Rainfall climatology and predictability over the Kingdom of Saudi Arabia at subseasonal scale

Abstract Precipitation over the Kingdom of Saudi Arabia (KSA) occurs mostly as isolated, episodic convective events during the cooler months (November–April). Rainfall forecasts at lead times on subseasonal to seasonal (S2S) timescales can potentially assist disaster‐risk mitigation and water resource management. Here we demonstrate predictability and forecast skill relevant for these sectors by performing retrospective ensemble forecast simulations over a 20‐year period from 1998 to 2017, out to a subseasonal timescale (up to four weeks ahead). We downscale the European Centre for Medium‐Range Weather Forecasts (ECMWF) S2S ensemble reforecasts dynamically using the Weather Research and Forecasting (WRF) model at convection‐permitting resolution (4 km). The simulations are initiated every Monday for 29 weeks, running for 30 days using 11 ensemble members targeting the forecast of November–April. A total of 191,400 hindcast days are generated to evaluate the predictability of KSA winter rainfall over the 20‐year period studied. Our results indicate that the WRF convection‐permitting model describes the precipitation patterns over the KSA adequately and improves the forecast skills statistically in comparison with its driving ECMWF fields. The model forecast skills generally decrease with increasing forecast range, with peak performance in the middle of the country and notable improvement in the southwest mountains. WRF extends forecast skills beyond ECMWF, by about one week. WRF forecasts outperform those of ECMWF by 87% in winter, emphasizing their value in simulating winter mesoscale convective systems. Rainfall forecasts in spring are already well forecast with ECMWF, and can be improved further by 8% with WRF.