A Review of Soil Moisture Mapping based on Remote Sensing Technology

Abstract

Soil moisture is one of the key components of the land surface that is highly sensitive to climate change. It plays a crucial role in the water exchange between the land surface and the atmosphere and serves as an important hydrological variable that directly influences droughts, floods, ecosystem stability, and agricultural productivity. In recent years, methods for estimating and mapping soil moisture based on remote sensing techniques have developed rapidly. However, studies that comprehensively evaluate the diversity of satellite data, modeling approaches, and algorithms used in this field remain relatively limited. This study aims to analyze soil moisture estimation studies based on remote sensing methods published in high-impact scientific journals and to identify key research trends and fundamental concepts. A total of 4,933 articles have been published in this field over the past nine years, showing a steady increase in publication output. Based on a content analysis of 66 selected articles, a keyword co-occurrence network analysis revealed three major clusters, focusing on satellite types, data processing methods, and application domains. From a methodological perspective, the studies can be classified into three main groups: index-based approaches, machine learning and deep learning approaches, and physical and physically-based modeling approaches. Index-based methods such as OPTRAM, STIM, and TVDI account for 15.1% of the studies, while machine learning approaches account for 59.1%, and physical or physically-based models represent 25.8%. Machine learning and deep learning approaches have demonstrated the highest performance in producing high-resolution soil moisture maps (R = 0.86–0.94, RMSE = 0.03–0.05 m³/m³). Therefore, recent studies increasingly focus on developing  approaches that integrate multi-source data and machine learning techniques.