結合地面與空載光達推估大葉桃花心木人工林林分材積

近年來光達系統(Light Detection and Ranging, LiDAR)為興盛之森林調查技術，各項光達系統具有不同尺度範圍使用之優缺點，需透過研究目的及現地狀況，選擇合適之光達系統進行調查。本研究區域位於雲林縣古坑鄉大葉桃花心木(Swietenia macrophylla King)之平地人工林，以地面及空載光達推估林分性態值並評估其使用效率與準確度，並建立此樹種之樹高曲線式及地方材積式，最後藉由空載光達建立空中材積式推估蓄積量。研究結果顯示，地面光達對於量測胸徑、枝下高及樹冠幅準確性佳；而空載光達對於量測樹高、枝下高、樹冠幅及活冠長準確性佳，而兩種光達在使用上各有利弊，將兩者點雲合併，亦可相輔相成並抵銷彼此不足，單獨使用地面及空載光達之MAPE為26.9%及25.8%，點雲結合後MAPE下降至僅7.5%，彌補地面光達對於獲取樹高及空載光達獲取胸徑及立木位置的不足；傳統調查時間需100小時，單獨使用地面及空載光達需89及81小時，結合點雲後獲取資料的時間需92小時，提升獲取資料之效率外，資料也可重複檢視；此外，亦可透過結合點雲建立樹高曲線式(H=-0.0369D2+1.6492D-2.5349)及地方材積式(V=-0.09227+0.02D+-0.007H)。後續透過空載光達建立空中材積式並推估林分蓄積量，以10項林分性態值與傳統調查所獲之蓄積量進行逐步迴歸分析後，建立空中材積式(V=-22.86598+0.18352×N+0.66025×H+2.99902×CW)，後續透過驗證樣區進行驗證，其估計之林分蓄積量為182.93 m3/ha-1，而現地真值之林分蓄積量為176.76 m3/ha-1，其MAE=0.71 m3、RMSE=0.79 m3，印證此式可應用在相同林況之森林推估蓄積量。因應不同調查目的及樣區現況選擇合適之光達系統，結合地面及空載光達點雲可有效提高調查林分性態值之效率及準確性，還可建立樹高曲線式及地方材積式，單獨使用空載光達也可建立空中材積式，透過本研究之結果，期望能提供未來大尺度人工林蓄積量推估方法論之相關基礎資訊。

關鍵字

空載光達；森林資源調查；遙感探測；林分蓄積量；地面光達

並列摘要

Light Detection and Ranging (LiDAR) technology has become popular for forest surveys recently. Different LiDAR systems implement in different scales due to the research objectives and local conditions with appropriate LiDAR system. This study was conducted in a plantation of Swietenia macrophylla King in Guken Township, Yunlin County, Taiwan. I aimed to evaluate the accuracy and efficiency of estimating forest stand attributes by TLS and ALS. I will assess the feasibility, accuracy, and efficiency of combining point clouds and use the combined point cloud to create height-diameter model and localized volume equationsand and establish a regression model using ALS to estimate forest volume. The results showed that TLS was accurate for measuring Diameter at Breast Height (DBH), Crown Base Height (CBH), and Crown Width (CW), while ALS was accurate for measuring Tree Height (TH), CBH, CW, and Live Crown Length (LCL). Both LiDAR systems have their advantages and limitations. By merging their point clouds, they can complement each other's disadvantages. The MAPE of using TLS and ALS alone is 26.9% and 25.8%, and the MAPE drops to only 7.5% after point cloud combination, compensate their limitations of TLS in measuring TH and ALS in measuring DBH and standing position. It takes 100 hours for a traditional survey, 89 and 81 hours for ground and airborne LiDAR alone, and 92 hours for data acquisition after combining with point cloud. In addition to improving the efficiency of data acquisition, data can also be checked repeatedly. In addition, the tree height curve formula (H=-0.0369D2+1.6492D-2.5349) and the local volume formula (V=-0.09227+0.02D+-0.007H) can also be established by combining point clouds. I use traditional survey-derived volume and ALS-derived data to establish a regression model for estimating forest volume and conducting stepwise regression analysis with 10 stand characteristics. To test the equation ability, I apply this equation to conduct into another plantation as same species. The predicted volume was 182.93 m3/ha-1, while the traditional survey-derived volume was 176.76 m3/ha-1. There was no significant difference between two method values, with Mean Absolute Error (MAE)=0.71 m3 and Root Mean Square Error (RMSE)=0.79 m3. As the result demonstrated that we can confirm the applicability of this equation for estimating volume in similar forest conditions. Select the appropriate LiDAR according to different survey purposes and current situation of plot. Combining TLS and ALS point clouds can effectively enhance the efficiency and accuracy of surveying forest stand attributes. height-diameter model and localized volume equations can also be established. Furthermore, we can use airborne LiDAR alone enables the establishment of a regression model for estimating forest volume. In future, we can provide basic information about the methodology for large-scale plantation stock volume estimation.