Multi-physics schema for sub-seasonal prediction of Indian summer monsoon

The biases due to model physics in simulating the Indian summer monsoon are investigated to construct a multi-physics forecast strategy for sub-seasonal prediction. A 15-year hindcast from May to September (model is initialized six times every month) is produced with Climate Forecast System version 2 (CFSv2) using four physics combinations. These physics schemes include simplified Arakawa Shubert (SAS) and revised Arakawa Shubert (NSAS) convection parameterization coupled with Zhao and Carr (ZC) and Ferrier (FER) microphysics. The spatial and temporal characteristics of predicted monsoon climatology and its biases are evaluated using observation and reanalysis datasets. All physics schemes could predict rainfall distribution but have biases in predicted spatial rainfall maxima, and low-level circulation resulted from erroneous diabatic heating. Stronger deep convection in NSAS increased rainfall over the Indian landmass and caused excessive precipitation over oceanic regions. The inclusion of critical mixed-phase microphysical processes tends to reduce high cloud fraction in FER. It is illustrated that such constraining refinements due to one or more modified aspects of these physics schemes can be utilized to create a multi-physics prediction scenario. The multi-physics ensemble’s mean showcases the considerable skill and controlled error-growth up to 20-day lead time compared to each physics combination for the monsoon zone. The study anticipates that the multi-physics approach could exploit individual physical schemes’ strengths to yield better sub-seasonal forecasts.