Our new paper, “Assessment of convection-permitting hydroclimate modeling in urban areas across the contiguous United States“, is published in Urban Climate (IF: 6.0). This work was led by undergraduate student Liam Thompson. Congratulations, Liam!

The paper can be downloaded at https://www.sciencedirect.com/science/article/pii/S2212095525000914.

Authors: Liam Thompson, Chenghao Wang, Cenlin He, Tzu-Shun Lin, Changhai Liu, and Jimy Dudhia

Abstract: Accurate representation of urban areas in weather and climate models is crucial for simulating interactions between urban surfaces and the atmospheric boundary layer, especially in high-resolution regional models that resolve deep convection. However, many continental-scale simulations use simplified urban parameterizations, raising questions about their ability to reproduce urban hydroclimate. This study evaluates CONUS404—a recent USGS-NCAR 4-km convection-permitting hydroclimate modeling dataset—in urban areas across the contiguous United States (CONUS). We assessed hourly near-surface air temperature, dewpoint, and wind speed simulations at 208 urban and 342 non-urban station locations from 2011 to 2020 using observations. Results show that CONUS404 performs better for air temperature in urban areas, with a slight mean warm bias (0.08 °C) at urban stations and a mean cold bias (−0.52 °C) at non-urban stations. Dewpoint simulations exhibit stronger dry biases at urban stations, suggesting underrepresented evapotranspiration from urban vegetation. Wind speed is generally underestimated, with average biases of −0.74 m s−1 at urban and −0.35 m s−1 at non-urban stations. Seasonal analyses reveal larger model errors for wintertime temperature and dewpoint that strongly depend on urban fraction. These findings highlight the limitations of the bulk urban parameterization in CONUS404, underscoring the need for enhanced urban representations to improve continental-scale hydroclimate simulations.

DOI: https://doi.org/10.1016/j.uclim.2025.102375

Fig. 9. Dependence of (a–c) MAE and (d–f) MBE for modeled hourly near-surface air temperature (a, d), dewpoint (b, e), and wind speed (c, f) on urban fraction across 208 urban grids. A linear or multi-linear regression line is fitted to the mean MAE and MBE of each plot, using the higher R2 of the fitted model. Each boxplot contains 40–43 urban grids/station locations. An F-test was performed to determine the overall goodness of fit of the model. N.S. indicates not significant, while *, **, and *** denote statistically significant with p < 0.05, 0.01, and 0.001, respectively. The box denotes the interquartile range (IQR) that represents the distribution of CONUS404 errors between the upper and lower quartiles, the whiskers represent the distribution of errors ±1.5 × IQR, and points outside the core box and whiskers are outliers.