Development of a NWP based Integrated Block Level Forecast System (IBL-FS) using statistical post-processing technique for the state Jharkhand (India)

Abstract

A statistical post-processing forecast system for medium range predictions using the GFS model has been developed for Jharkhand (India) with the aim of improving rainfall and temperature predictions for agricultural applications. The basis of the integrated block level forecast system (IBL-FS) build includes (i) Decaying weighted mean (DWM) bias correction technique, (ii) Value addition and (iii) Inverse distance squared weighted (IDSW) interpolation. In the first step, model bias corrected district level forecast for 24 districts of Jharkhand is generated from the output of numerical GFS model (T1534L64) by applying DWM bias correction technique. In the second step, these bias corrected forecasts are value-added using forecast from various NWP models and synoptic methods. Finally in the third step, the IDSW interpolation method is used to generate the forecast at an unmeasured block from the value-added district level forecast of the surrounding districts. The value-added forecast for 263 blocks for the state Jharkhand is prepared up to medium range time scale (120h). The performance skill of IBL-FS is evaluated for rainfall during monsoon season 2018 and 2019, for minimum temperature during winter season 2019, and for maximum temperature during summer season 2019 using different statistical metrics. The skill of IBL-FS is found to be higher than the direct model forecast (DMFC) by 15% to 43% for minimum temperature, by 18% to 41% for maximum temperature, and by 22% to 30% for rainfall forecast for day1 to day5 forecasts. This study concludes that the integrated approach is more skillful than DMFC for real time forecasts and useful for farming for the blocks of Jharkhand.