Characterize the spatial-temporal variations of urban heat island intensity using a land use regression approach

Human activity is one of the dominated driving forces for global warming, in particularly in the urban areas. The temperature within the urban areas is higher than that of the surrounding rural area, this phenomenon is known as Urban Heat Island Effects (UHIE). In this study, a Land-use Regression (LUR) approach was applied to estimate air temperature based on ambient land-use/land cover allocations, and then to assess the spatial-temporal variability of UHIE intensity in six metropolises of Taiwan. The study materials included in-situ observations of air temperature from 2000 to 2016, landmark database, digital road network data, National land use inventory, MODIS NDVI datasets and thermal power plant distribution database. The Spearman correlation coefficient and stepwise regression were employed to develop the prediction model. Variables with the erroneous direction of correlation, high collinearity (VIF>3) and p-value>0.1 were eliminated out during the variable selection procedures. Model robustness was verified by 10-fold validation and external data verification. The results showed that, with the adjusted R² of 0.87, a 10-fold cross validated R² of 0.87, and an external data validated R² of 0.92, the high explanatory power of the resultant model was confirmed. 19 variables related to green spaces, culture activities, road, traffic and transportation, and industry were selected as important predictors variables in the developed model. Finally, UHI intensity calculated from the resultant model showed that, Taichung City had the highest level of UHI intensity (4.6?) among the six cities, and then followed by Kaohsiung City (1.8?), Taoyuan City (3.3?), New Taipei City (2.6?), Tainan City (1.3?), and the level of Taipei City (0.9?) was the lowest. Regardless the location, only minor variations were observed among the studied 17 years, indicates more efforts were needed for heat mitigation.