Huang et al. (2025) Urban heat forecasting in small cities: evaluation of a high-resolution operational numerical weather prediction model

Identification

• Journal: Geoscientific model development
• Year: 2025
• Date: 2025-11-28
• Authors: Yuqi Huang, Chenghao Wang, Tyler Danzig, Temple R. Lee, Sandip Pal
• DOI: 10.5194/gmd-18-9237-2025

Research Groups

• School of Meteorology, University of Oklahoma, Norman, OK, USA
• Department of Geography and Sustainability, University of Oklahoma, Norman, OK, USA
• Department of Geosciences, Texas Tech University, Lubbock, TX, USA
• NOAA/Air Resources Laboratory, Oak Ridge, TN, USA

Short Summary

This study evaluates the performance of the High-Resolution Rapid Refresh (HRRR) model in forecasting urban heat dynamics, including temperature, humidity, nocturnal cooling, and urban heat advection (UHA), in a small, semi-arid city (Lubbock, Texas). Findings reveal systematic biases in HRRR forecasts and limitations in its urban representation, highlighting the need for improved urban parameterizations in numerical weather prediction models for small to mid-sized cities.

Objective

• To assess how accurately the HRRR model simulates near-surface hydrometeorological conditions (temperature and humidity) within and around a semi-arid urban environment across different times of the day and throughout varying seasons.
• To determine the extent to which the HRRR model reproduces observed nocturnal cooling rates across heterogeneous urban and rural landscapes.
• To investigate if the HRRR model can capture the spatial variability and magnitude of Urban Heat Island (UHI) and urban heat advection (UHA), particularly under varying wind regimes.

Study Configuration

• Spatial Scale: Lubbock, Texas, USA (population 266,878), a small-sized city in a semi-arid environment. HRRR model resolution is 3 km.
• Temporal Scale: One year, from 1 September 2023, to 31 August 2024. Evaluation focuses on 18-hour HRRR forecasts.

Methodology and Data

• Models used: High-Resolution Rapid Refresh (HRRRv4) operational numerical weather prediction (NWP) model, which incorporates the Rapid Update Cycle Land Surface Model (RUC LSM) with a slab urban parameterization scheme.
• Data sources:
  • In situ observations:
    • Urban Heat Island Experiment in Lubbock, Texas (U-HEAT) network: 23 stations providing 2 m air temperature and 2 m dew point temperature at 1-minute or 5-minute intervals.
    • West Texas Mesonet (WTM): 5 stations providing 10 m wind speed and direction at 1-minute intervals.
  • Model data: HRRRv4 18-hour forecasts and 0-hour forecasts (model initialization) for 2 m air temperature, 2 m dew point temperature, 10 m westward wind component, 10 m southward wind component, and total cloud cover.
  • Evaluation metrics: Mean Bias Error (MBE), Root Mean Square Error (RMSE), and Pearson correlation coefficient (r).
  • Analysis methods: Nocturnal cooling rate calculation (regression-derived, restricted to nights with <25% cloud cover), and Urban Heat Advection (UHA) assessment using a wind-direction-dependent analytical framework (adapted from Heaviside et al., 2015), classifying mean nighttime 10 m wind fields into eight directional sectors.

Main Results

• HRRRv4 forecasts show strong correlations for 2 m air temperature (average r = 0.98 ± 0.00, RMSE = 2.01 ± 0.12 °C) and 2 m dew point temperature (average r = 0.93 ± 0.01, RMSE = 4.06 ± 0.19 °C) across all sites.
• Wind speed forecasts exhibit persistent overestimations (average MBE = 0.29 ± 0.12 m/s) and weaker correlations (r ranging from 0.37 to 0.39) across all Mesonet sites, particularly at urban locations.
• Diurnal biases for 2 m air temperature: Rural sites show a slight warm bias during the day (average annual MBE = 0.10 °C) and a cold bias at night. Urban sites exhibit a consistent cold bias (average annual MBE = -0.27 °C), especially pronounced at night (average annual MBE = -0.57 °C).
• A pervasive dry bias is observed for 2 m dew point temperature across all seasons and sites (average MBE = -1.95 ± 0.22 °C), being most pronounced in spring. Urban-rural differences in dew point temperature biases are minimal.
• HRRR consistently underestimates nocturnal cooling rates for both urban (average MBE = -0.12 ± 0.27 °C/h) and rural sites (average MBE = -0.14 ± 0.26 °C/h), though it effectively captures the expected slower cooling in urban areas.
• The model reasonably captures large-scale wind-direction patterns but fails to accurately reproduce Urban Heat Advection (UHA) patterns under most wind regimes, showing limited sensitivity to wind speed. For instance, it incorrectly predicts a cold-core anomaly over the urban area under northerly winds.
• Comparison of 0-hour (near-real-time) and 18-hour forecasts indicates that data assimilation partially improves predictive performance, particularly by reducing extreme values and improving the diurnal cycle for temperature and dew point, but shows negligible improvement for wind speed.

Contributions

• Provides the first systematic, year-long evaluation of an operational high-resolution NWP model (HRRR) for urban heat dynamics, including urban heat advection, in a small-sized, semi-arid city.
• Identifies specific systematic biases and urban representation limitations in current high-resolution NWP forecasts, particularly concerning near-surface hydrometeorological conditions, nocturnal cooling, and urban heat advection.
• Highlights the critical need for improved urban parameterizations (e.g., explicit urban geometry, dynamic surface properties, anthropogenic heat fluxes) and enhanced evaluation frameworks focused on forecast skill for small and mid-sized cities.
• Offers valuable insights for advancing urban weather forecasting and improving heat-risk warning systems in vulnerable urban environments.

Funding

• National Oceanic and Atmospheric Administration (NOAA) Weather Program Office (grant No. NA21OAR4590361)
• National Aeronautics and Space Administration (NASA) (grant nos. 80NSSC24K1056, 80NSSC24K0357, and 80NSSC25K7496)
• National Science Foundation (NSF) (grant no. OIA-2327435)
• U.S. Geological Survey (USGS) (grant no. G24AC00475)
• University of Oklahoma Libraries’ Open Access Fund

Citation

@article{Huang2025Urban,
  author = {Huang, Yuqi and Wang, Chenghao and Danzig, Tyler and Lee, Temple R. and Pal, Sandip},
  title = {Urban heat forecasting in small cities: evaluation of a high-resolution operational numerical weather prediction model},
  journal = {Geoscientific model development},
  year = {2025},
  doi = {10.5194/gmd-18-9237-2025},
  url = {https://doi.org/10.5194/gmd-18-9237-2025}
}