An Efficient Method to Identify Regions of Initial Condition Uncertainty for Short-Range Rainfall Forecast during Indian Summer Monsoon Season

Abstract

Abstract Forecast errors can be reduced if observations are assimilated in regions with the largest percentage of analysis error growth. Adjoint or ensemble sensitivity analysis provides a way to determine the relationship between a scalar forecast metric and analysis/forecast state variable whereby regions with the largest sensitivity can be identified. In this study, the regions where additional observations will impact the forecasts over the Indian subcontinent during the summer monsoon season have been identified using the ensemble approach. Five years of ensemble data drawn from the National Centers for Environmental Prediction (NCEP) The Observing System Research and Predictability Experiment (THORPEX) Interactive Grand Global Ensemble (TIGGE) archive has been used to perform ensemble sensitivity analysis (ESA). Regions with the largest percentage of sensitive forecast cycles may be considered as potential locations for launching new observations. Results show that the climatological sensitivity fields of the target forecast region are most often located in its upstream regions in the analysis. It is found that the region of frequent sensitivity may significantly vary with the forecast variable used in the calculation. The 24-h and 48-h precipitation forecasts averaged over the Western Ghats are most often sensitive to far south of the target forecast region. The region of sensitivity for the forecast domain located over the Gangetic Basin and North-Eastern region of the Indian subcontinent is identified to be a zonally elongated region over central India and head Bay, respectively.