应用SGS和LHS分析数字土壤质量评价中的不确定性

土壤信息不确定性在空间分析中的传递是数字土壤评价中的关键问题。本文提出使用序贯高斯模拟(Sequential Gaussian simulation,SGS)和拉丁超立方抽样(Latin hypercube sampling,LHS)相结合的方法(即SGS-LHS),来应对该问题,目的在于充分利用SGS和LHS各自的优点,互补各自的缺点,以提高不确定性传递分析的准确性和效率。这种方法(包括两种途径:SGS-LHS1和SGS-LHS2)和SGS、LHS一起被应用于香港农田土壤质量评价中,并进行了比较。结果表明:(1)SGS-LHS分析所得的不确定性结果与SGS接近一致,与LHS则有一定的差别,但差别不大;(2)SGS-LHS估计不确定性的准确性与SGS接近一致,且两者均高于LHS,尽管LHS估计的置信度区间平均宽度略显精确。

Application of SGS and LHS to analyzing uncertainties in digital soil quality assessment

Propagation of uncertainty from soil information to digital soil assessment is a key problem. This study proposed to combine sequential Gaussian simulation (SGS) and latin hypercube sampling (LHS), i.e., SGS-LHS, to cope with this problem, aiming to take advantage of the two methods and compensate their respective drawbacks, so that accuracy and efficiency of uncertainty propagation analysis could be improved. The method, i.e., SGS-LHS, as well as SGS and LHS, was applied into soil quality assessment for agricultural farms in Hong Kong. Results from comparison of the methods in the application showed that: (1) uncertainty of soil quality estimated by SGS-LHS was similar to that by SGS, while different from those by LHS to some extent; (2) accuracies of the estimated uncertainties produced by SGS-LHS and SGS were very similar and both higher than those produced by LHS, although LHS estimated narrower confidential intervals. Thus, this paper recommended using SGS-LHS to analyze uncertainty propagation in spatial analysis.