Integrating UAV-Based RGB Data and Machine Learning for Accurate Assessment of Soil Salinity

Soil salinization poses a significant threat to agricultural productivity and sustainability, particularly in semi-arid regions such as Northeast Thailand. This study investigated the potential of unmanned aerial vehicle (UAV)-based RGB photogrammetry combined with machine learning regression models to estimate surface soil salinity at a high spatial resolution. A total of 250 soil samples were collected across the two agricultural seasons and analyzed for electiical conductivity (EC) as a proxy for salinity. RGB imagery was acquired using a DJI Mavic Air 1 UAV and processed to generate orthomosaics at four ground sampling distances (GSDs): 5, 25. 50, and 100 cm. From the imagery, seven color-based indices (GRVI. RGRI. GBRI. RBRI. m. rg. and rb) were computed and used as input features for three machine-learning models: Generalized Linear Model (GLM). Random Forest (RF). and Support Vector Regression (SVR). The results showed that RF at a GSD of 25 cm achieved the highest prediction accuracy (R2 = 0.68. RAISE = 4.56). These findings underscore the utility of RGB-derived indices and machine learning models in producing cost-effective, scalable, and accurate salinity maps. This approach is promising for supporting precision agriculture and land management in salt-affected regions.