WRF prediction of an atmospheric river‐related precipitation event: Sensitivity to cumulus parameterization schemes

Abstract

Abstract The present study aimed to evaluate performance sensitivity to the Cumulus Parameterization Scheme (CPS) used for the Weather Research and Forecasting (WRF) model to predict the atmospheric river‐related precipitation (ARP) event with 206 and 57 mm, highest and area‐averaged precipitation (AAP) per 24‐h, respectively, that occurred over the central mountainous basins of Iran on 31 March 2019. For this purpose, experiments were designed using the 12 (almost all) CPSs available in WRF v4. To verify the predicted precipitation (from the inner 4‐km domain), both point‐scale and grid‐scale comparisons were performed against gauge‐ and satellite‐based observational data at three accumulation time‐scales (12‐, 18‐, and 24‐h) and in three distinct sub‐regions. All scores obtained from the different statistical metrics used, are in complete agreement with a strongly dependent performance of WRF on the CPS used. In addition, the use of Kain‐Fritsch, KF‐CuP, and Grell‐3 CPSs could provide a realistic picture of impending heavy precipitation for WRF. Contrary, the New SAS, Tiedtke, and Zhang‐McFarlane CPSs did not perform satisfactorily in predicting the ARP event. As a result, CPSs with the “momentum transport” option in their modification mechanism are unlikely to adequately simulate the conversion of incoming low‐level moisture from atmospheric river to precipitation. However, precipitation predictions are more accurate at the 24‐h accumulation time‐scale than at the 12‐ and 18‐h. Also, a dry bias in the predictions is expected as the terrain elevation and accumulation time‐scale decrease and the distance from the core of the precipitation field increases.

Keywords

• atmospheric river
• Cumulus Parameterization Scheme
• dynamic downscaling
• heavy precipitation
• model verification
• sensitivity analysis