Soil salinity estimation based on satellite hyperspectral and synthetic aperture radar remote sensing image fusion

Soil salinization poses a significant threat to agricultural sustainability. High-precision remote sensing for mapping soil salinity is crucial for effective salinity management. Hyperspectral image (HSI) serve as essential data sources for regional-scale monitoring of soil salinity. However, the spectral response to soil salinity is highly susceptible to coupling effects from other soil physical properties. Synthetic Aperture Radar (SAR) remote sensing is highly sensitive to soil physical parameters and can effectively compensate for the limitations of HSI. This study proposes an HSI and SAR image fusion method based on a multi-scale, multi-depth Wasserstein Generative Adversarial Network with Gradient Penalty (MSD-WGAN-GP) to improve soil salinity estimation accuracy. A soil salinity prediction model was developed using a convolutional neural network based on 123 soil samples collected from Northeast China. The results show that: (1) HSI and SAR fusion can significantly improve the prediction accuracy of soil salinity. Compared with the prediction of soil salinity based on HSI, the R2 and RPIQ of the model increase by 0.22 and 1.13, respectively, and the RMSE is reduced by 2.68 ds·m⁻¹ . (2) Compared with the traditional image fusion method, the MSD-WGAN-GP model demonstrates superior performance in the fusion of HSI and SAR image, achieving a peak signal-to-noise ratio of 38.39 dB and a structural similarity index of 0.88. (3) The MSD-WGAN-GP model significantly improved the correlation between soil salinity and spectra, achieving an average increase of 0.32 in the correlation coefficient per spectral band, while effectively mitigating the prediction bias introduced by soil moisture and surface roughness. This study emphasizes the significance of integrating multi-source remote sensing data to comprehensively capture the multidimensional characteristics of soil, thereby enabling more accurate estimation of soil salinity.