Understanding Medium-Range Convective Scale Precipitation Errors from the Model for Prediction Across Scales-Atmosphere

The capability to accurately forecast precipitation events beyond a short-term horizon is crucial for those such as stakeholders or emergency management professionals. However, current modeling approaches are limited in this capacity. The Model for Prediction Across Scales (MPAS) variable-resolution capability enables the use of convection-allowing resolution in regions of interest and smoothly transitions to lower-resolution globally to effectively capture medium-range weather patterns. This research aims to analyze the spatial variability of errors at convective scales and assess how storm development in MPAS compares with observed storm evolution. Additionally, this study evaluates whether the variability exhibited by MPAS ensemble simulations can capture some of the forecast uncertainty. Ten-member 7-day forecast ensemble simulations, initialized daily over May 2019, were executed using MPAS with a 3-km horizontal mesh resolution over the contiguous United States (CONUS) and a 15-km horizontal mesh resolution for the remainder of the globe. A comparison is conducted between MPAS precipitation forecasts and the National Severe Storms Laboratory’s multi-radar/multi-sensor system (MRMS) precipitation estimates. The Method for Object-Based Diagnostic Evaluation (MODE) tool is employed to identify convective objects and their characteristics for this analysis, facilitating the assessment of spread and error in zonal displacement of forecasted versus observed objects through hypothesis testing. The ensemble forecast objects are then categorized into two groups: those that are under-dispersive and those that are not. Preliminary findings indicate that MPAS demonstrates superior predictive skills for storms likely more influenced by large-scale atmospheric dynamics, whereas it exhibits diminished accuracy for storms that are more subjected to localized forcing mechanisms.