The Heterogeneous Influence Mechanism Of Urban Factors On The Urban Heat Island

With the continuous expansion of cities,urban heat island(UHI)effect was becoming more and more serious.In order to seek mitigation strategies,many statistical methods have been applied to quantitatively explore the impact of urban elements on land surface temperature(LST).However,the commonly used statistical model in previous studies was the ordinary least-squares(OLS)model,which regarded the LST as general data.As a kind of spatial data,LST is spatial autocorrelation due to its spatial attribute.Spatial autoregression model(SAM)effectively deals with this problem,but it is still a global regression model which only explores the homogeneous influence mechanism of LST.In fact,the impacts of urban elements on LST might vary across units,namely the impacts present heterogeneity.Geographically weighted regression model(GWRM)identifies local features "point by point" and excavates potential heterogeneity.However,in regions with similar LST levels,the effects of the same element on LST tend to be consistent.One the other hand,whether these studies used SAM or GWRM,they can only deal with either spatial dependence or heterogeneity.Thus,this study innovatively introduces the spatial quantile regression(SQR)model to mine the heterogeneous influence mechanism of LST under different quantiles,while considering the spatial dependence,taking the fifth-ring area of Beijing as an example.And then SQR model is applied to LST prediction.The results are as follows:(1)Seven indicators of urban elements,including underlying surface elements(water proportion(WP),grass proportion(GP),forest proportion(TP),impervious surface area except buildings proportion(ISAP))and building morphology elements(building area(BA),sidewall area(SA)and sky view factor(SVF)),pass the significance and multicollinearity test.(2)Substantial variations of the coefficients under the estimation of SQR model are found.With the increase of LST quantile,the effects of spatial dependence,ISAP and SVF on temperature gradually decrease,while the absolute estimation coefficients of WP,GP and TP gradually increase.(3)The results of mean absolute error and normalized root mean square error show that the performance of SQR in temperature prediction is better than SAM,which is proved to be a promising empirical research method to address the problem of UHI.(4)Differentiated guiding measures should be taken for areas with different intensity.For low-LST areas,special attention should be paid to the volume of construction land to curb the warming effect of buildings from the source,and then vigorously protect the blue-green space.As for high-LST areas,increasing green and protecting blue,such as green roof and blue roof,might be more effective to alleviate UHI.Then,it is suggested to strictly control the construction density,especially the height of buildings on both sides of the park and rivers.What’s more,dredging the "capillaries" of roads for "supplying" clean and cool air for urban areas is also a useful way.