空載光達對立木樹冠及林分競爭解釋能力之探討

Study on the Explainable Ability by Using Airborne Lidar in Tree Canopy and Stand Competition

葉日嫈(Jih-Ying Yeh)

指導教授 ： 陳朝圳

國立屏東科技大學/農學院/森林系所/碩士(2015年)

https://doi.org/10.6346/NPUST.2015.00167

摘要

森林冠層結構是由林地上相同或不同的樹種組成，森林冠層則為影響陽光到達林間之主要因子，而林間光度大小則影響林木是否產生競爭，進而影響林木生長與樹冠特徵。疏伐撫育為控制森林冠層密度的重要作業法，而疏伐撫育作業必須先掌握林分之森林冠層與競爭資訊。近年來所發展的空載光達科技，在森林資源調查已發展為具有三度空間資料收集的標準作業，其對於森林冠層特徵的掌握，不僅止於平面資訊，已可收集三維立體之空間分布，具有探討水平及垂直林分結構的特性，對於森林冠層結構之研究更具發展潛力。本研究以溪頭營林區65年生之不同栽植距離柳杉人工林為研究範圍，透過空載光達資料，建構樹冠高度模型、三維像元分層法、多重回波與點雲強度值，計算光達特徵值，配合現地調查資料，探討推估林分性態值之準確度；並藉由不同尺度與不同林分鬱閉度，探討光達特徵值與林分性態值之相關性。根據三維像元分層法之點雲頻度分布圖萃取冠層厚度，並透過單因子變異數分析，瞭解其於不同栽植距離之差異性。藉由地真資料計算林分競爭指數，以多元迴歸分析法，推估光達之林分競爭指標。透過多重解析分割法進行三種不同分割尺度萃取樹冠輪廓，並依據局部最大值偵測立木位置，分別評估樹冠面積大小和精確度，以瞭解空載光達於三種不同分割尺度萃取樹冠輪廓與立木位置之可行性。研究結果顯示，空載光達推估之林分性態值，以林分高之精度最高，其平均精度為91.03%。而光達特徵值會隨著空間尺度增加而提升其相關性，以20 m × 20 m對於林分性態值之相關性最佳，但較高的林分鬱閉度會降低光達特徵值推估林分性態值之準確度。林木栽植距離顯著影響冠層厚度和林分競爭程度，而透過光達特徵值與林分競爭指數，推導所獲得冠層厚度、點雲強度值、林分枝下高、林分鬱閉度與單一回波比例等五項光達指標呈顯著相關，相關係數為0.81。在栽植距離小的情況下，透過最小分割尺度使樹冠輪廓之偵測較為精確，但整體仍呈現低估情形。當栽植距離較小時，最適合偵測立木之分割尺度為5 m × 5 m，而分割尺度為7 m × 7 m最適合偵測栽植距離大的立木。於溪頭試驗樣區中，因林分過於鬱閉，導致光達在推估上產生些許誤差，但整體結果顯示，空載光達資料確實可以推估林木生長與林分結構之相關資訊，其對於立木樹冠與林分競爭之解釋有所助益。

關鍵字

空載光達 ； 光達特徵值 ； 林分性態值 ； 林分結構

並列摘要

Forest canopy structure is composed by the various species. Sun light is a main factor to affect the crown structures after tree competition. However, thinning operation is an appropriate way to control canopy density, which can adjust the competition conditions in the different crown structures. Recently, Airborne Laser Scanning (ALS), also referred to as Light Detection and Ranging (LiDAR), has been established as a standard technology for high precision three dimensional forest data acquisition; it could get stand characteristics with three-dimensional information that had develop potential for the structure characteristics of forest canopy. The 65 years old, different planting density of Cryptomeria japonica experiment area was selected for this study in Chitou area. We used LiDAR image to estimate LiDAR characteristic values by constructed CHM, voxel-based LiDAR, multiple echoes, and assess the accuracy of stand characteristics with intensity values and field data; moreover, to explore the correlation between the LiDAR characteristic values and stand characteristics via different spatial scales and different canopy density. Furthermore, extracting canopy thickness with point cloud frequency distribution of voxel-based LiDAR, and discuss the vertical structure of the stand through variance in the different planting distances by one-way ANOVA analysis. The competition index was calculated with field data, and estimate competition index of LiDAR via multiple linear regression. The tree crown extraction by three different scales through multi resolution segmentation method, and used local maxima to detect individual tree location. Respectively assess the size of crown and accuracy, in order to understand the feasibility of extraction with crown profile and individual tree location of the LiDAR in three different segmentation scale. The results showed that the highest accuracy with stand characteristics was stand high which estimate by LiDAR, its average accuracy of 91.03%. LiDAR raster grid size was 20 m × 20 m for the correlation was the best, however, the higher canopy density will reduce the accuracy of the LiDAR characteristic values to estimate the stand characteristics. The significantly affect canopy thickness and the degree of competition in different planting distances. The canopy thickness, intensity of the point cloud, branch height, canopy density, and ratio of only echo of the five LiDAR indicators was obtained by competition index, that showed a high correlation (R2=0.81). Through the minimum segmentation scale was more accurate to extract tree crown in the small planting distance, but the overall situation was still showing underestimated. When segmentation scale was 5 m × 5 m for the most suitable of the detection individual tree location in small planting distance, and the segmentation scale of the 7 m × 7 m was optimal in the large planting distance. The result found stand canopy too closure in Chitou field, lead to the estimate of LiDAR caused few error. However, the LiDAR characteristic can estimate the information on the tree growth and stand structure, and profit the interpretation of the tree canopy and stand competition.

並列關鍵字

Airborne laser scanning ； LiDAR characteristic values ； Stand characteristics ； Stand structure

參考文獻

林莉萍 (2013) 應用空載光達資料估計森林樹冠高度模型及葉面積指數。國立成功大學測量及空間資訊學系碩士論文。69頁。

莊棨州 (2014) 溪頭柳杉林昆蟲垂直分層群聚研究。國立臺灣大學昆蟲學研究所碩士論文。101頁。

羅時凡、陳淯婷、沙學均、陳朝圳 (2008) 林木健康指標之建立與效度評估。林業研究季刊 30(3): 41-52。

彭炳勳、謝依達、陳朝圳 (2008) 空載光達雷射穿透率指數與柳杉林葉面積指數之關係探討。臺灣林業科學23: 53-63。

魏浚紘、陳朝圳、彭炳勳、李崇誠、陳建璋 (2013) 地面光達應用於林木競爭之研究。航測及遙測學刊 16(4): 279-287。