Validating HRRR Simulated Cloud Properties for Different Weather Phenomena Using Satellite and Radar Observations

In this study, we evaluate the ability of the HRRR model to forecast cloud top and in-cloud properties through comparison of observed and simulated satellite brightness temperatures (BTs) and radar reflectivity during different weather phenomenon. Three different phenomena were chosen over one week in December 2021: the Mayfield, KY tornado on 11 Dec 2021, a heavy snow event in Minnesota from 10-11 Dec 2021, and the Midwest Derecho on 15 Dec 2021. This analysis was accomplished using objects created by the Method for Object-Based Diagnostic Evaluation (MODE) tool.

First, the relationship between simulated satellite BTs and radar reflectivity is compared to the corresponding observations. Initial analysis shows that HRRR generally produced satellite and radar objects that were too large for the snow event, while the areal extent of the objects was too small at the beginning of the tornado event and for the duration of the derecho event. Yet, the HRRR forecasts did accurately depict the observed displacement between the observed radar and satellite objects.

HRRR forecast accuracy is assessed using the Object-based Threat Score (OTS). Based on the OTS, the HRRR model is the most accurate at forecasting the snow event, followed by the derecho and tornado events. This lower accuracy is due to the matches between simulated and observed images being worse for the tornado event compared to the snow event, as the smaller objects for the tornado event are generally less similar in size with a greater distance between the matching object centers than the snow event. This result illustrates the importance of examining model accuracy across a range of weather phenomenon.