Current Status and Prospects of Drought Forecasting in South Asia L. S. Rathore, Akhilesh Gupta and J.V. Singh National Centre for Medium Range Weather Forecasting Department of Science & Technology Government of India Website: www.ncmrwf.gov.in
PREDICTING DROUGHT • Difficult for most locations • Anomalies last for months to decades depending upon features of global circulation • Big picture: Global weather patterns • Little picture: High Pressure • Tropical outlook: SST, ENSO • Prediction needed for too many variables • Long range prediction of precipitation and temperature • Tools: Empirical, Statistical, Climate models
Information needed for drought prediction • Onset/termination dates of rainy season • Precip./Temp. forecast in medium range • Precip./Temp forecast in extended to seasonal scale • Dry-spell duration and its culmination into drought Predicting drought means predicting Precip. & Temp, the prime causatives, in all temporal ranges i.e. medium, extended and seasonal.
Multi-decadal changes in Break Days Data of past 50 years show that number of Break days are more in July as compared to August
Rainfall Percentage departure for the country as a whole in recent major drought years
July Rainfall in Recent Years Since 1991, number of years (11 out of 14 years) with less than LPA of July rainfall was more as compared to previous decades. 2002 was the worst year with All India rainfall 54% below normal.
First Half of Monsoon Rainfall: 1901-2004 Large inter-annual variability can be noticed for June-July rainfall. Years with above normal and below normal rainfall are nearly equally distributed
NCMRWF’s Forecast System for Drought Prediction • NCMRWF has developed a Deterministic Medium- Range Weather Forecasting System. • Weather forecasts up to 7 days is being produced using Global Data Assimilation and Forecast Systems at T80L18 and T170L28 resolutions. • High-Resolution Mesoscale Forecasts up to 72 hrs using MM5 and Eta Models. • Extended range prediction – Experimental (monthly scales). * Weather based farm advisory service
NUMERICAL WEATHER PREDICTION MODELS AT NCMRWF # Operational Model
GLOBAL DATA ASSIMILATION-FORECAST SYSTEM Satellite Aircraft Radiosonde Surface observations Pilot Balloon GTS DATA Ships RTH IMD DATA RECEPTION AT NCMRWF DATA PROCESSING & QUALITY CONTROL 06 12 18 00 ANALYSIS ANALYSIS ANALYSIS ANALYSIS 6 HR FCST 6 HR FCST 6 HR FCST GLOBAL DATA ASSIMILATION GLOBAL SPECTRAL MODEL 7 DAYS FORECASTS ARCHIVAL FORECAST DISSEMINATION TO USERS
Computing Resources at NCMRWF Cray SV1: 24-Processors- 1.2 GFlops per processor, 8 GB Main Memory, 800 GB Disk DEC-ALPHA:Parallel Processing System 2- Servers AS4100 @600 MHz, Memory– 1GB each 9-Work Stations @600 MHz, Memory– 512MB each Switch: Gigabit Ethernet Smart Switch Router ORIGIN 200:Parallel Processing System 2- Servers: 4 CPU each @225 MHz, Memory– 1GB each ORIGIN 200:Single CPU Servers 3- Servers @270 MHz, Memory– 512 MB each 1- Server @180 MHz, Memory– 512 MB 4- O2 WORK STATIONS:@200 MHz, Memory– 512MB each PARAM 10000:Parallel Processing System 2- SUN Ultrasparc-II Servers (4 CPU each) @300 MHz, Memory– 1GB each, ( Switch: MYRINET) LOCAL AREA NETWORK:on Fast Ethernet. 4- LINUX SERVERS: (WEB, FTP, PROXY, PRINT) Internet: 2 MBPS Leased Line
AGROMETEOROLOGICAL ADVISORY SERVICE OF NCMRWF NCMRWF PREPARATION OF LOCATION SPECIFIC FORECAST FEEDBACK FROM AAS UNIT Internet FAX PHONE • AAS UNITS (SAUs / ICAR institutes) PREPARATION OF AGROMET ADVISORY BULLETIN FARMER’s FEEDBACK AIR PERSONAL T.V. PRINT CONTACT FARMER
NETWORK OF AGROMET ADVISORY SERVICE (AAS) UNITS OF NCMRWF
Dry Spells in July 2004 • Monsoon 2004- Started well and remained in good phase till 3rd week of June. As on June end; All India rainfall was 2% below normal. 19/36 sub-divisions reported normal rainfall 9/36 subdivisions reported excess rainfall 61% of district received normal to excess rains. • July witnessed break conditions, which culminated into; All India rainfall of 15% below normal. Only 19 subdivisions received normal to excess rainfall. 45% districts received normal to excess rainfall. • Sub-divisions like W and E Rajasthan, Gujarat, Saurashtra, H.P., West U.P., Haryana, Chandigarh & Delhi, Punjab, Jharkhand, E and W M. P., Madhya Maharashtra and Vidarbha remained deficient for several weeks.
Monsoon-2004: Drought Monitoring Growing Concern: Expansion of R/F Deficient Area with time 1 June-14 July
Monsoon-2004:Medium Range Forecast (One week in advance) issued to Min. of Agri.
Monsoon-2004: Medium Range Drought Prediction A typical bulletin of NCMRWF on MRWF of Monsoon situation issued for the week 19-25 July • Monsoon is once again going into a BREAK PHASE. • Model predictions do not indicate revival of monsoon • during next 5 days. • NE States, Sub-himalayan West Bengal and North • Bihar may receive widespread rains with isolated • heavy to very falls during next 5 days. • Rainfall deficiency in some of the already highly • distressed sub-divisions such as East and West • Rajasthan, Haryana, Punjab, West UP, West MP, and • Vidarbha, may grow further during next week.
Extended Range Prediction • System of NCMRWF: • Model Climate • ii) Predicted Sea Surface Temperature (SST) fields • iii) Ensemble Integration of the Model • iv) Forecast Preparation • v) Down Scaling
Forecast Preparation : RAINFALL ANOMALIES • For each run, Area-Weighted Rainfall Anomalies • over 36 Meteorological Sub-Divisions in India • are computed • For each run, Area-Weighted Rainfall Anomalies • over 6 homogenous zones over India are computed • West Coast (ii) Peninsular India • (iii) Central East India (iv) North-East India • (v) North-West India (vi) Himalayan Belt
Forecast anomalies for each zone from each ensemble member runs are examined and given a category > Excess: > 20% Normal: between –20% to 20% Deficient: less than -20% Probability of prediction is computed by counting how many runs have predicted which category of rainfall anomaly for a zone (e.g. if 10 runs are made, and for North-East zone, 7 runs Predict excess rainfall, 2 predict normal and 1 deficient, probability of forecast is given in % as (70,20,10) or (7,2,1) as number. If the probability of forecast for a zone exceeds 80% for any category, sufficient confidence exists, and on the Anomaly Map, the zone is shaded.
No Shading denotes areas where category could not be determined as the spread is quite large
Observed Rain June 2002 (source: IMD) Performance of NCMRWF ERP Was generally GOOD for JUNE
DOWNSCALING: ? The ERP prepared at 1.4x1.4 degrees. DOWNSCALING NEEDED To generate forecast for smaller domains: DYNAMICAL DOWNSCALING Employing Eta Model (32x32km Grid) using the Forecasts of the Global Model STATISTICAL DOWNSCALING Perfect Prog Method at monthly scale Use Weather Generator for temporal downscaling from monthly to daily values
DOAC Initiative for ERP(IITD, IMD, NCMRWF, SAC, ICAR etc.) • Global Ocean Atmosphere Coupled Model Products (GOACM) • Nested Grid high resolution Regional Climate Model (RCM) • Multi model Super-ensemble approach based on dynamical model products and synoptic scale signals using statistical/GIS techniques • Deterministic and probabilistic prediction of precipitation, surface air temperature and soil moisture
CONCLUSIONS MRWF skillful and can play a role in indicating synthesis of drought over small spatial domains. Outreach system (AAS) existing at 83 zones. ERP System is under development. Evaluation of the Predictions in progress. Generally Rainfall Prediction over Peninsula, eastern and NE Regions good. Role of Initial Data, SST and other predictands on seasonal forecast need to be examined further.