Please use this identifier to cite or link to this item:
標題: 應用植物生理特徵推估新化林場大葉桃花心木面積
Application of Plant Physiological Mechanism for the Estimation of the Big-leaf Mahogany (Swietenia macrophylla) Area in Hsin-Hua Experimental Forest Station
作者: Tsai, Jeng-I
Lu, Chung-Ting
Feng, Fong-Long
關鍵字: SPOT 5 multi-spectral images;SPOT 5多光譜衛星影像;plant physiological mechanism;Swietenia macropnylla;NDVI;spectral variance;植物生理特徵;大葉桃花心木;常態化植生指標;光譜變異
出版社: 臺中市: 國立中興大學農學院實驗林管理處
Project: 林業研究季刊, Volume 34, Issue 3, Page(s) 215-226.
The main purpose of this study is using SPOT 5 multi-spectral satellite images and a land-use map to estimate the distribution of big-leaf mahogany (Swietenia macropnylla) in Hsin-Hua Experimental Forest Station. Mahogany has a specific mechanism, which was completely deciduous and germinates during a short period around March or April each year. This mechanism made the spectral variance. We collected two periods SPOT 5 images and obtained status and change of Mahogany with normalized difference vegetation index (NDVI) and image differencing algorithm. The training areas were chosen based on land-use map to set the ranges of threshold. We compared the classification results with ground truth (temporal and permanent sample plots) to determine the accuracy assessment. The results showed that it was not suitable to estimate the Mahogany area with SPOT images only (the overall accuracy was 63.58 % and Kappa coefficient was 0.28). After we removed the non-vegetation and no mahogany areas accorded to the landuse map, the overall accuracy increased to 85.87 % and the Kappa coefficient became 0.71. The reason might be the change of environmental factors and other physiological characteristics made the spectral variance either. The results also showed that the stand neighbor could reduce the overall accuracy, and the method was could not be used in the lower density stand. If we could realize the periods between leave falling and sprouting clearly, the method helps us to estimate the areas of mahogany in pure and mixed forest in very short time.

本研究主要目的為利用SPOT 5多光譜衛星影像與土地利用型圖,推估新化林場大葉桃花心木(簡稱大桃)面積。每年3月底至4月初,大桃有在極短時間內完全落葉,並在1至2星期內快速發芽的植物生理特徵,此物候現象造成影像光譜產生明顯變異,故研究計算兩期SPOT 5衛星影像的常態化植生指標(normalized difference vegetation index, NDVI)值與進行影像差異演算法,得到現況與變遷資料,以2009年的土地利用型為依據設定大桃門檻值,推估大桃分布與面積。為確認分類結果適合性,研究以永久樣區(36個)與臨時樣區(148個)做地真,計算整體分類準確度與Kappa統計值。結果顯示,單以多時期SPOT 5影像計算大桃面積,會有高估情形產生,整體分類準確度約63.58%(Kappa值0.28),可能係外在環境或其他植物生理特徵影響光譜,故研究將土地利用型圖納入分析,將非植生或無大桃區域去除後,分類準確度提升至85.87%(Kappa值0.71),屬分類良好的結果。研究也發現,林分邊界產生的混淆像元,以及林分密度太低區域會降低分類結果。如果使用者可以準確掌握大桃的落葉與萌芽時間,配合此法可在短時間內,將大面積的大桃於混淆林中選取出來,減少調查所需之人力與物力。
Appears in Collections:第34卷 第03期

Files in This Item:
File SizeFormat
151374-4.pdf1.04 MBAdobe PDFView/Open
Show full item record
TAIR Related Article

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.