應用空載光達資料推測阿里山地區單木樹高與林分高度之研究

Estimating Single Trees and Stand Heights in Alishan Area with the Application of the Airborne Lidar Data

彭炳勳(Bing-Syun Peng) ； 魏浚紘(Chun-Hung Wei) ； 陳朝圳(Chaur-Tzuhn Chen)

《中華林學季刊》 42卷1期 (2009/03) Pp. 167-180

https://doi.org/10.30064/QJCF.200903.0011

摘要

本研究於阿里山地區設置樣區，針對台灣赤楊、柳杉、紅檜（天然林與人工林）等樹種進行立木樹高之量測，配合該樣區之空載光達影像資料，建立樹冠高度模型(Canopy height model, CHM)，探討空載LiDAR在森林地區對於單木樹高與林分高的量測情形；單株立木高係以樣區立木位置為中心，萃取0.5 m 圓形環框範圍內之CHM平均值為其推估值；光達林分高則以6種林分面積(5 m×5 m、10 m×10 m、12.5 m×12.5 m、15 m×15 m、17.5 m×17.5 m、20 m×20 m)萃取範圍內之CHM平均值為其推估值。單木及林分高均使用6種不同網格大小(0.5 m、1 m、2 m、3 m、4 m、5 m)所建立之LiDAR_CHM進行樹高量測。研究結果顯示，當點雲密度介於2-4 pts m^(-2)，將每網格內的高層最大值視為樹冠高度時，在單木測計上，依照不同樹種、造林性質與地勢起伏進行比較，測量的正確性易受地表坡度影響，地表越平坦正確性越高，樹種或造林性質的影響次之；若將所有樣區共492株立木合併進行比較，由於森林複層結構，空載光達對於優勢木或上層木樹高測計表現良好，下層被壓木之樹高則為高估，492株立木中，63.21%之立木樹高相對誤差在20%以內，32.72%的立木樹高誤差絕對值在2m以內；將單株立木高轉換為平均林分高進行比較時，發現當LiDAR_CHM空間解析力為3m時，以15 m×15 m的林分面積（共25個像元）進行平均值的計算，能與相同範圍的實測林分高獲取最佳之相關性(R^2=0.979)。

關鍵字

空載光達 ； 林分結構 ； 林木樹高

並列摘要

In order to analyze the airborne LiDAR (Light Detection And Ranging) data corresponding to single trees and stand heights in the forestry area, we set up samples to measure tree heights of Alnus formosana, Cryptomeria japonica, and Chamaecyparis formosensis (in nature and planted forest) in the Alishan area. We build the Canopy Height Model (CHM) based on the data from the airborne laser scanning system. Single Tree heights were estimated by averaging the CHM values within a 0.5m radius, and six different sizes of sample stands (5×5, 10×10, 12.5×12.5, 15×15, 17.5×17.5, and 20×20) were set up to extract the average heights within the area as the calculated value. LiDAR_CHM build up by six various raster cells (0.5m, 1m, 2m, 3m, 4m, and 5m) was used to measure the heights of single trees and stands. The results indicated that if we considered the highest value in each raster cell as the canopy height and the density of the point clouds lied in between 2 to 4 pts m^(-2). When measuring the single tree heights, the accuracy was easily affected by the ground gradient. The flatter the ground was, the higher the accuracy was. The species of the trees and the characteristics of the stands were the secondary factors. Due to the complexity of the stand structure, airborne laser scanning system gave a better statistics on the dominant crops in opposed to the lower level trees. Out of all 492 sample trees, 63.21% of the trees had a relative tree height error under 20%. The absolute value of the tree height error for 32.72% of the trees was less than 2m. When comparing the transferred single tree heights with the average stand heights, we found that when the spacial analysis strength of the LiDAR_CHM was 3m, and the area of the stands was 15m*15m (with a total of 15 pixels), we could get the best calculated results of the relativity with the actual survey range of the stand (R^2=0.979).

並列關鍵字

Airborne LiDAR ； Stand structure ； Tree heights