Effects of Land and Ocean Initial Conditions on the Seasonal Prediction of Boreal Summer Monsoon Rainfall

Coupled climate models are essential tools for seasonal climate prediction, particularly in capturing the complex dynamics of tropical climate systems such as the monsoons. However, their predictive skill remains limited by uncertainties in initial conditions and model physics. This study evaluates the relative contributions of land and ocean initial conditions, 2 primary sources of seasonal predictability, on boreal summer monsoon prediction skill, with a specific focus on Indian Summer Monsoon Rainfall (ISMR). Observations and reanalysis datasets reveal statistically significant correlations between ISMR and antecedent (February) land surface and oceanic conditions, highlighting well-established sources of long-lead predictability. A suite of targeted sensitivity experiments using the Indian Institute of Tropical Meteorology-Climate Forecast System (IITM-CFS) shows that replacing realistic land initial conditions with climatological states does not degrade much of monsoon forecast skill and, in some cases, marginally improves it. Similarly, initializing land conditions using European Centre for Medium-Range Weather Forecasts Reanalysis v5 leads to moderate improvements in seasonal monsoon forecast skill, suggesting that enhancing land surface process in the model may yield greater benefits than solely improving initial conditions, likely due to the model’s unrealistic land–atmosphere coupling. Moreover, land initial conditions are found to have an impact on El Niño-Southern Oscillation statistics. Experiments using climatological mean ocean initial conditions result in a marked decline in forecast performance, including negative skill in key regions. These results underscore that oceanic initial conditions account for 50% to 60% of the observed covariance in seasonal boreal summer monsoon rainfall over Asia, Africa, and America. These findings provide clear evidences for further improving IITM-CFS development strategies aimed at enhancing predictability from other sources, such as the land surface processes.