Estimating impervious surface percent in plain river network regions using a refinement CART analysis

摘要

The rapid expansion of impervious surface has become a major factor affecting ecosystem health of the high density river network. In this paper, the ensemble learning of CART analysis was used to estimate impervious surface percent (ISP) through Variable Precision Rough Sets (VPRS). First, Landsat TM and ALOS imagery were utilized to construct the ISP predictive model; then, in order to get the best attribute variables of CART decision tree, VPRS was adopted to extract optimum feature subset from multi-source feature sets. Results illustrate the validity of this ensemble learning, and prove that this method can obtain higher accuracy than the traditional single CART method. However, in the initial estimation results, ISP's high value area was underestimated relatively seriously. It was found that there is an intensive relationship between the Temperature Vegetation Dryness Index (TVDI) and ISP. The increase of ISP will cause significant increase of local TVDI. Then post-processing rules extracted from the relationship was used to improve results. According to the verified results, the combination of VPRS reduction and post-processing rule in CART algorithm has higher analysis precision than the traditional single CART learning algorithm. The root mean square error between estimated ISP value and reference ISP is 10.0% and the correlation coefficient is 0.89. The method is viable for the estimation of the ISP in plain river network regions.