Spectral mixture analysis of ASTER images for examining the relationship between urban thermal features and biophysical descriptors in Indianapolis, Indiana, USA

Mapping urban biophysical and thermal conditions has attracted increasing interest. However, the relationship between them has not been fully understood. This paper explores thermal features and their relationship with biophysical descriptors in an urban environment by analyzing multitemporal ASTER images. Linear spectral mixture analysis was used to unmix the five thermal infrared bands of ASTER into hot-object and cold-object fraction images and to unmix the nine visible, near-infrared, and shortwave-infrared bands into impervious surface, green vegetation, and soil fractions. Land surface temperatures (LSTs) were computed from band 13 (10.25–10.95 μm) of the ASTER. Correlation analysis was then conducted to examine the relationship between LST and the five derived fraction variables across the spatial resolution of the pixels from the ASTER images, which ranged from 15 m to 90 m. Multiple regression models were further developed to reveal how LSTs were related to urban biophysical descriptors (i.e., impervious surface, green vegetation, and soil) and to the thermal feature fractions (i.e., hot-object and cold-object). Results indicate that impervious surface was positively correlated while vegetation was negatively correlated with LST. Hot objects displayed a more significant role in influencing LST patterns than cold objects.