Mapping Coastal Soil Salinity and Vegetation Dynamics Using Sentinel-1 and Sentinel-2 Data Fusion With Machine Learning Techniques

Coastal regions are vulnerable ecosystems where monitoring soil salinity and vegetation coverage dynamics is critical for environmental management and conservation. This study introduces a multisensor data fusion approach, integrating Sentinel-1 synthetic aperture radar (SAR) and Sentinel-2 multispectral imagery with advanced machine learning techniques, specifically a convolutional neural network (CNN) based classification model. This approach enables the precise mapping of coastal salt-affected soil and vegetation patterns, addressing spatial heterogeneity and dynamic environmental conditions. The analysis has been conducted for a coastal region in China, where derived features, such as normalized difference vegetation index (NDVI), salinity indices, and SAR-based soil moisture proxies, have been used as inputs to the CNN model. The model achieved an overall accuracy of 87% and a kappa coefficient of 0.82, outperforming traditional classification methods by leveraging spatial feature learning and data augmentation. Temporal NDVI trends revealed seasonal vegetation dynamics, while predicted soil moisture patterns showed strong alignment with observed ecological conditions. The results indicate that saline soils dominate the study area, with nonsaline soils and vegetated areas exhibiting scattered and localized distributions. These findings demonstrate the potential of integrating multisensor remote sensing with advanced machine learning techniques for coastal monitoring, providing a robust framework for sustainable land-use planning and ecological management.