應用空間迴歸模式探討都市綠地、PM_(2.5)與地表溫度之關聯

Applying Spatial Regression Model to Explore the Relationship among Urban Green Areas, PM_(2.5) and Land Surface Temperature

張晏菁(Yen-Ching Chang) ； 徐逸祥(Yi-Shiang Shiu) ； 林峰正(Feng-Cheng Lin)

《航測及遙測學刊》 24卷1期 (2019/03) Pp. 25-43

https://doi.org/10.6574/JPRS.201903_24(1).0003

摘要

本文利用MODIS衛星影像與政府開放資料取得相關的資料變數，加入五種空間單元個別討論地表溫度與PM_(2.5)濃度的空間關係，試圖從中找出冬夏季臺中地區綠地環境、細懸浮微粒與地表溫度三者的關聯性。以一般空間模式的分析結果得到地表溫度與PM_(2.5)濃度存在一定之關係，彼此都有提升另一變數的效果。透過Moran's I等指標評比，建議以500 m空間單元探討因子與地表溫度之關係較能處理空間自相關問題，其冬夏季Moran's I值分別為0.008與0.018；細懸浮微粒PM_(2.5)則以250 m較適當，其冬夏季Moran's I值分別為-0.046與-0.032。此外，本研究發現植生指數NDVI對於PM_(2.5)濃度的正負向相關性在冬夏季節出現不同的結果，推測冬季枯水期與市地重劃的土地裸露所造成地表揚塵，是可能造成植栽滯塵原理無法有效發揮的原因。

關鍵字

空間自相關 ； 一般空間模式 ； 植栽滯塵

並列摘要

This study used MODIS satellite imagery and government open data to discuss the relationships among urban green areas, PM_(2.5) and land surface temperature in the former Taichung city in winter and summer. We demonstrated the relationships with five different grid sizes in the general spatial model. According to the results of the general spatial model, there was a certain relationship between land surface temperature and PM_(2.5) concentration and also a positive effect on each other. The Moran's I index indicates that 500 m is the most suitable grid size to demonstrate the relationships between factors and land surface temperature and deal with the influence of spatial autocorrelation; the values of Moran's I is 0.008 and 0.018 in winter and summer, respectively. As for PM_(2.5), the most suitable grid size is 250 m with the values of Moran's I -0.046 and -0.032 in winter and summer, respectively. In addition, we found that the positive and negative effects of NDVI on PM_(2.5) concentrations were different in the winter and summer seasons. It is speculated that dry riverbed during the low flow season and barren land in land consolidation area in winter caused raised dust pollution, which is the reason why the dust retention with plants could not be effectively exerted.

並列關鍵字

spatial autocorrelation ； general spatial model ； dust retention