A Forecast Assessment Of Az Wrf-Hrrr In Its First Operational North American Monsoon Season
Presenter: Dylan Girone1
Co-Author(s): C. Bayu Risanto, Christopher Castro
Advisor(s): Dr. Whitaker
1Department of Hydrology and Atmospheric Sciences, University of Arizona
The risk for severe weather in the Desert Southwest is greatly exacerbated during the North American Monsoon (NAM) Season. The mesoscale forcing of the systems generated by these mechanisms necessitates a model capable of explicitly resolving convection (convective allowing model, CAM). Researchers in the Department of Hydrology and Atmospheric Sciences developed the University of Arizona weather research and forecasting (UA-WRF a.k.a. AZ WRF) model with the goal of using a localized model to capture the regionality and seasonality of the NAM. For this study, we aimed to evaluate the daily and sub-daily precipitation forecast skill of AZ WRF and the national model that its lateral boundary forcing is derived from. We evaluated both models against observations from Multi-Radar Multi-Senor (MRMS) gridded precipitation data. We re-gridded the national High-Resolution Rapid Refresh (HRRR) model and the newly-operational AZ WRF-HRRR to a uniform 1-km x 1-km spatial resolution but employed a 25-km x 25-km neighborhood-based evaluation method for contingency table analyses. This allowed us to quantify forecast skill with metrics such as probability of detection (POD) and false alarm rate (FAR) as well as perform Monte Carlo significance tests based on the results. We will present the spatial statistics graphically and explain the implications of how these models resolve convection in complex terrain; the results could possibly inform future CAM development. The preliminary findings indicate that the AZ WRF improves predictions in areas of convective initiation and propagation, especially near population epicenters below the Mogollon Rim in east-central Arizona.