Rapid urbanization, associated with land cover change such as increase of urban developed land and concentration of urban population, has resulted in many serious ecological and environmental problems, one of which is urban heat island (UHI). UHI, which is likely to be exacerbated in future years due to the synergetic effects of urbanization and global climate change, has significantly negative social, economic and ecological impacts, adversely affects human comfort and health, and finally restricts the resilient and sustainable urban development.
Urban vegetation bas considered to be the significantly effective avenues and solutions to mitigate the extreme urban heat because of UHI. Thus, exploring the influence of urban vegetation landscape pattern on urban temperature has become a major research focus in urban heat mitigating and urban ecology. Currently, numerous studies have shown that the landscape pattern of urban vegetation (e.g. especially urban tree, which was mainly considered in this study), including composition (e.g. percent of urban trees, Ptree) and configuration (e.g. mean patch size, MPS), could significantly affect urban temperature (e.g. land surface temperature, LST). However, the inconsistent or contradictory results usually exist in previous studies, for example edge density of urban trees has negative effect on urban temperature in some studies, but positive in others. And this inconsistency and contradictoriness prevent the application of results to urban greening planning and management. Detecting in-depth the variations of the relationship between urban vegetation landscape pattern and urban temperature among different cities, and exploring the reasons for such spatial variations could provide guidance for urban greening planning and management policies for cities with different background.
In this study, based on the theory and methodology of landscape ecology, we mostly focused on four study cities, which have different climate background and development, including Beijing and Shenzhen in China, and Baltimore and Sacramento in the United States. First, the spatial heterogeneity of relationship between spatial pattern of urban trees and LST is revealed and characterized. Second, we clarified the cooling paths of Ptree and spatial configuration on LST. And finally, we defined cooling efficiency (CE) as the LST reduction if 1% of urban trees increased, and revealed the spatial heterogeneity of CE at the city scale and continental scale. And further we explored the influencing factors (e.g. climatic factors) of such spatial heterogeneity. The main results and conclusions are as follows:
1) All configuration metrics of urban trees were significantly, negatively correlated with LST, across all analytical scales, at four cities. After controlling for the effects of percent cover of trees, however, the correlations (i.e., partial correlations) between configuration metrics and LST changed greatly, in terms of magnitude, significance, and even direction. Notably, mean patch size (MPS) had significantly positive effects on LST in Baltimore, but negative effects in Sacramento. This is due to the different effects of spatial configuration of urban trees on evapotranspiration and shading in different cities, resulting in different net effects on LST. Although both the increase of Ptree and optimization of spatial configuration of urban vegetation can affect LST, their interaction could explain most of urban heat mitigating.
2) The increase of Ptree can affect LST directly, and it is the most important way to alleviate LST, at the same time, the increase of Ptree can also affect LST via changing spatial configuration of urban trees, but with limited contribution, especially in hot and rainy cities. For hot and humid cities, the indirect paths can explain some of LST, for example using complex urban tree patches while increasing urban trees could explain 13.61% of urban cooling in Baltimore. However, in hot and dry cities, LST can be effectively alleviated only by optimizing the spatial configuration of urban trees, such as increasing the mean patch size (MPS) (68.13%). However, it should be noted the interaction of MPS and edge density (ED), for example, although the increasing of MPS can effectively reduce LST, it is also necessary to weigh the decrease of ED caused by the increase of MPS and which leads to the incraese of LST.3) Ordinary least squares (OLS) model was used to quantify the relationship between Ptree and LST. And we defined the absolute value of this coefficient as cooling efficiency (CE) which could quantify the magnitude of LST reduction by one unit of urban vegetation (e.g., 1% of Ptree) increase. CE varied greatly within a city and at the continental scale:
3.1) At the city scale, daily meteorological conditions nonlinearly affected CE. Taking air temperature as an example, the increase of air temperature can increase CE, and restrict it when the air temperature exceeded about 32℃℃—The extreme higher temperature will rapidly inhibit CE and weaken the mitigating effects. Air temperature could explain more variations of LST than humidity and wind speed. Additionally, the local CE (i.e. the absolute values of local regression coefficients after using GWR) also showed significant spatial heterogeneity, and affected by the local conditions non-linearly, for example local CE decreased sharply with the increase of local Ptree, but was stable after the certain threshold of local Ptree (e.g. 20-30%), suggesting that urban greening should give priority to places where short of urban trees for the higher local CE.
3.2) At the continental scale, CE, with the average value of 0.168 and ranging from 0.040 to 0.574, was higher in southwestern cities, especially in hot and dry biomes but not significantly different among other remaining biomes, within which CE was typically greater in biomes dominated by broadleaf trees compared to those dominated by coniferous or sparse trees. Additionally, climate context affected CE non-linearly with threshold, for example increasing temperature and wind speed first improve CE, but after a certain threshold, can reduce it which was the same at the city scale. Similarly, CE could be inhibited by increasing humidity firstly, and be stable after a certain point.