Evaluating hydrostatic vs. non-hydrostatic dynamics in RegCM5: a 50-year simulation of near-surface wind speed over the Yellow River Basin

Abstract

Accurate simulation of near-surface wind speed over complex terrain is essential for wind energy assessment, ecological conservation, and hydroclimatic change studies. This study evaluates the RegCM5 regional climate model over the Yellow River Basin (YRB), China, using 50-year (1971–2020) ERA5-driven simulations at 25-km grid spacing. Three dynamical configurations are compared: the traditional hydrostatic core (Model 1), the MM5-like non-hydrostatic core (Model 2), and the newly introduced Moloch non-hydrostatic core (Model 3). Validation against the CN05.1 observational dataset shows that all configurations reproduce the broad spatial climatology but systematically overestimate wind speed (+ 1.3 to + 2.8 m s−1), particularly in the rugged Upper Reaches, where sub-grid topographic smoothing and insufficient effective surface drag are likely important. Model 3 yields the lowest mean bias and the closest probability distribution match, whereas Models 1 and 2 retain slightly higher correlations in some temporal and Taylor-diagram diagnostics. This contrast indicates that the Moloch core improves mean-state and distributional fidelity under the present setup but does not by itself resolve phase or trend errors. All models capture part of the interannual variability but substantially underestimate the observed terrestrial stilling trend, highlighting the need for updated or time-varying land-use/land-cover forcing and independent assessment of driving-field and observational uncertainties. The three configurations also differ markedly in computational cost. Model 2 is computationally expensive because of strict CFL constraints (dt = 25 s), whereas the Moloch core remains stable with dt = 200 s, reducing total runtime by 58% relative to Model 2 and 17% relative to Model 1 under the same computing environment. A one-decade sub-grid land-surface sensitivity experiment further shows that refined land-surface representation reduces the whole-basin positive bias by about 39%, although spatial correlation decreases in some regions. Overall, the Moloch core, especially when combined with improved land-surface representation, appears to be the most practical configuration for long-term RegCM5 wind downscaling over complex terrain within the present experimental design.