How detailed should urban datasets be for accurate heat exposure modelling in cities?

Accurately capturing the spatial variability of heat exposure is important for planning heat-resilient cities. While regional climate models have historically simplified urban characteristics, grid-explicit urban morphological datasets now present an opportunity to produce spatially accurate heat maps. In this vein, this study evaluates four datasets representing Sydney, Australia in the Weather Research and Forecasting (WRF) model during the 10–20 January 2017 heatwave: 1) the default IGBP-MODIS data with no local morphology, 2) a class-based Local Climate Zones (LCZ) dataset requiring parameter interpretation by modellers, and explicitly defined parameter datasets from 3) World Settlement Footprint 3D (WSF-MB) and 4) Geoscape. The latter three used the WRF-Comfort urban canopy model, which outputs subgrid-scale Universal Thermal Climate Index (UTCI), a human thermal stress index. Comparison with observations showed using WRF-Comfort with any urban dataset over WRF defaults reduced 2 m air temperature mean absolute errors (MAEs) at peak solar hours by at least 1 °C on average, while localised instantaneous and median 2 m air temperature differences reached 13 °C and 4.5 °C across the domain. Grid-explicit experiments outperformed LCZ 2 m air temperature estimates by up to 0.37 °C (MAE) before sunrise. UTCI estimates differed substantially: LCZ estimated four times lower probability of extreme heat stress in the afternoon, and 2 °C lower average UTCI exceedances in Western Sydney. Urban dataset choice mattered most under weak synoptic conditions, though simplified datasets critically diverged in estimating localised heat stress even during strong forcing. The findings underscore the importance of grid-explicit data for identifying times and locations with highest heat exposures.