以衛星時序資料發展都市行道樹監測技術

Abstract

都市林可提供多樣生態系服務功能，提升市民之生活品質，若能以系統性的方法監測都市林之時空變化，應能協助增進都市林經營管理。本研究運用哨兵衛星2號遙測資料結合行道樹現地測量資訊，探討行道樹樹冠活動的年間變化與衛星影像之關聯性，以協助建立系統性監測方法。我選擇台北市種植數量較多的兩種行道樹台灣欒樹(Koelreuteria henryi)和樟樹(Cinnamomum camphora)為研究對象，運用三種不同植生指數，包括：NDVI、EVI與雙波段EVI(EVI2)時間序列資料，抽取試驗區中樹木訊號，以曲線擬合演算法將試驗區位置影像依不同植生覆蓋程度進行分類，並量化不同植生區域之植生指數時序資料，最後結合衛星資料和氣象站記錄所推估而得之試驗區地表溫度資料間進行相關性分析。 研究結果顯示以NDVI植生指數進行分類的結果有較高的準確率，台灣欒樹區域與樟樹區域分別獲得66%與72%的分類準確度、Kappa指數分別為0.39和0.58。而以EVI所建構而得的時序資料雜訊較少，適合用來捕捉行道樹一年中主要的物候事件特徵；分析結果亦發現落葉樹種的氣溫變化與年間樹冠活動之相關性較常綠樹種為高。因此，利用開源衛星資料配合地面基礎調查資料，結合氣象及植生指數時間序列資料，可協助了解行道樹之年間生長狀況，並建立系統性之都市林監測方法，提供相關單位於都市林經營規劃所需之重要資訊。

Urban forests provide multiple ecosystem services and functions to improve life qualities for city residents. Applying a systematic way to monitor the spatial and temporal changes in urban forests can help improve urban forest management. To establish a systematic monitoring method, this study uses the Sentinel-2 remote sensing data combined with in situ street tree inventory data to explore the relationship between annual tree crown dynamics and satellite imagery. I selected two of the widely-planted tree species in Taipei City, Koelreuteria elegans and Cinnamomum camphora as the research objects, and applied three different vegetation indices, including NDVI, EVI, and dual-band EVI(EVI2) time series data, to extract the tree signals from the plots. Then I classified the images with different vegetation coverages in the plots based on the curve-fitting algorithm. Following that, I quantified the vegetation index time-series data extracted from different tree-cover areas. Lastly, I explored the relationship between vegetation index and surface temperature estimated by combining satellite data and meteorological station records. Results showed that among the three vegetation indices, the classification based on NDVI had a higher accuracy, that the Koelreuteria elegans area and the Cinnamomum camphora area obtained 66% and 72% of the classification accuracy, and the kappa coefficient κ was 0.39 and 0.58, respectively. Results revealed that the vegetation index time-series data extracted by EVI had less noise, suitable for capturing the characteristics of the main phenological events in a year. The analysis results also found a higher correlation between temperature changes and the annual tree crown activities of the deciduous than those of the evergreen species. Therefore, using open-source satellite data in conjunction with the ground-based survey data, and combining with meteorological and vegetation index time-series data, can help understand the annual growth of street trees, assist in the establishment of a systematic method for urban forest monitoring, and provide critical information needed in the urban forest management and planning.