利用自動化雨滴譜儀所得到之北台灣降雨沖蝕指數進行其空間分佈推估之研究

Abstract

台灣因為特殊的水文和地文條件以及人為的活動使得土壤沖蝕問題嚴重。降雨沖蝕指數(Rainfall Erosion Index)是影響土壤沖蝕的一個重要因子。由於現有的雨量觀測站數目有限，且降雨觀測期長短不一；因此，對於沒有設立雨量觀測站或是沒有雨量資料之處，實有必要針對其降雨沖蝕指數進行空間分佈之推估。 本研究以北台灣地區為研究區域，對無觀測站地區的降雨沖蝕指數進行推估之研究。所使用之降雨沖蝕指數是由自動化雨滴譜儀之觀測資料分析所得。而空間分佈推估之方法有：(1)內插法(Interpolation Method):由等降雨沖蝕指數圖求得，或由估算地點附近三個已知地點之值以反距離加權法(Inverse-Distance Weighting Method)進行計算，這是台灣目前水土保持技術規範之做法；(2) 克利金法:以研究區域全部雨量觀測站資料透過克利金法進行推估，這是近年來國內外漸多使用且比方法一較為準確之方法；(3) 均一區劃分法和克利金法合併使用:此方法先進行降雨沖蝕特性均一區劃分，以同區之測站資料透過克利金法進行推估該區域內的降雨沖蝕指數，此為本研究所提出和使用之方法。本研究透過方法二和方法三來探討降雨沖蝕特性分區對於降雨沖蝕指數推估之影響。 利用方法三進行分析時，是先經由主成分分析(主成分包括經緯度、高程、降雨沖蝕指數及其變異係數)後再進行兩階群集分析，在保有各個群集內之空間相依性之同時，可將台灣北部各個站點大致劃分出明顯的區域。再對各降雨沖蝕特性分區以克利金法進行無觀測站地區之降雨沖蝕指數推估。由交叉驗證法驗證之結果顯示，進行降雨沖蝕特性均一區劃分之推估準確度比未進行分區更為良好，亦即方法三較方法二為優。

Soil erosion has been a serious problem in Taiwan due to special hydrological and physiographic conditions, and human activities. Rainfall erosion index (the R factor) is an important factor for soil erosion. Because of the limited number of rainfall observation stations and the different lengths of rainfall observation period, it is necessary to conduct spatial estimation of the R for the areas without rainfall observation station or lack of rainfall data. To accomplish this goal, the northern Taiwan was selected as the study area. The R values used in this study was obtained from the observed data using distrometer. There are three methods for spatial estimation. The first is interpolation method: isoerodent map method or inverse-distance weighting method computed using the data of the three near rainfall observation stations. This method is currently applied to the technical specification of soil and water conservation in Taiwan. The second is Kriging method: the observed rainfall data in the whole study area are used to estimate the R value using the Kriging method. Compared to Method 1, Method 2 is more accurate and more often used in recent years. The third is delimiting homogeneous region method accompanied with the Kriging method: firstly, delimit homogeneous regions of rainfall erosion properties, then using the Kriging method with each homogeneous region’s rainfall observation station data to estimate the R vales in each homogeneous region. This method is proposed and utilized in this study. By comparing Method 2 with Method 3, the effects of delimiting homogeneous regions of rainfall erosion properties on the R estimation in the areas without observed rainfall data are investigated in this study. While Method 3 is used for analysis, principal components analysis(Principal components include longitude, latitude, elevation, rainfall erosion index and its coefficient of variation.) is firstly conducted and two step cluster analysis is then carried out. Through the procedures, the northern Taiwan could be clearly delimited, while each region maintain its spatial dependence. Finally, in each rainfall erosion property region, the Kriging method is applied to estimate the R values for the areas without observed rainfall data. From the results of cross validation, it is found the R values estimated using Method 3 is more accurate than that using Method 2, i.e., Method 3 is better than Method 2.