Spatial variability of soil properties

建立了一种确定基质吸力、非饱和导电性和土壤水扩散系数的方法，该方法通过最小化测量值与尺度缩减值之间的平方和来确定。采用两种方法确定标度系数；利用实验方程的估计值作为有效饱和度的函数和测量值的平均值作为尺度缩减值。在任何情况下，标度效应都很明显，两种方法的标度系数值相近。不使用实验方程的平方和比使用实验方程的平方和要小。结论是，前者非常有用，因为缩放很容易完成——对于不同的属性，只需改变约简变量中的系数值。为了从有限的测量值中估计未知值，建立了克里金技术。该过程包括以下三个步骤。首先，确定半变异函数作为滞后距离的线性方程。其次，通过最小化估计误差来确定最优邻域大小。最后，从最优邻域大小内的测量值估计任意点处的值。绘制了沙丘砂和海填地土壤的粒径分数、土壤溶液离子浓度、土壤盐分和溶液pH的等高线图。由克里格估计绘制的等高线图比仅由测量值绘制的等高线图清晰得多。利用具有相似分布的两个性质之间的线性预测方程，可以从一个性质的值估计另一个性质的值。

A technique for determining the scaling coefficients for matric suction, unsaturated hydraulic conductivity and soil water diffusivity was established by minimizing a sum of squares between the measured values and the scale reduced values. Two methods were used in determining the scaling coefficients; the estimates from an experimental equation as the function of the effective saturation and the average of the measured values were used as the scale reduced values. Scaling effect was noticeable for any case and the similar values of scaling coefficients were obtained for the two methods. The sum of squares was smaller in the case that the experimental equation was not used than that the equation was used. It was concluded that the former was very useful because scaling is easily done-for the different properties only by changing the value of coefficients in the reduced variables.The kriging technique was established in order to estimate the unknown values from the limited number of measured values. The procedure includes the following three steps. First, a semi-variogram is determined as the linear equation of the lag distance. Next, the optimum neighborhood size is determined by minimizing the estimate error. Finally, the value at the arbitrary point is estimated from the measured values within the optimum neighborhood size. The contour maps were made for particle size fraction, ion concentration in the soil solution, soil salinity and solution pH in dune sand and reclaimed land from sea. The contour maps drawn by kriging estimates were far clearer than those drawn only by the measured values. The values of one properry could be estimated from those of another property by using the linear predictive equation between two properties with the similar distribution.