Using spatiotemporal fusion algorithms to fill in potentially absent satellite images for calculating soil salinity: A feasibility study

Currently, remote sensing technology has unique advantages in monitoring soil salinity information, but it is vulnerable to the quality of imaging, and spatiotemporal fusion algorithms can fill these missing images. The purpose of this study is to evaluate the applicability of four commonly used spatiotemporal fusion algorithms for monitoring soil salinity. The applicability of four spatiotemporal fusion techniques for generating fused images based on Landsat 8 and Sentinel-2 data was examined. To estimate soil salinity in the oasis intersection zone region of Xinjiang, China, random forest regression models were built using measured soil salinity and multiple fused images. The results show that the spatiotemporal fusion algorithm can generate fused images with good accuracy and the constructed random forest regression model can assess soil salinity well. We conclude that using a spatiotemporal fusion technique to generate fused pictures can solve the missing data problem caused by poor imaging quality, and that the fused images are well suited for monitoring soil parameters sensitive to environmental changes, such as soil salinity. One of the key influencing variables on the accuracy of the developed assessment model is the correctness of the fused images. When dealing with incomplete data, this work can serve as a scientific reference for monitoring soil salinity utilizing remote sensing methods.

Automated summary

This study evaluates the applicability of four commonly used spatiotemporal fusion algorithms for monitoring soil salinity and assesses soil salinity using a random forest regression model. The results show that the spatiotemporal fusion algorithm can generate accurate fused images, which is crucial for monitoring soil salinity.