Seasonal Variation of the Spatially Non-Stationary Association Between Land Surface Temperature and Urban Landscape

• Published in: Remote sensing (Basel, Switzerland) Vol. 11; no. 9; p. 1016
• Main Authors: Liu, Huimin, Zhan, Qingming, Gao, Sihang, Yang, Chen
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.05.2019
• Subjects: Albedo; Climate; Cooling effects; Density; Environmental impact; Goodness of fit; Heat; High temperature effects; Human-environment relationship; Indicators; Land surface temperature; Landsat; Landsat satellites; Landscape; landscape composition; LST; MGWR; Morphology; Multiscale analysis; Regression; Remote sensing; Satellite imagery; seasonal; Seasonal variations; Seasons; spatial non-stationarity; Summer; Sustainability; Temperature; Thermal environments; Time series; Urban environments; urban morphology; Urban planning; Urbanization; Vegetation cover; Winter; United Kingdom > UK; China
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs11091016

There has been a growing concern for the urbanization induced local warming, and the underlying mechanism between urban thermal environment and the driving landscape factors. However, relatively little research has simultaneously considered issues of spatial non-stationarity and seasonal variability, which are both intrinsic properties of the environmental system. In this study, the newly proposed multi-scale geographically weighted regression (MGWR) is employed to investigate the seasonal variations of the spatial non-stationary associations between land surface temperature (LST) and urban landscape indicators under different operating scales. Specifically, by taking Wuhan as a case study, Landsat-8 images were used to achieve the LSTs in summer, winter and the transitional season, respectively. Landscape composition indicators including fractional vegetation cover (FVC), albedo and water percentage (WP) and urban morphology indicators covering building density (BD), building height (BH) and building volume density (BVD) were employed as potential landscape drivers of LST. For reference, the conventional geographically weighted regression (GWR) and ordinary least squares (OLS) regression were also employed. Results revealed that MGWR outperformed GWR and OLS in terms of goodness-of-fit for all seasons. For the specific associations with LST, all six indicators exhibited evident seasonal variations, especially from the transition season to winter. FVC, albedo and BD were observed to possess great spatial non-stationarity for all seasons, while WP, BH and BD tended to influence LST globally. Overall, FVC exhibited certain positive effect in winter. The negative effect of WP was the greatest among all indicators, although it became the weakest in winter. Albedo tended to influence LST more complicatedly than simple cooling. BD, with a consistent heating effect, was testified to have a greater influence on LST than BH for all seasons. The BH-LST association tended to transfer into positive in winter, while the BVD-LST association remained negative for all seasons. The results could support the establishment of season- and site-specific mitigation strategies. Generally, this study facilitates our understanding of human-environment interaction and narrows the gap between climate research and city management.