Can a mix of surface temperatures and shade help predict street-level air temperature?

Details

A simple statistical model showing how different patterns of urban development intensify urban heat islands and stress human health would be useful, but has remained elusive. Street-level air temperatures are difficult to predict from land cover and thermal data: common remote sensing data are too coarse to observe key environments beneath urban tree canopies, while intensive, location-specific ground surveys require immense labor.

The study tested whether combining remotely sensed urban heat with land cover heterogeneity and sun/shade fractions at a sufficiently fine scale for linear addition could enable simple, accurate street-scale air-temperature modeling with minimal empirical fitting. Researchers used ground-based thermography in a sample of 12 residential streetscapes in Portland, Oregon, to characterize ground-measured land surface temperatures (LSTg) of 11 common urban surface cover types in sun and shade.

Surfaces were cooler in shade than in sun, but the differences varied by surface type and were not explained by greenery or whether surfaces were permeable or impervious. Even when exposed to midday sun, surfaces on streetscapes with more canopy cover were significantly cooler than comparable sun-exposed surfaces on streets with less canopy cover. This points to the importance of partial shading during the day, which urban thermal statistical models do not typically account for.

Using high-resolution orthoimagery, the researchers measured the area of each surface cover type in each streetscape and calculated an area-weighted average surface temperature (Ts) that accounted for differences in sun and shade. Calculated Ts had a significant, nearly 1:1 relationship with sun-shielded air temperature (Ta). By contrast, relationships between Ta and tree cover, impervious area, or the LSTg of dominant surface cover types were all statistically insignificant.

The results suggest that models may more reliably connect remotely sensed urban surface temperatures to street-scale air-temperature predictions when their resolution is sufficiently high, for example less than 10 m, to avoid some known scale dependence and allow simple weighted linear models. The models should also distinguish the thermal contributions of sunlit and shaded surfaces and include the influence of shading over the day. Such models may provide effective, low-cost predictions of local urban heat islands and help inform street-level approaches to reducing urban heat.