assea level pressure, 925 hPa temperature, low level wind and precipitation usingMM5. The Regional Climate Model version 3 (RegCM3) has also been success-fully integrated to simulate the salient features of Indian summer monsoon14 S.K. Dash et al.circulation and rainfall (Dash et al. 2006; Shekhar and Dash 2005). They foundthat the RegCM3 successfully simulate the 850 hPa westerly flow and the200 hPa easterly flow. Also the seasonal mean summer monsoon rainfall sim-ulated by the model is close to the GPCC values. They have also found thatRegCM3 successfully simulates the temperature pattern at 500 hPa over theIndian Peninsula and Tibet. In their study it was inferred that RegCM3 can beeffectively used to study monsoon processes over the South Asia region. Shekharand Dash (2005) have examined the effect of Tibetan snowfall in the month ofApril on the Indian summer monsoon circulation and associated rainfall usingRegCM3 with 55 km resolution. Model simulations show that when 10 cm ofsnow-depth in April is prescribed over Tibet, summer monsoon rainfall in entireIndia reduces by about 30%. Singh et al. (2006) used RegCM3 over East Asiaregions and showed promising performance of this regional model in simulatingimportant characteristics of monsoon circulations.The objective of the current paper is to examine the performance of RegCM3 attwo different resolutions 55 and 30 km so as to compare the important features ofsimulated summer monsoon in the contrasting years 2002 and 2003. A briefdescription of RegCM3, experimental design and data used are given in Sect. 2.2.Sections 2.3–2.6 discuss the characteristics of upper level wind, lower level wind,temperature at 500 hPa and rainfall for both the resolutions of RegCM3 respectivelyand the conclusions are summarized in Sect. 2.7.2.2 Model, Initial Data and Experimental DesignRegCM3, an upgraded version of the Abdus Salam International Centre forTheoretical Physics (ICTP) regional climate model RegCM2 originally developedby Giorgi et al. (1993a, b), is a compressible, grid point model with 14 verticallayers and hydrostatic balance. There are two categories of landuse such as MM4vegetation and Global Land Cover Characterization (GLCC) which determinesurface properties like albedo, roughness, moisture etc. at each grid point. MM4vegetation has 13 different types and GLCC has similarly 20 types of lands. Themodel dynamical core is essentially the same as that of the hydrostatic version ofMM5 (Grell et al. 1994). The model includes cumulus parameterization schemes,large scale precipitation scheme, planetary boundary layer (PBL) parameterization,state-of-the-art surface vegetation/soil hydrology package, the Biosphere-Atmo-sphere Transfer Scheme (BATS), Ocean flux parameterization, pressure gradientscheme, explicit moisture scheme, the radiative transfer scheme and the ocean-atmosphere flux scheme.The complete RegCM3 modelling system consists of four modules: Terrain,ICBC, RegCM and Postprocessor. Terrain and ICBC are the two components ofRegCM preprocessor. These program modules are run in sequence. Following Dashet al. (2006) in this study, Grell convective precipitation scheme with Arakawa-Schubert (AS) closure has been used. It has been widely used within both MM5 and2 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 15RegCMmodeling frameworks. It is a mass flux scheme that includes the moisteningand heating effects of penetrative updrafts and corresponding downdrafts.For simulating the monsoon circulation features the central grid point for modeldomain is chosen at 80�E and 20�N. In the 55 km resolution domain chosen is 50�Eto 109�E and 4�S to 41�N and there are 101 grid points along a latitude circle and120 points along the longitudinal direction. In case of 30 km resolution there is one-way nesting with a larger domain at 90 km resolution as shown in Fig. 2.1. In the90 km resolution the domain chosen is 51�E to 109�E and 4�S to 42�N and there are64 grid points along a latitude circle and 72 points along the longitudinal direction.While in case of 30 km resolution the domain chosen is 66�E to 98�E and 6�N to36�N and there are 120 grid points along both the latitude circle and the longitudinaldirection.The elevation data used are obtained from the United States Geological Survey(USGS). Terrain heights and land use data are used at 30 min resolution. In both theyears 2002 and 2003, simulations at 55 and 30 km resolutions cover the months40N35N30N25N20N15N10N5NEQ55E 60E 65E 70E 75E 80E 85E 90E 95E 100E 105EabFig. 2.1 One way nesting of domains at (a) 90 km and (b) 30 km for the Indian region16 S.K. Dash et al.from April to September in the ensemble modewith nine-members starting from25th April to 3rd May. The data for initial and lateral boundary conditions havebeen obtained from the National Centers for Environmental Prediction (NCEP)Reanalysis (NNRP1) and Sea Surface Temperature (SST) that are collected fromthe National Oceanic and Atmospheric Administration (NOAA) 4-times daily andweekly datasets respectively.Fig. 2.2 Mean winds (JJAS) at 850 hPa in 2002 (a) NCEP/NCAR reanalysis, (b) RegCM3 at55 km and (c) RegCM3 at 30 km2 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 172.3 Lower Level WindThe averages of seasonal mean winds of the nine ensemble members simulatedby RegCM3 in 2002 and 2003 with 55 and 30 km resolutions at 850 hPa levelsare depicted in Figs. 2.2 and 2.3 respectively. In the year 2002, the maximumFig. 2.3 Mean winds (JJAS) at 850 hPa in 2003 (a) NCEP/NCAR reanalysis, (b) RegCM3 at55 km and (c) RegCM3 at 30 km18 S.K. Dash et al.strength of the Somali jet at 850 hPa is 12 m/s at 30 km resolution which is closerto NCEP/NCAR reanalysis while in case of 55 km the difference is more. Acrossthe monsoon trough the maximum values are 8 m/s and 4 m/s in case of 30 and55 km respectively. The westerly wind over the peninsula has the speeds of10–12 m/s, 4–6 m/s and 8–10 m/s in case of 30 km model resolution, 55 kmmodel resolution and NCEP-NCAR respectively. Similar pattern is seen for theyear 2003. The maximum strength of the Somali Jet at 850 hPa level in case of30 km resolution is 12 m/s which is very close to the NCEP/NCAR reanalysisvalue of 14 m/s. While in 55 km resolution the corresponding wind strength is6 m/s, along the west coast the wind speeds are 12 m/s, 4 m/s and 10 m/s forresolutions 30, 55 km of RegCM3 and NCEP/NCAR respectively. The westerlywind simulated by RegCM3 over peninsula has the same speed of 10–12 m/s forboth the years 2002 and 2003.2.4 Upper Level WindThe observed NCEP/NCAR reanalysed wind and the ensemble averages of theseasonal mean winds simulated by RegCM3 at 200 hPa level in the years 2002 and2003 are shown in Figs. 2.4 and 2.5 respectively. In the year 2002maximum strengthsof the easterly jet across the peninsula are 22 m/s and 18 m/s at 30 and 55 km modelresolutions respectively. Both the values are close to the NCEP/NCAR reanalysisvalue of 20m/s. Upper level easterly wind strength over Tibet is 20m/s at both 30 and55 km model resolutions. In 2003, the wind speed ranges over the peninsula are20–22 m/s, 18–20 m/s and 16–18 m/s in case of 30 km model resolution, NCEP/NCAR reanalysis and 55 km respectively. Upper level wind strength over Tibet is24 m/s in the 30 km model resolution whereas both in NCEP/NCAR reanalysis and55 km resolution the corresponding strength is 20 m/s. At 30�N and 85�E, the windstrength is 12 m/s at both 30 and 55 km model resolutions.2 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 19Fig. 2.4 Mean winds (JJAS) at 200 hPa in 2002 (a) NCEP/NCAR reanalysis, (b) RegCM3 at55 km and (c) RegCM3 at 30 km20 S.K. Dash et al.2.5 Temperature at 500 hPaDepict the NCEP/NCAR reanalysis and RegCM3 simulated JJASmean temperaturesat 500 hPa with 55 and 30 km resolutions respectively for the years 2002 and 2003.As shown in Fig. 2.6 the mean maximum temperature over Tibet is 272 K at both 30Fig. 2.5 Mean winds (JJAS) at 200 hPa in 2003 (a) NCEP/NCAR reanalysis (b) RegCM3 at55 km (c) RegCM3 at 30 km2 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 21and 55 kmmodel resolutions in 2002. The correspondingmean temperature at 500 hPain the NCEP/NCAR reanalysis is 271 K. Over the Indian Peninsula the meansimulated temperature in the 55 km resolution is similar to that of NCEP/NCAR butit is more in case of 30 km model resolution for both the years 2002 and 2003.The corresponding value in NCEP/NCAR reanalysis is 272K Figures 2.6 and 2.7.Fig. 2.6 Mean temperatures (JJAS) at 500 hPa in 2002 (a) NCEP/NCAR reanalysis, (b) RegCM3at 55 km and (c) RegCM3 at 30 km22 S.K. Dash et al.2.6 RainfallThe monthly and seasonal rainfall simulated in two consecutive years 2002 and2003 with two different model resolutions of 55 and 30 km are discussed here.For better and close comparison of rainfall simulated by the model and the IMDobserved rainfall, area weighted averaging has been done at the model grids. ThisFig. 2.7 Mean temperatures (JJAS) at 500 hPa in 2003 (a) NCEP/NCAR reanalysis, (b) RegCM3at 55 km and (c) RegCM3 at 30 km2 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 23method is similar to that used by IMD with the observed sub-divisional rainfall. It isknown that IMD calculates ISMR based on the area weighted average sub divi-sional rainfall. As mentioned earlier RegCM3 at 55 km has 118 � 99 grids acrossthe Indian landmass and at 30 km it has a total of 118 � 118 grids. The grid boxesmarked ‘1’ in Fig. 2.8a, b are those, which cover all the meteorological subdivisionsof IMD as closely as possible. In this study, the model simulated monthly andseasonal rainfall are computed at the grids over Indian landmass. Figures 2.9 and2.10 show the grid averaged rainfall of IMD and those simulated by RegCM3 with55 and 30 km resolutions in July of both the years 2002 and 2003 respectively.Figures 2.11 and 2.12 show corresponding values for JJAS.IMD has generated monthly mean rainfall at regular 10 � 10 grids over theIndia landmass (Rajeevan et al. 2005). For the analysis, daily rainfall data of6,329 stations were considered during 1951–2003. There were only 1,803 stationswith a minimum 90% data availability during that period. In their analysis, theinterpolation method proposed by Shepard (1968) has been followed. This methodis based on the weights calculated from the distance between the station and the gridpoint and also the directional effects. Standard quality controls were performedbefore carrying out the interpolation analysis. Quality of the present gridded rainfallanalysis was also compared with similar global gridded rainfall data sets. Compari-son revealed that the present gridded rainfall analysis is better in more realisticrepresentation of spatial rainfall distribution. The IMD gridded rainfall dataset isbeing extensively used for many applications in validation of climate and numericalweather prediction models and also for studies on monsoon variability (Rajeevanet al. 2006; Dash et al. 2009).Fig. 2.8 Grid boxes used for calculating rainfall over Indian landmass in RegCM3 simulations at(a) 55 km and (b) 30 km resolutions24 S.K. Dash et al.In July 2002 very little rain was observed except in the west coast and north-eastof India (Fig. 2.9a) while in July 2003 good amount of rainfall was recordedthroughout the country (Fig. 2.10a). This contrasting rainfall is clearly seen forhigher resolution simulation of RegCM3 in comparison to resolution of 55 km.Figure 2.9a, c show that there is good amount of rainfall in west coast, some parts ofcoastal Andhra, coastal Orissa, along the foothills of Himalayas and north east Indiawhereas there is very less rainfall in north-west and some parts of central Indiawhile Fig. 2.9b shows very less rainfall in coastal Orissa and north east India. Onthe other hand in 2003, most of the country received good amount of rainfall exceptnorth-west India as shown in Fig. 2.10.30N 30N20N 20N10N 10N30N20N10N70E 70E80E 90E0 15 30 45 80 75 90 105 12080E 90E70E 80E 90Ea bcFig. 2.9 July 2002 mean rainfall (cm) at the grids over Indian landmass (a) IMD 1-degree, (b)RegCM3 55 km and (c) RegCM3 30 km2 Monthly and Seasonal Indian Summer Monsoon Simulated by RegCM3 25Figure 2.9c shows that in the simulations with 30 km grid size maximumrainfall is 75–90 cm. The model with 55 km simulates maximum rainfall of90–105 cm (Fig. 2.9b) over west coast whereas the corresponding observedvalue is 50