Ocean–atmosphere processes associated with enhanced Indian monsoon break spells in CFSv2 forecasts

Abstract

This paper analyzes the role of ocean–atmosphere processes associated with break days and their impact on dry-land biases of Indian summer monsoon in Climate Forecast System version 2 (CFSv2)’s sub daily and monthly hindcasts, which are produced by initializing the forecast system every 5 days in each month from January 1982 to December 2007. Each initialized forecast produces 24 ensemble members and they are characterized by a systematic dry-land bias over central India and have been interlinked with higher break days in majority of the ensemble members. Analyses on 15–45 day band-passed summer precipitation anomalies indicate that the monsoon intraseasonal oscillations are very weak in Mar- and Apr-initialized forecasts while they get stronger in May- and Jun-initialized forecasts. The persistent low-level anticyclonic activity in the Findlater jet in the longer lead months (Mar- and Apr-initialized forecasts) is systematically diminished in the summer forecasts that are produced by initializing the forecast system closer to the June–September (JJAS) season (e.g., May- and Jun-initialized forecasts). This low-level negative vorticity bias extends across central India and diminishes with the systematic oceanic mixed layer shoaling in the northern Bay of Bengal (BoB) and Indian Ocean (IO). The deeper ocean mixed layer in BoB and northern IO in conjunction with persistent negative vorticity biases suppresses intraseasonal variability of Indian summer monsoon rainfall and induces a large number of break days, which lead to the dry-land biases in July and August in the core monsoon zone. CFSv2-produced Pacific sea surface temperatures exert a strong influence on mixed layer depths (MLDs) across the basins in Arabian Sea (AS), BOB and IO unlike in observations. The forecasted extended break spells are systematically and highly correlated with the model produced Nino3 index while no such strong correlation is found between total monsoonal break days and ENSO in the observations. The May- and Jun-initialized CFSv2 forecasts reproduce the Indian summer monsoon rainfall in July–August reliably by reasonably alleviating the ocean–atmosphere coupled biases relative to earlier initializations.