王院民, 陈东湘, 仝桂杰, 等. 基于决策树模型的水稻镉超标空间识别及预测研究[J]. 生态与农村环境学报, 2019, 35(11): 1475-1483. DOI: 10.19741/j.issn.1673-4831.2019.0205
    引用本文: 王院民, 陈东湘, 仝桂杰, 等. 基于决策树模型的水稻镉超标空间识别及预测研究[J]. 生态与农村环境学报, 2019, 35(11): 1475-1483. DOI: 10.19741/j.issn.1673-4831.2019.0205
    WANG Yuan-min, CHEN Dong-xiang, TONG Gui-jie, et al. Spatial Recognition and Prediction of Rice Cd Over-standard Based on Decision Tree Model[J]. Journal of Ecology and Rural Environment, 2019, 35(11): 1475-1483. DOI: 10.19741/j.issn.1673-4831.2019.0205
    Citation: WANG Yuan-min, CHEN Dong-xiang, TONG Gui-jie, et al. Spatial Recognition and Prediction of Rice Cd Over-standard Based on Decision Tree Model[J]. Journal of Ecology and Rural Environment, 2019, 35(11): 1475-1483. DOI: 10.19741/j.issn.1673-4831.2019.0205

    基于决策树模型的水稻镉超标空间识别及预测研究

    Spatial Recognition and Prediction of Rice Cd Over-standard Based on Decision Tree Model

    • 摘要: 为探究水稻籽粒中Cd元素的超标风险及其与环境要素之间的关系,筛选了11个环境因子,基于决策树模型,识别出影响水稻Cd超标的主控因子,并建立超标空间预测技术。结果表明,距交通运输用地距离、土壤有机质含量和无定形铁含量为研究区水稻Cd超标的主控因子。经验证,提取的研究区水稻籽粒Cd超标识别规则的精度为85.71%。进一步运用隶属度空间制图法,以有限的样点对研究区水稻籽粒Cd超标进行空间分布预测制图,分类精度达到91.67%,该结果精度比单纯的决策树和传统插值制图方法都有较大提高。构建的决策树模型能较好地预测水稻籽粒Cd超标的空间分布,对研究区水稻Cd超标空间识别与分区管控有实践意义。

       

      Abstract: Cadmium contamination of rice grains has become an important factor affecting the quality and safety of rice in China. To explore relationships between Cd contamination in rice and various environmental factors, spatial mapping was carried out. Eleven environmental factors were selected for study. Using a decision tree model and a reasoning mapping method, the main factors controlling Cd contamination in rice were identified in a case study area, and a spatial prediction technology for identifying rice with Cd levels in excess of the Chinese standard was established. The results show that transportation distance, soil organic matter content and amorphous iron content were the main factors correlated with Cd contamination in rice. Rice having Cd contents exceeding the standard in the study area were extracted and classified into five grades according to their associated risk degree. The empirical accuracy of this classification method was 85.71%. In addition, a subordinate degree spatial analysis method was used to make spatial predictions of rice areas having Cd levels exceeding the standard threshold, based on information from limited sampling points. The classification accuracy using this method reached to 91.67%, representing a marked improvement, compared with the simple decision tree and traditional interpolation mapping results. However, the decision tree model developed in this study predicts the spatial distribution of Cd contamination risk in rice, making it of practical significance to mitigating Cd contamination risks in rice.

       

    /

    返回文章
    返回