Implementation of a COM-based decision-tree model with VBA in ArcGIS

The problem of soil pollution by heavy metals has been receiving an increasing attention in the last few decades. Geostatistics module in ArcGIS, could not however efficiently simulate the spatial distribution of heavy metals with satisfied accuracy when the spatial autocorrelation of the study area severely destroyed by human activities. In this study, the classification and regression tree (CART) has been integrated into ArcGIS using ArcObjects and Visual Basic for Application (VBA) to predict the spatial distribution of soil heavy metals contents in the area severely polluted. The overall CART accuracy of assigning samples to the right Pb classes is 89.62% and 85.71%, the Kappa coefficient is 0.8444 and 0.7575, respectively for learning data and test data. This is a great improvement comparing with ordinary Kriging method in ArcGIS. The integrated approach allows for relatively easy, fast, and cost-effective estimation of spatially distributed soil heavy metals pollution. The methods and results described in this study are also valuable for understanding the relationship between heavy metals pollution risk and environmental