Day and night synergy to improve subpixel urban impervious surface mapping in desert environments at 30-m Landsat resolution

Abstract

ABSTRACT Few studies provide effective guidance for detecting subpixel urban impervious surface in desert environments. Such an environmental setting is substantially different from the study areas of most existing studies, especially with complicated desert landforms. To improve the accuracy of subpixel mapping of urban impervious surface in desert environments, an integrative approach is proposed to estimate fractional urban impervious surface at the 30-m resolution. This is done by the synergistic use of open-source datasets taken from day and night that characterize different aspects of human activities at a different time of a day. We carefully analyse the performance of three methods with model input combinations from different sources, i.e. the day and night image synergy, daytime images only, and georectified night-time International Space Station (ISS) photographs only. Three major findings are concluded from the analyses. First, the collective usage of remote sensing images derived from day and night yields reasonable results in all desert cities and outperforms any single-source data for subpixel urban impervious surface mapping. Second, among all the input features from multiple data sources, night light variables have a higher contribution than other variables of daytime images for subpixel urban impervious surface mapping in desert cities, regardless of lighting types. Third, since desert cities remain understudied in previous studies, the proposed synergy method fills this gap in the literature. This data fusion of multi-source remotely sensed data is promising, which can be employed for better subpixel urban mapping in support of various environmental and socioeconomic applications in the future.