Resolving the street-scale Urban Heat Island with a coupled WRF-LES-SLUCM atmospheric modeling system

Urban heat islands (UHIs), exacerbated by heat waves, can lead to dangerously high temperatures, posing significant risks to public health, the economy, and urban infrastructure. These extreme heat events disproportionately impact underserved communities due to past policies and inequitable development patterns. To better understand and mitigate these effects, it is essential to simulate UHIs at fine resolutions, particularly at the neighborhood scale, to unravel the physical and socioeconomic drivers of extreme heat and evaluate potential mitigation strategies. The Weather Research and Forecasting model coupled with the Single-Layer Urban Canopy Model (WRF-SLUCM) is commonly used for UHI simulations at a 1-kilometer scale. In this study, we evaluated the applicability and sensitivity of WRF in large-eddy simulation mode coupled with SLUCM (WRF-LES-SLUCM) for simulating UHIs in Baltimore, MD. We modeled the city at a 150-meter resolution over a 97.2 km² domain from August 19 to August 22, 2017, a period during which temperatures exceeded 33°C. Various scenarios were analyzed to assess sensitivity to soil moisture initialization and sea surface temperature (SST) inputs. The WRF-LES model performed well in capturing land surface temperatures (LST) over urban areas, with root-mean-square errors (RMSE) ranging from 2.68 to 3.34°C compared to Landsat measurements. Comparisons with 2-m air temperature (AT) observations from eight NOAA AWOS/ASOS stations showed good performance, with RMSE ranging from 1.5 to 3.0°C over the 96-hour simulation. Daytime AT errors were generally lower (<0.2°C) than nighttime errors across most stations and scenarios. Our findings demonstrate that WRF-LES-SLUCM is a robust tool for simulating UHI effects at fine spatial scales. This modeling can inform interventions and illuminate how neighborhood-scale urban forms and land cover influence heat exposure. Moreover, the results reveal the superior efficacy of the SLUCM parameterization in predicting LST and AT over urban land covers compared to the Noah model's performance over non-urban landscapes. However, there is a need for further refinement of intraurban SLUCM parameterizations to enhance model accuracy.