Journal of Ecology and Rural Environment ›› 2020, Vol. 36 ›› Issue (3): 308-317.doi: 10.19741/j.issn.1673-4831.2019.0326

Previous Articles     Next Articles

Spatial Prediction Method of Regional Soil Heavy Metals Content Based on Multiple Model Optimization

HUANG Zhao-lin1, DING Yi2, WANG Jun-xiao1, JIA Zhen-yi1, ZENG Jing-jing1, ZHOU Sheng-lu1   

  1. 1. School of Geography and Ocean Science, Nanjing University, Nanjing 210023, China;
    2. College of Software, Tongji University, Shanghai 201800, China
  • Received:2019-05-08 Online:2020-03-25 Published:2020-03-25

Abstract: Spatial prediction of soil heavy metals is of great significance for utilization of regional soil resources, environment protection and pollution prevention. Jintan District in Changzhou City, Jiangsu Province was chosen as study area. Spatial variation model (BP-K), source-sink model (BP-S) and comprehensive model (BP-SK) were applied to simulate and predict the spatial distribution of heavy metals, including Cd, Pb, Cr, Cu and Zn. The results of the Models' performance were evaluated by coefficient of determination, root-mean-square deviation, mean absolute error and so on. Furthermore, the prediction results from the best performance model were selected to calculate heavy metal spatial distribution in each sub-region. The results show that:(1) The coefficient of determination of BP-SK model was higher than that of BP-S and BP-K for Cd, Cr, Cu and Zn. Only for Pb, the coefficients of determination of BP-S and BP-K were higher than that of the BP-SK model. BP-SK can strengthen local characteristics and uncover more information. (2) The comprehensive model performed better than original single model. The root-mean-square deviation of Cd, Pb, Cr, Cu and Zn were improved by 15.15%, 20.71%, 19.19%, 1.75% and 9.24%, respectively. (3) In prediction results, the high-value areas for Cd, Pb, Cr, Cu, and Zn were located in the central and northeastern parts, and the low-value areas were located in west hilly and mountainous areas. The BP-SK model has a better performance in areas with severe human influence, while the BP-K model has better applicability in hilly area where natural factors have a greater impact.

Key words: soil heavy metal, source-sink, spatial differentiation, back-propagation network, spatial prediction

CLC Number: