Convective Biases in the US DOE Global Storm‐Resolving Model: Insights From Regionally Refined Simulations During the CACTI Campaign

Abstract Accurately simulating convective processes in complex terrain remains a critical challenge for global storm‐resolving models (GSRMs). This study systematically evaluates moist convective biases in the Regionally Refined Mesh configuration of the U.S. Department of Energy Simple Cloud‐Resolving E3SM Atmosphere Model (RRM‐SCREAM) using comprehensive observations and large‐eddy simulations from the Cloud, Aerosol, and Complex Terrain Interactions (CACTI) campaign in the mountainous area of central Argentina. Comparisons of simulations with high‐resolution observations and reanalysis data indicate that RRM‐SCREAM effectively captures large‐scale meteorological patterns, including regional atmospheric gradients and diurnal variability. However, RRM‐SCREAM disproportionately produces smaller precipitation clusters referred to as “popcorn convection,” and exaggerated rainfall intensities compared to observations and reference models. Detailed examination of a representative orographic shallow‐to‐deep convective transition case shows that RRM‐SCREAM delays initial shallow convection growth due to lower‐tropospheric dryness and sustained convective inhibition, but once triggered, deep convection becomes overly vigorous with excessively strong vertical velocities and elevated cloud ice content, linked to a thermodynamic structure characterized by suppressed low‐level moistening and excessive upper‐level moisture retention. Our results highlight specific deficiencies in the model representation of convective vertical velocity, cloud microphysical processes, and convective precipitation organization within RRM‐SCREAM. Addressing these biases is essential for improving the predictions of convective clouds and precipitation in the global high‐resolution atmospheric models.