面向环境质量等级划分的加密采样空间布局优化方法研究

Environmental quality category delineation is one of the most important aims in resource and environment field, for instance delineation of forbidden cultivated areas, delineation of areas for environmental remediation, statistics of areas of different water quality categories, functional partition of sea closed to shore. Different from global mean estimation and spatial interpolation, environmental quality category delineation only focuses on estimation the relationship between the unknown value of the target variable and the threshold. And also the environmental quality categories of spatial units are usually spatially continuous because of spatial autocorrelation. Thus to improve the delineation precision of environmental quality categories, more sampling points should be laid in the transition area of environmental quality categories in additional sampling. However, because the sampling points of primary survey are usually sparse and would produce large errors and uncertainty in estimated result, it’s hard to recognize the true qualities transition area. What’s more, application has different requirements on two types of errors in categories delineation. The above-mentioned problems cannot be solved by available spatial sampling optimization method. Aiming at these problems, this program plans to put forward a spatial layout optimization method for additional sampling of environmental quality categories delineation. In this method, the non-parametric estimation method would be employed for inference and its estimation result with confidence interval would be used to constitute the error index quantizing the probability of wrong category delineation. Also the control parameters of two types of errors and balance between local and global delineation precision would be integrated with the error index to design the spatial layout optimization objectivity function. The proposed method would save the survey and monitoring cost, improve the delineation precision of environmental quality categories, serve for the differentiated, delicacy and scientific management of resource and environment, and contribute to the development of theories and methods of multi-phased spatial sampling design.

环境质量等级划分（如禁产区划分、土壤修复区划定、海域功能区划等）是环境调查和监测的一项重要目标。不同于总体估计和空间插值，等级划分只需要准确估计目标变量取值与等级划分阈值之间的大小关系，而同时由于环境质量等级具有空间连续性，加密采样时应该将样点布设在等级过渡地带。但是由于初步调查的样点一般较为稀疏，导致等级估计结果具有较大的不确定性，难以反应真实的等级过渡情况，而且具体应用对两类等级划分错误也有不同的控制要求，已有采样优化布局方法不能有效解决上述问题。因此，本研究拟基于非参数估计结果与概率阈值的关系构造反映错划可能性的指标，结合两类错误的控制参数以及研究区局部与全局等级划分的准确性构造优化目标函数，提出面向环境质量等级划分的加密采样空间布局优化方法。研究成果能节约调查和监测成本，提高等级划分精度，为差别化、精细化和科学化的环境管理、利用和治理提供数据基础，并推动多阶段采样设计方法的发展。

环境质量等级划分是环境调查和监测的一项重要目标。不同于总体估计和空间插值，等级划分只需要准确估计目标变量取值与等级划分阈值之间的大小关系，同时由于环境质量等级具有空间连续性，加密采样时应该将样点布设在等级过渡地带。但是初步调查的样点一般较为稀疏，导致等级估计结果具有较大的不确定性，难以反应真实的等级过渡情况，而且具体应用对两类等级划分错误也有不同的控制要求，给等级划分加密采样空间布局带来了困难。对此，本项目在实际需求调研的基础上，基于空间统计量及其概率分布，提出了针对正态分布数据和非正态分布数据以及单指标和多指标的环境质量等级自适应划分方法和加密采样布局优化方法，与支持多目标采样的多路径空间模拟退火同步优化算法，并研发了相应的实现软件，为环境质量等级划分统计推断与采样优化提供了一整套解决方案。项目研究成果充实了面向等级划分目标的统计推断方法和采样优化方法，丰富了空间采样与统计推断方法体系，已经在土壤与海洋环境调查监测与统计推断中的到了初步的应用。将进步一推进项目成果在农用地土壤环境质量类别划分、工矿企业土壤污染修复区边界划定、水环境监测与水质面积估算中的应用。