Soil salinity inversion based on novel spectral index
Xiaohong Zhou,Fei Zhang,Changjiang Liu,Hsiang-te Kung,Verner Carl Johnson
DOI: https://doi.org/10.1007/s12665-021-09752-x
IF: 2.8
2021-08-01
Environmental Earth Sciences
Abstract:Soil salinization is one of the most important causes for land degradation and desertification and is an important threat to land management, farming activities, water quality, and sustainable development, with the management of saline soil crucial in arid and semi-arid areas. The salt-affected soil is predominant in the Ebinur Lake Wetland National Nature Reserve (ELWNNR) in the Northwestern China. It influences the development of agricultural economy. Rapid and accurate measurement of the soil salt content (SSC) is significant for the soil salinization control. However, the traditional method of obtaining soil salt is time-consuming and laborious. Nowadays, it is an unprecedented perspective to monitor soil salinity through Sentinel-2A multispectral remote sensing image to construct three-dimensional spectral index. In this study, through soil salt data of 97 ground-truth measurements, Sentinel-2A data-derived spectral indices, based on particle swarm optimization support vector machine (PSO-SVM), gray wolf optimization support vector machine (GWO-SVM) and differential evolution support vector machine (DE-SVM) algorithm to construct a best soil salt inversion models. The results show that three-band (3D) spectral index has better correlation with soil salinity than single band and two-band (2D) spectral index, among TBI5 and TBI7 has a high correlation with the salinity of the soil, and the points are concentrated on the 1:1 line. Therefore, this approach could be applied to estimate the soil salinity in the Ebinur Lake region. The established models were validated using the Machine learning algorithm. The DE-SVM model performed the best by the three Model of accuracy with R2, Bias, and SEP2C of 0.56, − 2.03, and 8.62, respectively. Therefore through the soil salinity value predicted by the modeling constructs a linear relationship with the indexes TBI5 and TBI7, and draw the soil salt inversion map, soil salinity around the lake is relatively high, and decreases outward along the lake, which is consistent with the field. The result from this model will be useful for soil salinization monitoring in the study area and can provide theoretical support for the estimation of SSC in arid and semi-arid areas.
environmental sciences,water resources,geosciences, multidisciplinary