Estimating urban impervious surface percentage with multi-source remote sensing data

In this article large-area impervious surface mapping is carried out within the Shenzhen urban area by employing the CART (classification and regression tree) method with remote sensing data acquired by 4 satellites. The experimental results show that near-infrared (NIR) band is of most importance for ISP estimation. Another conclusion is that higher estimation accuracy can be obtained with multispectral imagery of higher spatial resolution and better radiometric quality. However, an unfavorable phenomenon can be consistently observed from all results that over-estimation and under-estimation do exist around the lower and upper bound of actual ISP value range respectively.