Methods to extract impervious surface areas from satellite images

TLDR

Impervious surface area is a key environmental and socioeconomic metric, and remote sensing—particularly high‑resolution imagery—is the main source for its mapping, yet challenges remain in urban–rural frontiers, large‑scale coverage, and integrating coarse‑resolution data for improved accuracy. The study aims to summarize general ISA mapping procedures and major techniques, discuss how scale influences data and algorithm selection, and develop an optimal procedure tailored to specific study areas for accurate mapping. The authors review ISA mapping procedures and major techniques, highlighting how scale affects data and algorithm choice, and describe the use of texture and object‑based methods in high‑resolution imagery, as well as multi‑resolution data fusion and spectral mixture analysis to.

Abstract

Impervious surface area (ISA) is an important parameter for many environmental or socioeconomic relevant studies. The unique characteristics of remote sensing data made it the primary data source for ISA mapping at various scales. This paper summarizes general ISA mapping procedure and major techniques and discusses impacts of scale issues on selection of remote sensing data and corresponding algorithms. Previous studies have indicated that ISA mapping remains a challenge, especially in urban–rural frontiers and in covering a large area. Effectively employing rich spatial information in high spatial resolution imagery through texture and object-based methods is valuable. Data fusion of multi-resolution images and spectral mixture analysis are common approaches to reduce the mixed pixel problem in medium spatial resolution images such as Landsat. Coarse spatial resolution images such as MODIS and DMSP-OLS are valuable for national and global ISA mapping but more research is needed to effectively integrate multisource/scale data for improving mapping performance. Development of an optimal procedure corresponding to specific study areas and purposes is required to generate accurate ISA mapping results.