Weather Within Climate: Sub-seasonal Predictability of Tropical Daily Rainfall Characteristics

Abstract

Abstract The daily characteristics of tropical rainfall, such as rainfall frequency, are important application-relevant targets for sub-seasonal to seasonal (S2S) forecasts. Their potential predictability is assessed here based on observational estimates of spatial coherence of tropical rainfall anomalies estimated from the mean spatial autocorrelation in a 500-km radius. For the Indian summer monsoon, it is shown using 0.25-degrees data that daily rainfall frequency aggregated over 15-day running windows is more potentially predictable than the total rainfall amount, but that predictability decreases during the core of the monsoon. Extending the analysis to the global tropics, spatial coherence of interannual ranks of running 15-day amounts is shown to be generally much smaller over land, especially over regions of topography, than over ocean, with minima tending to coincide with the peak of the monsoon season across most of the continents and maxima usually occurring around the onset or withdrawal of the regional-scale monsoons. The areas of high spatial coherence of observed 15-day rainfall anomalies are shown to coincide partly (i.e., higher skill across oceans than over landmasses) with high anomaly correlation skill of European Centre for Medium-Range Weather Forecasts (ECMWF) week 3 + 4 precipitation reforecasts from the S2S database.