Deep Learning-Driven Mapping of Pre-Monsoon features for Indian Summer Monsoon Precipitation Forecasting

Accurate forecasting of the Indian summer monsoon is critical for water resource management, agricultural planning, and disaster mitigation. In this study, we developed a unique DL methodology to assess whether pre-monsoon atmospheric features can reliably predict monsoon season rainfall. Our model integrates several key parameters, namely, SST, SKT, near-surface air temperature (t2m), TCWV, and a convective index (vimd), which are merged to form a unified thermal variable that captures both marine and terrestrial influences. The domain-summed inland Indian region rainfall is analyzed over a test period from 2012 to 2024 using CDFs, spatial precipitation maps, frequency-of-exceedance curves, and grouped annual performance metrics. The CDF analysis reveals that observed cumulative monthly precipitation for test data has a median of approximately 1.2 m and a mean of 1.3 m, while the predicted distribution exhibits a lower median of about 1.1 m and a mean near 1.0 m, indicating systematic under-prediction, particularly in the upper tail, where the 90th percentile of observed values reaches roughly 1.8 m compared to 1.5 m for predictions. Spatial maps demonstrate that although the model captures broad rainfall patterns along the Western Ghats and Bay of Bengal, regional biases persist, with coastal areas often overestimated and inland regions underestimated. The frequency-of-exceedance analysis further indicates that the model underestimates the occurrence of extreme events. Grouped annual performance metrics, featuring an average correlation of 0.78 and a Kling–Gupta Efficiency of 0.72, underscore the model’s moderate skill across varying monsoon conditions. The statistical analysis of model performance suggests that although there are associated specific systematic biases intrinsic to the model, overall, our integrated DL model effectively captures the general properties of monsoon precipitation and hence can be utilized for extended-range precipitation forecasting.