應用地理資訊系統與多變值統計分析於水源涵養保安林之規劃－以台大實驗林為例

Applications of GIS and Multivariate Statistical Analysis in Planning a Water Conservation Protected Forest-An Example of the Experimental Forest of National Taiwan University

賴晃宇(Hun-Yeu Lai) ； 鄭祈全(Chi-Chuan Cheng)

《台灣林業科學》 12卷4期 (1997/12) Pp. 421-441

https://doi.org/10.7075/TJFS.199712.0421

摘要

在森林經營上，林地分類為森林經營作業的基礎。本研究以McHarg生態規劃方法為基礎，台大實驗林為試區，以規劃水源涵養保安林之林地分類為例，採用數值地形模型萃取穀系與集水區，以做為分類單元之用；同時並應用台大森林地理資訊系統資料庫中之12種變數，包括高程、坡度梯度、地形位置、坡向多樣化指標、地形多樣化指標、剖面曲率、平面曲率、曲率、蝕溝指數、潤濕指標、坡長-坡度指數、水份梯度等，做為應用多變值統計分析於水源涵養保安林規劃時所須之影響變數來源。
統計分析上為了改進McHarg生態規劃方法的缺點，主要是採用多變值統計分析中的因素分析與群落分析二種方法，分別直接將12種變數與經由多變值統計分析之逐步判別分析所選取之8種變數，應用於水源涵養保安林之規劃。
經由比較二種不同變數的規劃結果，並將其與目前台大實驗林的經營現況和前人研究結果比較可知，在進行因素分析之前，先收集相關可能的影響變數，再經由逐判別分析來選擇影響變數，有助於規劃結果準確性的提昇。

關鍵字

林地分類 ； 數值地形模型 ； 因素分析 ； 群落分析 ； 逐步判別分析

並列摘要

Forest land classification plays an important role in forest management. This study utilized the Experimental Forest of National Taiwan University (NTU) to illustrate planning a water conservation protected forest based on the McHarg ecological planning approach. The procedure is summarized as follows: 1. Watershed generated by Digital Terrain Model (DTM) automatic extraction was used as a classification unit. 2. Twelve dependent variables for planning the water conservation protected forests from the CIS database are elevation, slope gradient, terrain position, aspect diversity, terrain position diversity index, profile curvature, plan curvature, curvature, gully index, wetness index, length-slope index, and moisture gradient. 3. Factor analysis and cluster analysis of multivariate statistical analysis were used to improve the McHarg ecological planning approaches with which the above 12 variables and S variables selected by stepwise discriminate analysis were applied in planning the water conservation protected forest.
Compared with the management circumstances of the NTU Experimental Forest and previous work, the results indicate that planning accuracy can be improved if the derived variables are selected by stepwise discriminate analysis before using factor analysis.

並列關鍵字

forest land classification ； digital terrain model ； factor analysis ； cluster analysis ； stepwise discriminate analysis