Development of operational power regression models for long range monsoon rainfall forecasts

1. Eight parameter and ten parameter power-regression models (Rajeevan et al. 2004)

During the period of 1988–2002, IMD’s operational forecasts were based on the 16-parameter power regression and parametric models (Gowariker et al.
1989, 1991). The sixteen parameter model give out its forecast on May 25. A particular advantage of having two power-regression models is that the first forecast can be generated in April itself with the 8-parameter model. The 10-parameter model provides flexibility for fine-tuning the April forecast on the basis of developments in June, particularly the El Nino/La Nina tendency.

The 8-parameter and 10-parameter models were developed using data of 38 years (1958–95). Another seven years’ data (1996–2002) were used for independent verification of the models. As part of the development of new power regression models for monsoon rainfall forecasts, every parameter in the 16-parameter model was examined for statistical stability over time by calculating running 21-year window correlations. The analysis revealed that all six April–May parameters and four winter–spring parameters showed weakening correlations. This lead to the removal of ten parameters from the set. The extensive data analysis yielded four new more stable and physically related predictors, for use in the long-range forecast model. A new parameter set consisting of six old and four new parameters was thus formulated for the purpose of further model development.

In the following list the parameters 1, 2, 4, 6, 7 are 8 old parameter taken from 16 parameter model and retained in the new model. All the others are new parameters freshly identified and used in the new models. The first eight parameters in the list are used for eight parameter model while the ten parameter model take first ten parameters in the list. The three other parameters in the list are used in some models other than these two.

1. Arabian sea SST (January and February)
2. Eurasian snow cover (December)
3. Northwest Europe temperature
4. Nino3 SST anomaly (previous year July to September)
5. South Indian Ocean SST index (March)
6. East Asia pressure (February and March)
7. Northern hemisphere 50 hPa wind pattern (January and February)
8. Europe pressure gradient (January)
9. South Indian Ocean 850 hPa zonal wind (june)
10. Nino3.4 SST tendency (April to June and January to March)
11. South Indian Ocean SST index (March to May)
12. North Indian Ocean and North Pacific Ocean 850 hPa zonal wind diffrence (May)
13. North atlantic Ocean SST (December , January and February)

2. six parameter models (Rajeevan et al. 2006)

The statistical models (Rajeevan et al. 2006) for the long range forecast of Indian Summer monsoon rainfall is based on the following techniques

1 Ensemble multiple linear regression
2 Projection pursuit regression

The parameters used for first stage (April) forecast are

A1 North Atlantic SST anomaly (20N–30N, 100W–80W)December and January
A2 Equatorial SE Indian Ocean SST anomaly (20S–10S, 100E–120E) February and March
A3 East Asia surface pressure anomaly (35N–45N, 120E–130E) February and March
A4 Europe land surface air temperature anomaly Five stations January
A5 Northwest Europe surface pressure anomaly tendency(65N–75N, 20E–40E) DJF(0)-SON(–1)
A6 WWV anomaly (5S–5N, 120E–80W) February and March

The parameters used for second stage (June) forecast are

J1 North Atlantic SST anomaly(20N–30N, 100W–80W) December and January
J2 Equatorial SE Indian Ocean SST anomaly(20S–10S, 100E–120E) February and March
J3 East Asia surface pressure anomaly(35N–45N, 120E–130E) February and March
J4 Nino-3.4 SST anomaly tendency(5S–5N, 170W–120W) MAM(0) – DJF(0)
J5 North Atlantic surface pressure anomaly(35N–45N, 30W–10W) May
J6 North Central Pacific zonal wind anomaly at 850 hPa(5N–15N, 180E–150W) May


1. Rajeevan,M.;Pai†,D. S.; Dikshit, S. K.; Kelkar, R. R.; 2004
IMD’s new operational models for long-range forecast of southwest monsoon rainfall over India and their verification for 2003

2. Rajeevan, M.; Pai,D. S.; Anil Kumar,R.; Lal,B.; 2006 New statistical models for long-range forecasting of southwest monsoon rainfall over India