Evaluation of soil environmental quality in karst mountain area based on support vector machine: A case study of a tea plantation in northern Guizhou

The content of heavy metals in soil directly affects the quality and safety of tea, and even has a potential threat to human health. It is hence important to monitor, evaluate and control the content of heavy metals in the tea plantation soil. In this paper, we select a tea plantation in karst mountain area of northern Guizhou as a study area. The area is located in the transitional zone from Yun-Gui Plateau to Hunan hilly area, which belongs to the humid monsoon region of tropical plateau, with the annual precipitation of 1，000 -1，300 mm and the annual average temperature of 12.6-13.1 ℃. In this study area, because the Cambrian and Ordovician carbonate strata are widely exposed, karst landform is well developed and is characterised by interlacing occurrence of peak clusters and karst valleys. According to present situation and the characteristics of land use, in the area, 80 surface soil samples were collected for the analyses of heavy metal (such as mercury (Hg), arsenic (As), cadmium (Cd), lead (Pb), chromium (Cr) and copper (Cu)) contents and the environmental quality of the tea plantation soil. To classify and evaluate the sample analytical results, the Support Vector Machine (SVM) model coded in MATLAB was employed. Meanwhile, by comparing the result from Nemerow comprehensive pollution index method with that of fuzzy comprehensive evaluation method, the applicability of SVM in soil heavy metal pollution evaluation was discussed. These results show that,(1) There are significant spatial differences in soil heavy metal contents, with the variation coefficients in the order from high to low of Cr>Hg>Cu>As>Pb>Cd. By comparing the average value of evaluation factors with the soil background value of Guizhou Province, it is found that the values of Cd and Cu are lower than the soil background values, and the others fall in between the background values and the secondary standard values of the national soil quality standard. In fact, the chemical concentrations of 91.25% of the soil samples are below the standard limits for tea producing areas, which represents a soil environment of non-pollution and high-quality for tea plantation.(2) The quality of soil in the study area is good, with its soil environmental quality ranging between category I and II.The evaluation results of SVM method are quite similar to those of fuzzy comprehensive evaluation and Nemerow comprehensive pollution index methods, with a similarity of 82.5% and 80.0%, respectively. During the application of these methods it was found that the results of SVM were more accurate, which showed that the model is suitable for the evaluation of soil environmental quality in karst mountain area. (3)In addition, the SVM can solve complex nonlinear problems, with much easier manual operation and less artificial intervention, comparing with the application of traditional assessment models. It provides a new idea and method for the evaluation of soil environmental quality in karst mountain area.