GOVERNMENT OF INDIA

INDIA METEOROLOGICAL DEPARTMENT

LONG RANGE FORECAST UPDATE FOR
2003 SOUTH-WEST MONSOON SEASON RAINFALL

1. Introduction

This year, IMD introduced several new models for the Long Range Forecast of the South-West Monsoon rainfall. With this it has become possible to issue the long range forecasts in two stages: first in mid-April using data up to March and an update in mid-July using data up to June.

The first forecasts for the 2003 South-West Monsoon season rainfall (June-September) for the country as a whole using newly adopted power regression and probabilistic models were issued on 16th April. In the second stage, the following forecasts are now issued:

The three broad homogeneous regions are defined as follows:

2. Details of the 10-Parameter Power Regression and Probabilistic models

The 10-Parameter Power Regression and Probabilistic Models use 10 parameters, out of which 2 need data up to June. The parameters used in the model are: Nino-3 Sea Surface Temperature (July + August + September of Previous year), Eurasian Snow Cover (December of previous year), North-west Europe Mean Temperature (January), Europe Pressure Gradient (January), Arabian Sea Surface Temperature (January + February), 50 hPa Wind Pattern (January + February), East Asia Pressure (February + March), South Indian Ocean SST Index (March), Nino 3.4 SST tendency (April to June January to March) and South Indian Ocean zonal wind at 850 hPa (June). Data of 38 years (1958-1995) have been used for the model development and data for 7 years (1996-2002) have been used for an independent verification of the model performance. The model error of the 10-Parameter model is 4%. The Probabilistic Model uses the statistical discriminant analysis technique applied to the same 10 parameters. The model estimates the probability of South-West Monsoon season rainfall for the country as a whole in 5 categories viz., Drought (less than 90% of LPA), Below normal (90 to 97% of LPA), Near normal (98 to 102% of LPA), Above normal (103 to 110% of LPA) and Excess rainfall (more than 110 % of LPA).

3. Details of the Power Regression Model for July Rainfall

The Power Regression Model for July rainfall for the country as a whole uses the following 8 parameters: Eurasian Snow Cover (December of previous year), North Atlantic Sea Surface Temperature (December to February), Arabian Sea Surface Temperature (January + February), East Asian Pressure (February + March), South Indian Ocean SST Index (March to May), Zonal Wind Difference at 850 hPa between North Indian Ocean and North Pacific Ocean (May), Nino 3.4 SST Tendency (April to June January to March) and South Indian Ocean Zonal Wind at 850 hPa (June). The model was developed with 38 years of data (1958-1995) and 7 years of data (1996-2002) have been used for the independent verifications. The model has an error of 9%.

4. Details of the Power Regression Models for Broad Homogeneous Regions

The Power Regression Model for NW India was developed with the following 7 Parameters: Eurasian Snow cover (December of previous year), Europe Pressure Gradient (January), East Asian Pressure (February + March), South Indian Ocean SST Index (March to May), Zonal wind Difference at 850 hPa between North Indian Ocean and North Pacific Ocean (May), Nino 3.4 SST Tendency (April to June January to March) and South Indian Ocean Zonal Wind at 850 hPa (June).

The Power Regression Model for NE India uses the following 6 Parameters: Eurasian Snow cover (December of previous year), 50 hPa Wind pattern (January + February), East Asian Pressure (February + March), South Indian Ocean SST Index (March to May), Zonal wind difference at 850 hPa between North Indian Ocean and North Pacific Ocean (May) and Nino 3.4 SST Tendency (April to June January to March)

The Power Regression Model for Peninsula uses the following 7 Parameters: Nino 3 SST (July to September of Previous Year), Eurasian Snow cover (December of previous year), Arabian Sea Surface Temperature (January + February), East Asian Pressure (February + March), South Indian Ocean SST Index (March to May), Nino 3.4 SST Tendency (April to June January to March) and South Indian Ocean Zonal Wind at 850 hPa (June).

These Power Regression Models have a model error of 8%.

5. Forecast Updates for 2003 South-West Monsoon Season Rainfall

IMD's Long Range Forecast update for the 2003 South-West Monsoon Season (June-September) is that for the country as a whole the seasonal rainfall is likely to be 98% of the Long Period Average (LPA) with a model error of 4%.

IMD's updated probabilistic forecasts for the 2003 South-West monsoon season for the country as a whole are:

Rainfall in the month of July 2003 for the country as a whole is likely to be 102% of LPA with a model error of 9%.

Over the three broad homogeneous regions of the country, rainfall for the South-West Monsoon Season is likely to be 97% of LPA over Northwest India, 100% over Northeast India and 99% of LPA over the Peninsula, all with a model error 8 %.

R. R. Kelkar
Director General of Meteorology
New Delhi
The 9th July 2003