Since its launch in February 2013, the Thermal Infrared Sensor (TIRS) on Landsat 8 has provided two channel thermal imagery of urban environments worldwide. The 100 m resolution of TIRS combined with the 30 m resolution of the Operational Land Imager (OLI) makes it possible to distinguish spatial variations in aggregate brightness temperature and subpixel land cover fractions estimated using linear spectral mixture models. In the summer of 2015, Landsat 8 began collecting nocturnal thermal imagery of 150 cities worldwide. We use day-night acquisitions of TIRS thermal imagery in conjunction with subpixel land cover fractions to investigate the relationship between land cover and diurnal variations in surface temperature in a variety of human-modified landscapes. Using two sets of diurnal thermal acquisitions and coincident land cover fraction maps in six diverse environments and climatic zones, we quantify the relationships between diurnal thermal response and land cover in different seasons for different urban typologies along a variety of urban-rural gradients. Representing land cover as subpixel fractions of soil and impervious Substrate, Vegetation and Dark (SVD) surfaces distinguishes the most functionally and physically distinct components of the land cover mosaic and provides a physical basis for quantifying the effects of vegetation, water, shadow and impervious surface on aggregate brightness temperature and diurnal thermal response. Using multi-season composites of SVD fractions provides additional constraints in the form of spectral stability because it allows spectrally stable impervious surfaces to be distinguished from temporally variable soil reflectance. In all six environments, daytime and nighttime brightness temperatures show varying degrees of correlation, depending on season, atmosphere, and land cover diversity. Although nighttime brightness temperature distributions span smaller ranges than corresponding daytime temperature distributions, nighttime temperature maps show much greater diversity and stronger correlations to land cover than daytime temperature maps. Specifically, the inverse relationship between vegetation abundance and temperature is considerably reduced at night while the retrograde cooling effect of high albedo substrates persists. On the basis of these preliminary observations, we expect that ongoing collection of nocturnal thermal imagery by Landsat 8 could contribute significantly to our understanding of the thermal environment of urban areas worldwide.
