Abstract/Summary

Daily initialized coupled and uncoupled numerical weather prediction (NWP) forecasts from the global Met Office Unified Model (MetUM) are compared for the 2016 Indian summer monsoon. Three MetUM configurations are used: atmosphere‐only (ATM), coupled to a mixed‐layer ocean model (KPP), and coupled to a dynamical ocean model (NEMO). The analysis focuses on the impact of air‐sea coupling, particularly in the Bay of Bengal (BoB), on NWP for monsoon rainfall. Seasonal‐mean biases in all three configurations are highly consistent and driven by errors in atmospheric processes. Rainfall is initially overestimated over India, but underestimated over the BoB, the latter associated with too much shortwave radiation and too little cloud cover in MetUM. The excess shortwave radiation (>40 Wm‐2 over the northwest BoB) is partially compensated by additional latent cooling, primarily due to overestimated surface wind speeds. In NEMO and KPP, coupling improves the timing of intraseasonal active and break phases over India, primarily the end of these phases, which are systematically too late in ATM. NEMO and KPP show a more realistic intraseasonal local phase relationship between sea surface temperature (SST) and rainfall throughout the BoB, but no configuration reproduces the observed significant lagged relationship between BoB SST and Indian rainfall. The lack of this relationship may be partly attributed to weak heat flux feedbacks to northern BoB SST, with the forecast shortwave feedback having systematically the wrong sign (positive) compared to satellite radiation, and thus contributing to SST warming at all lead times. Based on these MetUM forecasts, there is a limited impact of coupling on NWP for monsoon rainfall, both for the mean rainfall and intraseasonal variability. Further research to improve NWP for monsoon rainfall should focus on reducing MetUM atmospheric systematical biases.