Prediction of Spatial Distribution of Lime Requirement in Strongly Acidic Soils by Geostatistics

地理統計預測強酸性土壤中石灰需要量之空間分佈

吳振記(Chen-Chih Wu) ； 陳仁炫(Jen-Hshuan Chen)

《臺灣農業化學與食品科學》 45卷3期 (2007/06) Pp. 163-171

https://doi.org/10.6578/TJACFS.2007.020

摘要

正確預測和繪製石灰需要量圖對強酸性土壤之管理上是必要的。本研究利用SMP迴歸模式預測石灰需要量，並結合一般克利金法以預測石灰需要量的地理分佈。驗證已建立之SMP迴歸模式及建立採樣策略以作為預測台灣強酸性土壤中石灰需要量之空間分佈為本研究之目的所在。以網格採樣方式，採集台灣省台中縣栽種番薯和蘿蔔之表土20公分樣品共110個，每個土樣均以SMP-single緩衝溶液法和降低鋁飽和度至10%法測定其石灰需要量。利用不同採樣數之絕對平均誤差(meanabsoluteerror)和均方根誤差(meansquareerror)作為評估最少採樣點的決定依據。結果顯示，以0.5公頃網格，即每公頃兩個採樣點之採樣方式，並利用一般克利金法結合球狀模式可預測此研究區域之石灰需要量之空間分佈。本試驗建議，可藉由已建立之SMP迴歸模式來預估改善台灣強酸性土壤的適宜石灰需要量，並以0.5公頃網格方式決定採樣方式，配合一般克利金法可預描繪石灰需要量分佈圖。

關鍵字

石灰需要量 ； 空間分佈 ； SMP迴歸模式 ； 地理統計 ； 一般克利金法

並列摘要

Accurate prediction and mapping of lime requirement (LR) are essential for managing strongly acidic soils. The SMP regression model was used to predict the LR and ordinary kriging method was used to predict the spatial distribution of LR in this study. The objective of this research was to verify the established SMP regression model and develop a sampling strategy for the spatial distribution of LR for strongly acidic soils in Taiwan. 110 soil samples were collected from a depth of 0-20 cm by grid sampling in a sweet potato and turnip field located in Taichung county, Taiwan. The SMP-single buffer and reducing soil Al saturation to 10% methods were used to measure the LR of each sample. The least sampling size was evaluated by the mean absolute error (MAE) and mean square error (MSE) calculated from different sampling sizes. The results showed that the least sampling size for LR determining of this study area was 0.5-ha grids and the spatial distribution of LR of the study area could be predicted by using ordinary kriging with spherical model. It suggested that the suitable amount of liming for ameliorating strongly acidic soils in Taiwan can be predicted by the established SMP regression model and the map of LR can be estimated by using ordinary kriging with 0.5-ha grids sampling size.

並列關鍵字

Lime requirement ； Spatial distribution ； SMP regression model ； Geostatistics ； Ordinary kriging method