协方差矩阵上-下三角分解法在区域土壤水盐条件模拟的应用

Conditional simulation for spatial variability of soil water-salt by means of lower-uppertriangular decomposition of covariance matrix

本文运用协方差矩阵的上-下三角分解法对黄河河套平原上土壤水盐的空间变异性进行了条件模拟,利用55个大网格的规则采样点模拟了小尺度待估点的土壤水盐含量,模拟结果的空间分布趋势、半变异函数以及其统计特征值与普通kriging的相应估计值进行了比较。结果表明,kriging估计结果大大缩小了实测值的变异系数具有明显的平滑效应,为条件模拟的变异系数则接近于实测值,能够很好的保持土壤水盐含量的空间结构;多个条件模拟能给出土壤特性的一个波动范围及极端值。这一效果对改造中低产田、提高灌溉效率和水土资源的监测和管理决策都十分重要。由于协方差矩阵的上-下分解法避免了常用的条件模拟实现法中转向带法和傅立叶转换法的一些缺陷,其理论简单,约束条件少,可将模拟和条件化同时进行。本文的研究说明该方法应用于水土科学是可行的。

The soil water-salt in 25m×25m-spacing points of alluvial field in Hetao area of the Yellow River is conditionally simulated by means of upper-lower(LU) triangular decomposition of covariance matrix according to the observation data at 100m×100m-spacing nodes of a regular grid. The statistical characteristics, spatial distribution trend and semivariograms of the simulated result are compared with that of the estimated value by using ordinary Kriging method. The result indicates the spatial variability of ordinary Kriging method is greatly smaller, while conditional simulation can maintain the original spatial variability and is closer to the observation value. The multiple conditional simulation result can well illustrate the fluctuation range of soil water-salt and gives the extreme values of their distribution. It is concluded that the LU triangular decomposition method is feasible in the conditional simulation of spatial variability of water and soil resources.