Subseasonal prediction at km-scale: the 2015 Texas-Oklahoma Extreme Rainfall-Flood event

Severe convective storms pose significant challenges to societal resilience and represent a critical test for Subseasonal-to-Seasonal (S2S) forecasting at longer lead-times. This study investigates the predictability of the torrential 2015 May Texas-Oklahoma extreme rainfall event, during which record-breaking rainfall abruptly terminated a multi-year drought, only to be followed by a second wave of heavy rainfall by Tropical Storm Bill in June. We evaluated the performance of the MPAS-NoahMP S2S prediction system in capturing this extreme rainfall event. Three sets of global mesh are designed, a global 60-km uniform mesh, two regional refinement mesh centered in the US for 60-15km, and 60-4km going down to convection-permitting resolution. 

At the 1-week lead time, ensemble forecasts demonstrate high fidelity, skillfully capturing the timing, magnitude, and spatial pattern of precipitation anomalies. At 2- and 3-week lead times, the model maintains a persistent signal of the May wet event, albeit with a damped magnitude and significantly larger ensemble spread, which itself is a useful indicator of potential high-impact weather. We further investigate the added values of regional refinement for this extreme rainfall event, in terms of extreme precipitation distribution, diurnal cycle, and land-atmosphere interactions priori to the rainfall. 

This study discusses the applications of km-scale convection-permitting simulation in subseasonal forecasts (2-6 week) and the valuable findings translating probabilistic S2S forecasts into actionable intelligence for stakeholders, such as water managers, who must navigate these increasingly volatile weather regimes.