Risk Assessment of Drinking Water Source Based on High Spatial Resolution Remote Sensing

This study combines object-based image analysis (OBIA) and deep learning (DL) techniques for automatic risk source extraction and risk assessment in drinking water sources using high-spatial-resolution (HSR) imagery. HSR images are first segmented to obtain the object primitives (OP), and the DL training samples are collected based on the OPs. U-Net semantic segmentation and object-based classification methods are then used to extract the risk source targets. Analytic hierarchy process (AHP) method is used to obtain the weights of risk sources. The risk assessment model is then established which comprehensively evaluates the safety level of drinking water source environment. Experiments at the Jiajiang water source of Nanjing city are carried out based on china's GaoFen-2 (GF-2) imagery, in which the risk sources such as farmlands, woodlands, grasslands, roads, and industrial and residential buildings are extracted and evaluated of their risk influences by the comprehensive risk assessment index (RAI). It shows that on December 12, 2018, the environment of Jiajiang drinking water source is at a safety level of “high”.