Abstract
The thermal environment is closely related to human well-being. Determinants of surface urban heat islands (SUHIs) have been extensively studied. Nevertheless, some research fields remain blank or have conflicting findings, which need to be further addressed. Particularly, few studies focus on drivers of SUHIs in massive cities with different sizes under various contexts at large scales. Using multisource data, we explored 11 determinants of surface urban heat island intensity (SUHII) for 1449 cities in different ecological contexts throughout China in 2010, adopting the Spearman and partial correlation analysis and machine learning method. The main results were as follows: (1) Significant positive partial correlations existed between daytime SUHII and the differences in nighttime light intensity and built-up intensity between cities and their corresponding villages except in arid or semiarid western China. The differences in the enhanced vegetation index were generally partially negatively correlated with daytime and nighttime SUHII. The differences in white sky albedo were usually partially negatively correlated with nighttime SUHII. The mean air temperature was partially positively correlated with nighttime SUHII in 40% of cases. Only a few significant partial relationships existed between SUHII and urban area, total population, and differences in aerosol optical depth. The explanation rates during daytime were larger than during nighttime in 72% of cases. The largest and smallest rates occurred during summer days in humid cold northeastern China (63.84%) and in southern China (10.44%), respectively. (2) Both the daytime and nighttime SUHII could be well determined by drivers using the machine learning method. The RMSE ranged from 0.49°C to 1.54°C at a national scale. The simulation SUHII values were always significantly correlated with the actual SUHII values. The simulation accuracies were always higher during nighttime than daytime. The highest accuracies occurred in central-northern China and were lowest in western China during both daytime and nighttime.
Highlights
In 2018, 55% of the world’s population lived in urban areas and this proportion is expected to reach 68% in 2050 [1]
(2) Both the daytime and nighttime surface urban heat islands (SUHIs) intensity (SUHII) could be well determined by drivers using the machine learning method. e root-meansquare errors (RMSEs) ranged from 0.49°C to 1.54°C at a national scale. e simulation SUHII values were always significantly correlated with the actual SUHII values. e simulation accuracies were always higher during nighttime than daytime. e highest accuracies occurred in central-northern China and were lowest in western China during both daytime and nighttime
urban area size (UAS) was only significantly positively partially correlated with daytime SUHII in region II in summer (p < 0.01) and autumn (p < 0.05) and with nighttime SUHII in region I in IV 0 400 800 km Cities Ecological zoning boundaries Provincial boundaries Figure 1: Locations of the 1449 cities selected for this study in the four environmental subareas
Summary
In 2018, 55% of the world’s population lived in urban areas and this proportion is expected to reach 68% in 2050 [1]. Is effect can directly and indirectly affect regional climate [4], energy use [5], air quality [6], urban hydrology [7], soil physicochemical properties [8], creature distribution and activities [9], and human health, comfort, and quality of life [10]. UHIs are called “surface urban heat islands (SUHIs)” when derived from remote sensing data in order to distinguish them from traditional UHIs analyzed using air temperatures [15, 19,20,21]. Many determinants of SUHIs have been extensively studied and can be divided into seven categories [22]: land use and land cover types, surface biophysical conditions, landscape components and configurations, manners and intensities of human activity, meteorological conditions and geographical location, policy factors, and synthetic analyses of the abovementioned factors. Many determinants of SUHIs have been extensively studied and can be divided into seven categories [22]: land use and land cover types, surface biophysical conditions, landscape components and configurations, manners and intensities of human activity, meteorological conditions and geographical location, policy factors, and synthetic analyses of the abovementioned factors. e analytical methods applied to assess the relationship between drivers and SUHI intensity (SUHII) have mainly included Pearson’s correlation analysis, Spearman correlation analysis, comparative analysis, ordinary least squares regression, geographically weighted regression, regression tree model [16], and machine learning models [23, 24]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
