透過您的圖書館登入
IP:18.188.175.182
  • 期刊

以有限資料建立土壤轉換函數推估坡耕地之總體密度

Estimating a Pedotransfer Function for Predicting Soil Bulk Density of Rural Sloplands by Using Limit Dataset

摘要


總體密度(bulk density, Bd)為一重要之土壤物理性質,它可直接影響土壤孔隙度、有效水分含量及水力傳導度並間接影響作物生長。總體密度因現地採樣耗時,通常較難於野外測定獲得,因而該性質於土壤資料庫中常付之闕如。本研究於台灣南部具相同母質來源之老埤臺地上,以網格法(間距為250 公尺)採集165 個表土(0-20公分)土樣。本研究目的在於以較易於野外判定或室內分析獲得之土壤特性,如:土地利用類型、pH、土壤有機質含量、質地、鹽基含量、游離鐵含量等性質,以逐步回歸方式建立一適合台灣地區高度風化之坡耕地下,可準確預測Bd 之土壤轉換函數(pedotransfer function, PTF)。研究結果顯示,pH、黏粒與土壤有機質為影響研究區域內Bd 之主要性質,約可解釋30%之變異。本研究所建立預測Bd 之PTF 為Bd = 0.875+ 0.073 pH + 0.002 Clay - 0.065 ln (SOM),R^2 為0.30。此外,本研究於研究區域內另採集59 個土樣,實際測定Bd,以驗證所建立之PTF 準確度;結果顯示該PTF 平均預測誤差 (mean prediction error, MPE)及均方根誤差(root mean square prediction error, RMSPE)分別為-0.02 及0.13,顯示此PTF 可準確預測研究區域內之Bd。

並列摘要


Bulk density (Bd) is an important soil property to be considered to directly affect soil porosity, available water content and water conductivity in determining crop productivity. Due to time-consuming sampling procedure in-situ, Bd is usually absence in soil database. This study aims to establish a pedotransfer function (PTF) to predict Bd for highly weathered rural soils at Laopi terrace, southern Taiwan. One hundred and sixty five surface soils (0-20 cm) were collected at the terrace, and then were separated into 2 data sets. Data set 1 (n=106) was used for PTF establishment, and the data set 2 (n=59) was used for validation. Soil properties easily obtained in-situ or from indoor analysis, such as land uses, pH, soil organic matter (SOM), texture, base contents, and free Fe contents etc. were used to establish the PTF in this study. The stepwise regression was selected in this study to establish the PTF. The results indicated that pH, clay fraction and SOM are major factors to influence Bd at the terrace, and they compose a PTF to well predict Bd as follows: Bd = 0.875 + 0.073pH + 0.002Clay - 0.065ln(SOM). This PTF’s R^2 is 0.30. Regarding validation of this PTF, the data set 2 was used to fit the PTF to validate the predictive quality. Mean prediction error (MPE) and root mean square prediction error (RMSPE) are -0.02 and 0.13, respectively, implicating this PTF could well predict Bd at this terrace.

並列關鍵字

Bulk density Pedotransfer function Land uses

延伸閱讀