支持向量机在地统计学中的应用研究——以土壤重金属砷的空间分布及其对农产品安全生产的影响为例  

Application of support vector machine in geostatistics: a study on the spatial distribution of arsenic in soil and its influence on the safety of agricultural products

作　　者：高铭[1] 王季槐[2,3] 卢大远[1] 戴家佳[1] 丰锋[3] 汪建文[2] 王大霞 杨杰文[3] 

机构地区：[1]贵州大学数学与统计学院,贵州贵阳550025 [2]贵州科学院,贵州贵阳550001 [3]广东海洋大学农学院,广东湛江524088 [4]贵州省分析测试研究院,贵州贵阳550003

出　　处：《贵州科学》2017年第3期58-64,共7页Guizhou Science

基　　金：国家自然科学基金项目(30960222);贵州省科技计划项目(黔科合NY字[2008]3050号;黔科合NY字[2011]3026号)

摘　　要：重金属砷广泛存在于地壳中,它的开采及使用造成了对农田土壤的污染,进而危害人体健康。Kriging地统计学方法是解决基于少量样本估计大面积内物质异质性分布的主要手段。但是由于传统的Krining方法对变异函数的拟合的主观性和局限性影响对物质分布的估计。为解决这一问题,引入了支持向量机(SVM)方法对变异函数进行拟合并将这一方法应用于对农田土壤重金属砷的分布的估计,即基于在目标地块实地取样获得的少量数据来估计砷在农田内的分布,结果表明:SVM回归的拟合优度为0.944,高于其他两种传统方法的拟合优度(0.842,0.744);基于该拟合函数计算的验证样点估计值与实测值的平均相对误差为4.75%;砷在目标地块内呈异质性分布。为了研究土壤砷的分布对农产品安全生产的影响,根据文献提供的实验数据建立了9种蔬菜对重金属砷的富集模型,据此建立了农产品安全生产的风险评价模型。评价结果表明,如果在目标地块中种植9种蔬菜,其中8种蔬菜是无风险的,即风险值p=0。然而,尽管整个目标地块土壤砷含量低于《土壤环境质量标准》^([14])的土壤二级标准,9种蔬菜中仍有一种蔬菜(青菜)的产品是完全超标的,即风险值p=1。因此,单纯用土壤重金属含量来评价农田土壤并不能很好地反映农田土壤污染对人类的影响。Arsenic exits widely in earth＇ s crust. Its exploitation has led to soil pollution in farmland, thus harming human health. Kriging is the main method used to estimate heterogeneous distribution of matter in large areas based on a few samples. However, the estimating precision of traditional Kriging method may be influenced by its subjec- tivity and limitations. In order to solve this problem, the SVM was introduced to fit the variograms and to estimate the distribution of As in farmland soil. The results showed that, the regression goodness-of-fit of SVM method （0. 944） was better than that of the other two methods （0. 842,0. 744） ; the average relative error between the esti- mated and tested values of the samples was 4.57 % ; there was heterogeneous distribution of As in the target farm- land. In order to analyze the influence of As distribution on the safety of farm products, we built the As concentrating models of 9 vegetables based on experimental data from the literature. Accordingly, the risk evaluation model of farm product safety was built. The evaluation results showed that,8 out of the 9 vegetables planted in the target farmland were safe, i. e. the risk value p = 0. However, although the As content in the soil of the target farmland was under the secondary standard stipulated in the ＂Environmental Quality Standard for Soils＂, the As content in one of the 9 veg- etables exceeded the standard, i. e. the risk value p = 1. Therefore, it is not fully reliable to evaluate the influence of polluted soil on human health simply by the content of heavy metals in soil.

关 键 词：支持向量机(SVM) KRIGING方法 变异函数 重金属分布 重金属富集模型 风险评价 

分 类 号：O29[理学—应用数学] TP181[理学—数学]