Evaluation of seasonal prediction skills of summer monsoon precipitation over Pan India

Abstract

The Southwest (SW) monsoon season (June, July, August, and September) is the major period of rainfall activity in India. This study was mainly concerned with the prediction of the SW monsoon using output of several General Circulation Models (GCMs) with three different statistical approaches, namely, singular value decomposition-based multiple regression, supervised principal component regression and canonical correlation analysis. In the present study, the Multi-Model Ensemble (MME) approach was evaluated based on the results of General Circulation Models (GCMs). Precipitation forecast was generated as model output from coarse resolution MME-GCMs under ERFS project for summer monsoon 2020 and 2021 with April and May month’s initial conditions over 34 meteorological sub-divisions of India. The performance of the MME-GCMs based precipitation outputs was evaluated against IMD observed precipitation for the period 1982 to 2019. The evaluation of simulated GCMs precipitation was done with various statistical analysis like climatological mean, standard deviation, standardized anomaly index, forecast bias, root mean square error (RMSE), correlation coefficients, and phase coherency index etc. It was observed that the spatial pattern of precipitation was well captured and presented. Most MME-GCMs underestimates the distribution of precipitation over central India and the Western Ghats, while overestimates in the peninsular India. The spatial and temporal correlation coefficients were well captured by the different MME-GCMs for each region. The verification of ERFS based prediction was done for the summer monsoon season.  The skill scores like Forecast accuracy (ACC), bias score (BIAS), detection probability (POD), false alarm rate (FAR) and threat score (TS) were calculated using 3*3 contingency table over 34 meteorological sub-divisions of India. For the country as a whole, the observed (IMD) seasonal summer monsoon precipitation in 2020 and 2021 was 109% and 99% of the long-period average (LPA), respectively. The rainfall forecast from ERF using two different initial conditions for April and May, were 107% and 112% of the LPA in the 2020 summer monsoon and 105% and 102% of the LPA in the 2021 summer monsoon. The precipitation forecast was better represented at metsub-division level i.e. 0.53 and 0.59 during the 2020 monsoon and 0.44 and 0.62 during the 2021 monsoon for the April and May initial conditions, respectively.