Ensemble sensitivities of WRF-ARW forecasts at high resolution

Decision making for many weather-sensitive operations depends on the availability of highly localized predictions which are critical for emergency management by municipal and state government agencies, and private transportation and utility companies. To provide high resolution forecasts relevant to such decisions with a focus at the meso-gamma-scale up to 72 hours for the greater New York City metropolitan area, an operational mesoscale prediction system, dubbed “Deep Thunder”, has been implemented at the IBM Thomas J. Watson Research Center. Its forecasting component is based on the WRF-ARW model with data assimilation of surface observations from more than 100 sites in the northeastern United States.

In this study we examine ensemble-based sensitivities of high resolution near-surface temperature, wind and precipitation forecasts to errors in initial conditions that are important for data assimilation within “Deep Thunder“, and for predictability studies. Numerical experiments are performed for horizontal grid resolutions of 10-20 km and 4-6 km, with a larger ensemble size used for higher resolution case to obtain reliable sensitivity patterns. The predicted impact of surface observations on forecast accuracy based on computed sensitivity is analyzed for both cases. Spatial and temporal aspects of ensemble-based sensitivities for high resolution forecasts are discussed.