Response of a Global Spectral Model for Simulation of Indian Summer Monsoon Rainfall

The present work highlights response of a global spectral model T80L18 with respect to Indian summer monsoon rainfall (ISMR) during 8 years period of 1996-2003. The model performance is evaluated for day-1, day-3 and day-4 retrospective 24-hour accumulated rainfall forecasts from 0300 UTC to the next day 0300 UTC using in-situ rainfall observations of 4491 stations. The model performance is evaluated by assessing: (i) percentage departure and root mean square error (RMSE) of seasonal rainfall forecast, (ii) coefficient of variation (CoV) of seasonal rainfall forecast and observation, along with percentage departure of monthly rainfall forecast and (iii) model performance during a drought and a normal year of 2002 and 2003, respectively. Generally, it is noted that the T80L18 model underestimated high rainfall and overestimated low rainfall, however, with increasing forecast duration prediction over low rainfall areas improved. The model RMSE over central and western India is found to increase with increasing forecast duration; however, the same was found to decrease over Jammu and Kashmir. The CoV of day-1 rainfall forecast is found to be low over all India in comparison to the observed data. In the case of model performance evaluation during a drought and a normal year of 2002 and 2003, it is noted that the model produced higher rainfall over the rainfall deficit regions of observed distribution; whereas the heaviest observed rainfall region (>250 cm) is not well resolved by the model. In general, the T80L18 model performance is noted to be better over central India for mean seasonal rainfall prediction.