應用光達資料於溪頭柳杉人工林分調查之研究

A study on investigating Cryptomeria japonica stand in Chitou using LiDAR data

張瑞文(Rui-Wen Chang)

指導教授 ： 邱祈榮

國立臺灣大學/生物資源暨農學院/森林環境暨資源學研究所/碩士(2007年)

https://doi.org/10.6342/NTU.2007.02868

摘要

在大面積的森林資源調查中，採用遙測（分為多光譜、多解像度、多時序性遙測影像）之調查方法因具有可大幅度減少人力經費等物資的消耗之優點，因而近年來多為各森林調查所採用。本研究採用工研院能資所提供的台灣大學實驗林溪頭營林區柳杉實驗林的光達影像作為研究材料，探討使用原始光達資料利用於森林資源調查之可行性。研究內容分為兩大部分，首先是光達資料的準確性評估。基於不同栽植密度的15個研究樣區的特性，以不同搜尋面積的局部最大值演算法、樹冠層分離法、光達體積參數的統計方式比較各樣區內單株立木的位置（水平誤差）及樹高（垂直誤差）之準確性、樹冠面積的篩選與分層。基於三種篩選原則，固定光達林分高界限值、林分高圖層相對斜率變化、光達百分比分層體積。初步研究結果確立光達圖層的準確度與誤差範圍後，進入樣區為基準的材積估算。採取柳杉材積通式與迴歸樹高式比較我們迴歸的結果。研究結果顯示光達是一個準確性高的高程量測工具，局部最大值的個數與樣區株數相關性有0.80，且使用光達分層體積、林分百分位高、局部最大值個數應用於林分中估算樣區材積的相關性達0.72。顯示使用光達資訊估算樣區林分材積有不錯的結果，光達百分位林分高與光達百分比分層體積是不可或缺的樣區材積估算參數。光達材積式應用於估算樣區材積如有樣區樹高、栽植密度分佈均勻林分的特性可得較準確的估算結果。

關鍵字

光達林分高 ； 局部最大值演算法 ； 光達百分位林分高 ； 光達林分體積

並列摘要

Recent researches have concluded that it is better to use remote sensing data to gain a time and effort-consuming result when applying in a large scale forest canopy monitoring. The LiDAR data is one of the remote sensing data that can be classified as multi-spectral, multi-scale and multi-temporal. The LiDAR system has many advantages, allowing quick access to large scale 3-dimensional information that is not available from field investigations. In this study, we used the LiDAR data provided by Industrial Technology Research Institute (ITRI) to understand its effects and abilities when applied to the NTU Chi-Tou experimental forest. To understand the accuracy of LiDAR data, field investigations are still required for calibration. The primary results in horizontal and vertical errors are rather good, and the vertical residuals caused by the inaccuracy of the measuring equipment resulted in the majority of the errors in this study.
When applying to the different plantation density of the 15 sampling plots, we used different searching ranges in the LiDAR local maximum algorithm, which are also the basis in the separation of tree crowns. In the scale of plot, to understand the stem volume concerning the volumes of LiDAR DCHM, we used single threshold value, relative slope variation and classified percentile volumes in DCHM to subtract layers of pixels within a plot. The argument will be more persuasive when some background data such as field investigation and canopy mean height is served.
To sum up, using LiDAR data is a powerful tool to obtain the multi-layer forest vertical structure and to gain a large-scale canopy height. In this study, we can use LiDAR local maximum number to predict number of trees in a plot, and predict canopy volume using LiDAR local maximum number, LiDAR CHM percentile and classified LiDAR volume, with the former correlation is up to 0.8 and the later 0.72.

並列關鍵字

LiDAR digital canopy height model ； local maximum algorithm ； LiDAR DCHM percentile ； LiDAR volume

參考文獻

林國銓、洪富文、游漢明、馬復京，1994。福山試驗林林生態系生物量與葉面積指數的累積與分佈。林業試驗所研究報告季刊 9(4): p299-315.

林務局，1997。台灣林產物處分調查用立木材積表。行政院農業委員會林務局。台北。

鄭祈全、邱祈榮、陳燕章，1997。應用遙測方法估測台灣杉林分之葉面積指數。台灣林業科學 12(3): p309-317.

LiDAR data analysis. 3D Remote Sensing in Forestry.

Daniel A. Zimble, David L. Evans, Geroge C. Carlson, Robert C. Parker, Stephen C. Grado, Patric D. Gerard, 2003. Characterizing vertical forest structure using small-footprint airborne Lidar. Remote Sensing of Environment 87, p171-182.