Assessing subkilometre‐scale weather models for an extreme heat event over <scp>Paris</scp>

Abstract National meteorological agencies currently run their operational, regional numerical weather prediction (NWP) models at horizontal grid spacing of the order of a kilometre. There is an increase in demand for finer, subkilometre (subkm) spatial resolution to provide more localized and accurate services, especially over spatially heterogenic areas like urban centres. However, significant practical and scientific questions remain around the widespread adoption of subkm models for operational NWP. One of the main questions is about the added benefits they provide over kilometre‐scale models for high‐impact weather events and whether these benefits, if any, are substantial enough to justify the increase in computational cost involved to run them routinely. The present study attempts to bridge this knowledge gap by evaluating subkm models for an extreme heat event over Paris using the Met Office Unified Model. The study explores grid resolution strategies that strike a good balance between model skill and computational cost by comparing a subkm‐scale model of grid length 300‐m to a coarser 1.5‐km grid length model and a finer 100‐m model. To elucidate the impact of vertical resolution, three grids of 70, 90 and 140 vertical levels are employed in both the kilometre‐ and subkm‐scale models. Taken together, the results from this study show that while subkm models provide improvements in the representation of some processes, an increase in spatial resolution does not necessarily translate into an improvement in model skill across all variables of interest. This study demonstrates the need for further research to improve the skill of subkm models in the context of providing operational forecast guidance.