滨海地区土壤盐分空间分布预测方法比较研究
Comparative Study of Methods for Spatial Distribution Prediction of Soil Salinity in Coastal Areas

以天津滨海地区为例, 利用4种确定性方法(反距离权重、径向基函数、全局多项式、局部多项式)和5种地统计方法(普通克里格、协同克里格、泛克里格、指示克里格、序贯高斯模拟)对典型样区进行土壤盐分空间分布估计、土壤盐渍化风险区域识别及其不确定性评价, 并针对不同方法的预测效果进行了综合对比分析．结果表明：地统计方法的插值精度整体上优于确定性方法, 序贯高斯模拟具有相对最高的预测精度; 模拟结果的统计特征与实测数据的一致性较高, 预测表面对局部细节的反映能力较强; 在以含盐量0.6% 为阈值的土壤盐分含量不确定性评估中, 序贯高斯模拟可通过多次模拟实现将环境风险性定量化, 是科学管理滨海地区盐渍化土壤资源的有效手段．
Focused on Tianjin coastal area，four deterministic methods(inverse distance weighting，radial basis functions，global polynomial interpolation and local polynomial interpolation)and five geostatistical interpolation methods(ordinary kriging，co-kriging，universal kriging，indicator kriging and sequential Gaussian simulation)were applied to estimate the spatial distribution of soil salinity of a typical sample area，identify risky areas of soil salinization and evaluate outcomes’ uncertainty. On this basis，the paper carried out comparative analyses according to the forecast effects of different methods. Results show that on the whole，the interpolation precision of geostatistical interpolation methods is higher than that of deterministic methods and sequential Gaussian simulation has relatively the highest prediction accuracy. Statistical features of simulation results and the measured data are of high consistency，and the interpolation surface of sequential Gaussian simulation can reflect the local details accurately. Furthermore，in the uncertainty assessment with the threshold of soil salt content at 0.6% ，sequential Gaussian simulation can quantify the environmental risk through multiple simulations. It is an effective?measure to manage saline soil resources in coastal areas scientifically