基于空间自回归模型的广州市NDVI和NDBI与气温关系研究

Relationship of air temperature to NDVI and NDBI in Guangzhou City using spatial autoregressive model

为了探索城镇化地区热岛的时空变化特征,采用2015年覆盖广州市的1 km空间分辨率MOD13A3月合成归一化植被指数(normalized difference vegetation index,NDVI)数据、用Landsat8 OLI提取的归一化建筑指数(normalized difference build-up index,NDBI)数据以及不同季节的气象站点近地表气温数据,运用相关性分析方法,研究近地表气温与NDVI和NDBI的相互关系;在此基础上,应用空间自回归方法构建不同季节的近地表气温与NDVI和NDBI的空间自回归模型,定量分析广州地区近地表气温与NDVI和NDBI的空间关系,并与普通回归模型进行比较分析。结果表明,不同季节的NDVI与近地表气温呈负相关,NDBI与近地表气温呈正相关;与普通线性回归模型相比,空间滞后模型与空间误差模型的拟合效果最优;通过比较分析相关系数(R^2)值、赤池信息量准则(Akaike information criterion,AIC)值及回归模型残差的莫兰指数(Moran index,Moran’s I),发现空间滞后模型的拟合效果略优于空间误差模型;从春季到秋季,NDVI对近地表气温的影响大于NDBI对近地表气温的影响;在空间滞后模型中,显著的、正的空间自回归系数表明,气象站点的近地表气温受到相邻气象站点的近地表气温的显著正影响。

To study the spatio-temporal pattern of the air temperature in Guangzhou City,the authors used MODIS monthly normalized difference vegetation index（ NDVI） acquired in 2015 and extracted the normalized difference built-up index（ NDBI） with Landsat8 OLI data. The correlation analysis method was used to explore the relationship between air temperature and NDVI,NDBI. The experimental results show that there is a negative relation between NDVI and air temperature and a positive relation between NDBI and air temperature. On such a basis,the spatial lag model（ SLM） and spatial error model（ SEM） were established to discuss the spatial relations between air temperature and NDVI,NDBI in different seasons,respectively. The SLM and SEM results were compared with the ordinary least square regression（ OLS） model,which shows the best performance of the SLM and SEM models. The SLM model with higher R^2 and lower AIC values performs slightly better than the SEM model. NDVI has more influence on air temperature from spring to autumn than NDBI. In the SLM model,the positive and significant spatial autoregressive coefficients indicate an active influence from neighboring meteorological stations.