How well does ERA5 reanalysis depict low-level winds associated with nocturnal rainfall in Sichuan Basin

Reanalysis data, while widely used to study rainfall mechanisms, requires rigorous uncertainty assessment, especially over complex terrains such as the Sichuan Basin (SCB). This study evaluates the performance of ERA5 in capturing low-level winds associated with SCB nocturnal rainfall, using hourly wind profiles from four wind profiler radars. Analysis confirms key features of SCB rainfall dynamics previously identified using reanalysis and model outputs, including nocturnal flow enhancement over the Yungui Plateau (YGP) and its positive correlation with low-level vorticity in the SCB, which ERA5 generally reproduces. However, substantial discrepancies emerge between ERA5 and observations, particularly in the vertical structure of low-level winds. ERA5 underestimates the height of the maximum wind speed over the YGP while overestimating its magnitude. This creates distinct vertical wind structures during SCB nocturnal rainfall events: observations exhibit an elevated peak (“prominent nose”), whereas ERA5 shows a stronger but lower-altitude maximum (“flatter nose”). Correspondingly, ERA5 overestimates lower-level intensity of the circulation over SCB and fails to reproduce the observed increase in circulation with height. These vertical structure biases highlight limitations in ERA5's representation of terrain-modulated nocturnal flows, emphasizing that uncritical use of reanalysis may misrepresent key mechanisms driving SCB rainfall. Our results underscore the necessity for enhanced observational data collection and analysis to constrain uncertainties in reanalysis and model outputs, thereby improving their reliability for applications such as low-level aviation and wind energy assessment, particularly over complex terrains. • First evaluation of ERA5 winds vs. profiler data for SCB nocturnal rainfall. • ERA5 reveals systematic biases in vertical wind structure over complex terrain. • Underscores need for enhanced data to constrain reanalysis uncertainties.