台灣地區雖然擁有豐富之森林資源,其森林覆蓋約58.5%,其中人工林約有295,500 ha,佔森林覆蓋面積的20%。由於人工林部份區域未施行中後期撫育作業,導致林分生長競爭激烈;或因生育地的不適應及天然災害等因素,形成林木生長不良之生育地,影響人工林的正常發展。如何利用遙測影像的光譜特性,進行大面積且快速的掌握人工林林分的健康程度為本文之主要研究目的。由於林木葉片葉綠素含量及水分含量高低,會影響林木光合作用速率,而林木生育環境逆壓及林木生長狀態的良好與否,亦會反映在林木葉片之葉綠素含量及水分含量的高低;本研究欲應用遙測影像光譜對葉綠素含量與水分含量的偵測,進而建構林分健康指標。本研究應用2006年MODIS-Terra衛星影像,利用光譜特性,計算可評估林木健康狀態之葉綠素含量及水分含量之植生指標,包括常態化差異植生指標(NDVI)、重新常態化差異植生指標(RDVI)、差異植生指標(DVI)、修正簡單比例指標(MSR)、比例植生指標(RVI)、常態化差異水分指標(NDWI)及總體植生水分指標(GVMI)。觀測其季節變化,結果顯示以冬季影像(2006/01/15)對於檢測植生狀態之差異性最為敏感。以2006/01/15影像進行主成份分析給予各指標權重,得到森林健康指標線性組合。應用地面樣區資料與森林健康指標進行準確度評估,其總體準確度達71.42%,Kappa係數為0.56,驗證以MODIS指標,透過主成份分析所組成之森林健康指標,可反映森林範圍內健康與非健康之林木分布情形。而不同時期森林健康指標推估模式之穩定度試驗,其結果顯示兩期影像呈現極顯著相關性。建議未來若應用植生指標進行森林健康監測,可以NDVI、MSR、RVI及NDWI配合其他相關變數進行監測。
Although there is abundant in forest resource in Taiwan, and the ration of forest coverage is above approximately 58.5%, which has 295,500 ha of artificial forest and possess 20% of total forest area, but some part of the artificial forest without execute the tending, or non-adaptation of the habitat, if this condition goes on, and coincides with any other natural calamity, then the stand will occur competition drastically. Finally, the whole forest land may go downhill, and influence health of the timber. How to use spectrum character of satellite data to know well with health condition of large area of artificial forest that is the objective of the paper. Because of chlorophyll content and canopy water content which could be influent the speed of photosynthesis. The environment stress and growth condition of timber would reflect on variation of leaf chlorophyll content and canopy water content. In this paper, we will use chlorophyll content and canopy water content monitoring of satellite data to construct stand health index. We use MODIS-Terra image data to obtain Normalized Difference Vegetation Index (NDVI), Renormalized Difference Vegetation Index (RDVI), Difference Vegetation Index (DVI), Modified Simple Ratio (MSR), Ratio Vegetation index (RVI), Normalized Difference Water Index (NDWI) and Global Vegetation Moisture Index (GVMI) which can indicate healthy condition of vegetation. Observe that seasonal variation of these indices, the result shows that the image of winter season (2006/01/15) is more sensitive than others for monitoring the variation of vegetation. We use Principal Components Analysis to give scores for each index that in order to obtain the linear formula of forest health index. Validation of the ground truth data and forest health indices resulted overall classification accuracy are 71.42%, and overall kappa statistics are 0.56. We use different periods of MODIS image proceed stability experimentation. The result showed high relationship between the two periods of forest health indices which indicated that the forest health index was stable to reflect vegetation health condition. The result of this paper propose that applications of vegetation index could use NDVI, MSR, NDWI and GVMI combine with other correlation variable for forest health monitoring.