Physically-based land surface temperature modeling in urban areas using a 3-D city model and multispectral satellite data

Mitigation of urban heat islands requires understanding of the factors affecting the interaction of solar radiation and urban surfaces. In this study, we developed a novel algorithm for high-resolution spatial modeling of land surface temperature by combining the r.sun solar radiation model in GRASS GIS, the Stefan-Boltzmann Law and raster input data derived from a 3-D city model and Sentinel 2 multispectral imagery. The 3-D city model was the source of the urban land surface geometry and land cover properties, Sentinel 2 data was used to estimate broadband albedo and thermal emissivity. The developed modeling tool also implements the attenuation effect of high vegetation (transmittance) on beam solar irradiance at the ground beneath the vegetation canopy. The transmittance was estimated using the leaf area index as a proxy derived from a linear relationship between Sentinel 2 NDVI and terrestrial LiDAR. The computed land surface temperature was compared to the Landsat 8 derived land surface temperature with a good correlation expressed by a Pearson's correlation coefficient of 0.88. The proposed approach has the advantage of providing the high spatial detail coupled with the flexibility of GIS to evaluate various geometrical and land surface properties for any daytime horizon.