Long-term Assessment Of Urban Thermal Environmental Effects In Hefei

With the rapid development of urbanization,the problem of urban heat island effect has became increasingly serious,which brings about an adverse impact to urban ecological environments and residents'health,undermining the momentum of sustainable urban development.The traditional meterological station based observations are not liable to create reliable and spatially explicit patterns of urban heat island effects(UHIE)due to the limited representativeness of the discrete stations.Remote sensing data,with its advantages of high timeliness,wide coverage,easy access and high revisit frequency,has become an important means in the dynamic monitoring of UHIE.Based on long time sereis Landsat and MODIS acquisitions spanning from 2000 to 2020,the spatial distribution characteristics and evolution trend of thermal environment in Hefei were quantified by using STARFM(Spatial and Temporal Adaptive Reflection Fusion Model)and ESTARFM(Enhanced Spatial and Temporal Adaptive Reflection Fusion Model)algorithms.Next,the land surface temperature(LST)and related factors of urban heat island(UHI)in each year were extracted to analyze the relationship between the driving factors and the high temperature thermal environment of UHI.Afterwards,based on the object-oriented method,the land cover classifications in Hefei were made,and the spatial associations between land cover types and urban heat island intensity levels were modelled from multivariant stepwise regression and random forest algorithm.Ultimately,using a CA-Markov model,the distribution characteristics of future high-temperature thermal environment and land cover types in Hefei were simulated,followed by a validation of the simulated distribution using the 2020 real patterns and a projection of the thermal environment of Hefei in 2030.The main findings were as follows:(1)The spatial-temporal fusion algorithms of STARFM and ESTARFM were both acceptable in terms of land surface temperature fusion,but STARFM's performance was slightly better than ESTARFM in terms of multiple evaluation indexes.Based on an independent real image's validation,it was revealed that STARFM had a R~2 value at 0.6005,with a SD(Standard Deviation)at 2.6589,an AAD(Average Absolute Deviation)at 1.6580,a MSE(Mean Squared Error)at1.9309 and a SSIM(Structural Similarity Index)at 0.9792,whereas ESTARFM algorithm obtained a R~2 of 0.5855,a SD at 2.3748,an AAD at 1.9886,a MSE at 2.2879 and a SSIM at 0.9377.(2)The stepwise regression model and random forest model were selected to first identify those most important related factors then model the statistical relationships between land surface temperature and the identified factors.In terms of model fitting and verification accuracy,random forest model was superior to the stepwise regression model.The fitting R~2 of random forest was at0.7433,higher than that of the stepwise regression at 0.5173,and the verification R~2 of RF was0.6957,also higher than that of the stepwise regression at 0.4116.Moreover,other indicators including the mean absolute error and root mean square error also confirmed this comparison relationship.In stepwise regression modeling,NDBI,DEM,NDSI and NBR were identified as important factors.In random forest modeling,DVI,NBR,B1(blue band),B3(red band),B4(near infrared band),TC?G(tasseled cap transform-green),NDVI,TC?B(tasseled cap transform-brightness),NDBI and NDSI were identified as modeling factors.(3)From 2000 to 2020,the UHIE in Hefei showed an increasing trend.The expansion of the high temperature region was very striking,rising from 638.8092 km~2 in 2000 to 886.3101km~2 in2020.The area of the middle and sub-high temperature regions decreased and increased year by year,with a total decrease of 520.1578km~2 and an increase of 243.3762km~2.The areal change in the sub-low temperature and low temperature regions was not obvious.(4)Based on the heat island grading map and land cover map in 2010,a CA-Markov model was used to simulate the spatial distribution of heat environment and land cover type in Hefei in2020,and the simulation accuracy was validated according to the actual heat island intensity grading map and land cover type map in 2020.The corresponding overall Kappa coefficients of the simulated heat island intensity grading map and land cover map were at 0.6751 and 0.8827,respectively.The CA-Markov model was further applied to predict the distribution of heat island in Heifei in 2030.The prediction results showed that the high temperature area and urban area of heat island in Hefei would continue to increase during 2020-2030,and the UHIE would continue to rise.Some measures can be taken,such as rational planning of urban layout,reducing anthropogenic heat emission,increasing urban greening and constructing sponge city.