天山北坡城市群季节性地表温度与景观格局空间关系分析

The spatial relationship between seasonal surface temperature and landscape pattern of the urban agglomeration on the northern slope of the Tianshan Mountains

探究景观格局与干旱区地表温度(LST)的空间关系,对于促进干旱区生态环境稳定发展具有重要的意义。基于Google Earth Engine(GEE)解析天山北坡城市群2003—2020年昼夜及季节性(白天)LST时空分布特征,使用Mann-Kendall非参数检验、Sen’s斜率并结合Hurst指数探究天山北坡城市群LST变化趋势并预测未来发展方向,运用标准差椭圆与重心迁移模型分析LST空间迁移特征,采用双变量空间自相关与多尺度地理加权回归(MGWR)模型分析LST与景观格局的空间关系。结果表明:2003—2020年天山北坡城市平均LST白天23℃,夜间-0.5℃,昼夜温差大,温度的季节性变化特征显著。LST未来变化趋势表明,夜间温度上升速度将高于白天上升速度,即冷岛强度白天高于夜间的现象将有所缓解。代表建设用地和植被的中温和低温,其温度重心在乌鲁木齐市。斑块密度(PD)、边缘密度(ED)和平均形状指数(SHAPE_MN)与LST呈空间负相关关系,而最大斑块指数(LPI)、平均斑块面积(AREA_MN)、聚集度指数(AI)和景观类型比例(PLAND)与LST呈空间正相关关系,表明斑块聚集且连续的景观LST越高。比较OLS、GWR和MGWR模型,MGWR模型的拟合效果最好,ED和AI对于LST的影响在研究区的右侧最显著,而PLAND对于LST的影响在研究区的中部南侧最为显著,其他景观格局指数影响强度较为缓和。

Exploring the spatial relationship between landscape pattern and land surface temperature (LST) in arid regions is of great significance for promoting the stable development of ecological environment in arid regions. Based on Google Earth Engine (GEE), we examine the temporal and spatial distribution characteristics of day and night and seasonal (daytime) LST in the urban agglomeration on the northern slope of the Tianshan Mountains from 2003 to 2020, and use the Mann-Kendall nonparametric test and Sen's slope and Hurst index to explore the change trend of LST and predict future development direction, adopt standard deviation ellipse and gravity center migration model to analyze the spatial migration characteristics of LST, apply bivariate spatial autocorrelation and multi-scale geographically weighted regression (MGWR) model to identify the spatial relationship between LST and landscape pattern. The results show that from 2003 to 2020, the average LST of the study area is 23℃ during the day and -0.5℃ at night. The future change trend of LST shows that the temperature rise rate at night will be higher than that in daytime, that is, the phenomenon that the intensity of cold island is higher during the day than at night will be alleviated. The center of gravity of the medium and low temperatures, which represent construction land and vegetation, is observed in Urumqi. Patch Density (PD), Edge Density (ED), and Mean Shape Index (SHAPE_MN) were spatially negatively correlated with LST, while Largest Path Index (LPI), Mean Patch Size (AREA_MN), Agglomeration Index (AI), and Percent of Landscape (PLAND) were spatially positively correlated with LST, indicating that patches of aggregated and continuous landscapes had higher LST. Comparing the OLS, GWR and MGWR models, we found that the MGWR model has the best fitting effect, the effects of ED and AI on the LST are the most significant on the right side of the study area, and the effect of PLAND on the LST is the most significant in the central-south part of the study area, and other landscape pattern indices have a moderate impact intensity.