关键词: 空间自回归模型, 空间自相关, 城市热岛, 局部气候区, 热环境

Abstract:

Urban heat island (UHI), defined as a human-induced urban climate phenomena characterized by higher temperatures in urban areas than in their surrounding rural areas, is widely acknowledged as one of the most significant urban environmental problems caused by urbanization. Urbanization has significantly transformed natural and semi-natural surfaces into impervious urban structures, which has disturbed the balance of the Earth's surface radiation and energy, as well as the composition of the atmospheric structure in the near-surface. Many studies have explored the complex mechanisms of urban heat islands by examining the relationship between land surface temperature and land surface factors. Few, however, have explored the relative contributions of impact factors to urban thermal environment using spatial statistical analysis. In addition, referring to the heterogeneity of urban land surface characteristics, the relationship between urban thermal environment and the related land surface factors is expected to be different within urban area.In order to explore the influence mechanism of related land surface factors on urban heat island intensity at various locations, Xiamen city is taken as an example. Based on remote sensing images and building census data, the concept of Local Climate Zone (LCZ) and spatial autocorrelation concept are applied to analyze the spatial response rules between related surface factors and heat island intensity in all LCZs. Spatial autoregressive model is used to quantitatively analyze the spatial relationship between heat island intensity and related land surface factors in LCZs of Xiamen. The results show that: 1) Heat island intensity in the study area has significant positive spatial correlation and obvious spatial agglomeration. The high value area concentrates on the construction land, cultivated land and bare land in the east and south, while the low value area concentrates on lakes, rivers and other water bodies, wetlands and forest in the north and northwest. 2) Ordinary linear regression model(OLS) can not effectively explain the relationship between related surface factors and heat island intensity in space. The spatial lag model(SLM) and spatial error model(SEM) results are compared with the OLS, which shows the best performance of the SLM and SEM model. By comparing the R ², AIC, SC, and LIK values, it can be seen that the SEM model has better fitting effect than spatial lag model, which can more scientifically analyze the spatial relationship between surface factors and heat island intensity. 3) In the SEM Model, lambda is always positive and significant in all LCZs, indicating strong spatial dependence of model error. In the SLM Model, the positive and significant spatial autoregressive coefficients indicate an active influence from neighboring regions. 4) The land surface factors which can be independent variables of the regression model are different in each LCZ. In all LCZs, as independent variables of the regression model, vegetation index, water index and sky view factor show a negative correlation with urban heat island intensity, while building density and proportion of impervious water surface show a positive correlation with urban heat island intensity. However, the correlation between building volume density, average building height, building height difference and heat island intensity is not consistent in various LCZs. To a certain degree, the mathematical relationship between urban heat island intensity and land surface factors is discussed. According to the research conclusions, it is suggested to protect the "compensation area" and separate the "function area", optimize the land use structure. In addition, it is necessary to consider the difficulty of adjusting land surface factors, cooling efficiency and population density to determine feasible planning strategies.

Key words: spatial autoregressive model, spatial autocorrelation, heat island, local climate zone, thermal environment