土壤性質空間與時間分佈之推估應用於土壤定址管理

Abstract

為降低農業土壤不當的磷肥施用對環境造成的危害，並且提高肥料施用之效益，需對磷肥施用進行定址管理。對磷肥施用進行定址管理，為考量作物差異與土壤性質變異下，將田間土壤規劃出差異性施肥之磷肥施用管理區。管理區之界定需要有準確之土壤性質空間分佈圖，因此本論文第一章提出了一個結合土壤圖類別資訊的地理統計模式，稱為克利金法結合土壤圖繪圖界線 ( The kriging combined with soil map-delineation, KSMD)，用以改善土壤性質空間分佈之推估。KSMD法為依據土壤圖之土系界線，將採樣點資料分組以將資料之變異源區分成兩部分，分別為土系間變異與土系內變異。KSMD法與一般克利金 (ordinary kriging, OK) 法模擬推估之結果比較，兩者之乖離率 (bias) 相似，而KSMD法相對於OK法之推估不精密度的減少率 (percentage of the decrease of estimation imprecision, DIP %) 於土壤性質之砂粒、坋粒、黏粒含量、pH、Mehlich-3 Ca與P，分別為34 %、40 %、48 %、20 %、42 %、與3 %，表示KSMD會有較高之精密度，因此能改善土壤性質空間分佈之推估。 土壤性質除了在空間分佈上具有變異性以外，部份土壤性質亦會隨著時間而變化，例如磷肥有效性指標 (fertilizer-phosphorus availability index, Fp)。磷肥有效性指標為描述磷肥施用後其仍能維持有效性之比率，其為反映土壤固定磷之能力。土壤固定磷之能力會隨時間增加而加強，因此本論文第二章探討了不同時間下之磷肥有效性指標的空間分佈之推估的兩種方式：先以克利金法進行空間分佈推估再套配動力學模式；先建立動力學模式再推估動力學模式參數之空間分佈。其中先建立動力學模式再推估動力學模式參數之空間分佈之方式，會因動力模式之參數已有偏差之情形下再進行空間分佈推估，使得推估結果有明顯之偏差。因此在時間變化上之資料較少的情形下，先進行空間分佈推估再套配動力模式的空間分佈推估結果較穩健，故較為適合用於推估不同時間下磷肥有效性指標之空間分佈。 而現行磷肥施用之推薦，係以速測法測得之土壤有效磷量將土壤肥力分級，再依據不同作物別推薦其磷肥施用量。此種施肥推薦之方式僅考慮施肥量與作物產量之關係，並無法評估施肥後土壤有效性磷量的變化。在維持作物產量之條件下，同時考慮施肥後土壤有效性磷量的變化，以進行磷肥施用之定址管理，將可提高施肥之效率。因此本論文第三章由土壤有效磷量 (P0) 與磷肥有效性指標 (Fp) 及不同作物之土壤有效磷臨界濃度 (Pc)，計算出不同作物之需磷量 (Pneed)，將需磷量等級距的劃分作為磷肥施用管理組之分組。此種管理組之建立方式，同時考慮到作物最適產量與施肥後土壤有效磷量變化，因此可提高磷肥施用之定址管理的效率。

To mitigate the impact on the environment from agriculture and to improve the efficiency of fertilizer application, soil site-specific management is needed. The site-specific phosphorus management is proposed in this study by delineating management zones for variable rate fertilizer application to manage in-field variability. The quality of mapping soil properties would impact the performance of soil site-specific management. Thus, a kriging model combined with soil map-delineation (KSMD), taking into account the variation components of soil type effect and residual to improve the estimation of soil properties, was proposed. The KSMD estimation was based on the soil map-delineation of soil types to group the sampled observations, and separated the variation of soil properties into two parts: one between soil types and the other within each soil type. When comparing KSMD and ordinary kriging (OK), the mean errors (ME) of KSMD and OK estimations were similar. However, decreases in estimation of imprecision for KSMD relative to OK (DIP %) for sand, silt, and clay contents, pH, and Mehlich-3 Ca and P were 34 %, 40 %, 48 %, 20 %, 42 %, and 3 % respectively. These results suggested that the proposed KSMD method could increase the precision of the interpolation of soil properties and improve the estimation of soil properties. In addition to spatial variability of soil properties, some soil properties have temporal variability, such as fertilizer-P availability index (Fp). The Fp is a ratio of the increase in extractable soil P to amount of P added, which represents the fixation tendency of soils. In chapter 2 of this dissertation, the estimation of Fp at different incubation time was performed by using the kriging-prior-to-Elovich-fitted and the Elovich-fitted-prior-to-kriging estimations. The results showed that the parameter a and b of the kinetics of Fp had bias prior to the kriging estimation of spatial distribution of Fp and hereafter could lead to a deviation of the Elovich-fitted-prior-to-kriging estimation. Compared to the Elovich-fitted-prior-to-kriging estimation, the kriging-prior-to-Elovich-fitted estimation was an unbiased estimator, and thus was a more robust and suitable estimating method for the estimation of Fp at different incubation time under the constraint of limited temporal data. Traditional recommendation for fertilizer P application was based on the soil test result to evaluate the soil fertility level for the recommendation of fertilizer application for different crops. This way of recommendation for fertilizer P application considered the effect of amounts of fertilizer applied on crop yield only, but didn’t evaluate the change of availability of P in soil after fertilizer P applied. For improving the efficiency of fertilizer P application, site-specific P management zones should take into account the maintainability of desired crop yield and the variation of available P in soil after fertilizer P applied. In the last chapter of this dissertation, site-specific P management zones were delineated by dividing the amounts of P needed (Pneed) with equal range. The Pneed was calculated from the amount of available P (P0) in soil prior to fertilizer applied and fertilizer-P availability index (Fp) under different criteria concentration of P (Pc) in soil for various crops. Both of the optimal crop yields and the variance of availability of P in soil after fertilizer P applied were considered, and then the efficiency of site-specific P management could be improved.