| 摘要: | 土壤深度為預測淺層山崩的定值分析法中重要參數之一，也與淺層山崩崩塌量的估測密切相關。本研究以石門水庫集水區為例，建立自然斜坡土壤深度推估經驗式，並利用手動螺旋鑽進行野外調查量測土壤深度。本研究調查點選在單純由土壤生成作用及側向的擴散作用所形成之自然斜坡上，避開平地、河床、河階、崩塌地、侵蝕溝及人工開挖回填之坡地。除了土壤深度量測，坡度及坡向也將一併紀錄，而地形曲率、濕度指數及岩性則透過DTM 計算及地質圖取得。將土壤深度與影響因子進行統計迴歸分析建立石門水庫集水區土壤深度推估經驗式，並以R2 值、殘差分析及誤差均方根來評估不同的因子組合及不同迴歸模型所得推估經驗式之適切性。結果顯示以採用多變量迴歸方法，並以坡度百分比、地形曲率、濕度指數、坡向及岩性為因子，所得之評估結果最佳，其R2 值高達0.91，殘差落在±0.2m 以內的資料組數為75 組（占資料總數90%），殘差值之標準差為0.14m，誤差均方根為0.14m。而只以坡度百分比為因子之單變量迴歸亦有不錯結果，其R2 值達0.82，殘差落在±0.2m 以內的資料組數為56 組（占資料總數67%），殘差值之標準差為0.22m，誤差均方根為0.22m。
Soil depth is an important factor in the physical based method of the prediction of shallow landslides. It is also critical in the estimation of the shallow landslide volume. This reaserch selected the Shihmen Reservoir catchment area as a study area to develop empirical fomulas for estimation of soil depth on natural slope. By using field obvervation and hand auger, I performed the investigation of soil depth at several test sites. The measurement spots were selected on natural slope, which are formed by soil production and
modified by mild slope processes, avoiding falts, river banks, river terraces, collapeses, gullies, and artificial slopes. Accompany the depth measurement, coordinate as well as slope gradient and aspect at the spot were also measured
and recorded. Terrain curvature, wetness indexe and lithology at a spot are then reduced from DTM and gelogical map, and put together for use. Regression analysis of soil depth and controlling factors is performed in order to build an empirical estimation model of soil depth at Shihmen Reservoir catchment. Different combinations of factors and form were testd in the study, and R2 value, residual analysis and root mean square of error were used to evaluate the appropriateness of an empirical estimate model. Result reveals that the best estimate is the mutiple regression using slope gradinet, terrain curveture, wetness indexe, slope aspect, and lithology as factors
the R2 value is 0.91, the residual value of 75 data (90% of all data) are within ±0.2m, the standard deviation of residual is 0.137m, and the root mean square of error is 0.14m. A simple formula using only slope gradient is also recommended
the R2 value is 0.82, the residual value of 56 data (67% of all data) are within ±0.2m, the standard deviation of residual is 0.218m, and the root mean square of error is 0.22m. |
