Analysis of Monsoon Mission Coupled Forecasting System (MMCFS) model simulations of sub-division scale temperatures over India for the hot weather season (April–June)

Abstract

This study examines skill of the Monsoon Mission Climate Forecasting System (MMCFS) model simulations of monthly and seasonal maximum, minimum and mean temperatures of hot weather season (April, May and June) for the period 1982–2008 over India. The hindcast skill at the sub-division and all India scales were computed. The hindcasts were prepared using initial conditions (ICs) pertaining to January, February and March. The bias-corrected forecast for the 2016 AMJ season was also verified with the high resolution gridded temperature data of the India Meteorological Department (IMD). Standard verification skill scores, namely correlation coefficient (CC) and root mean square error (RMSE) have been used to assess the hindcast skill at various lead times. The grid point level statistical bias-correction was successful in reducing the bias and RMSE of the MMCFS model hindcast at the sub-division and all India scales. The hindcast analysis showed that for all the considered ICs and during all the months (April, May and June) and AMJ season for maximum, minimum and mean temperature bias of four sub-divisions Jammu & Kashmir (J&K), Himachal Pradesh, Uttarakhand, and Arunachal Pradesh showed bias $${\le }{-}2.0^{\circ }\hbox {C}$$ and four sub-divisions Saurashtra and Kutch, Bihar, Gangetic West Bengal, Sub-Himalayan West Bengal (SHWB) and Sikkim showed bias $${\ge }2^{\circ }\hbox {C}$$ . Hindcast based on the February and March ICs showed the best skill both in sub-division scale as well as in all India scale. Similarly, among the three months of the AMJ season, model skills based on considered ICs were best for the April month. Many of the sub-divisions from northwest India and neighboring central India and along the west coast showed significant hindcast skill for simulations based on February and March ICs. For the all India averaged temperature, March IC based forecast showed the highest skill for all the months and AMJ season followed by February IC. The MMCFS model forecasts for the 2016 monthly and seasonal temperatures were able to indicate correct signs of the observed temperature anomalies in most of the sub-divisions. The pattern anomaly correlations for the May and June forecasts based on March IC were significant at $$\ge $$ 95% level.