Integrating Active and Passive Remote Sensing Data for Mapping Soil Salinity Using Machine Learning and Feature Selection Approaches in Arid Regions

The prevention of soil salinization and managing agricultural irrigation depend greatly on accurately estimating salinity. Although the long-standing laboratory method measuring salinity composition is accurate for determining parameters, its use frequently constrained by high expense difficulty long-term in situ measurement. Soil northern Nile Delta Egypt severely affects agriculture sustainability food security Egypt. Understanding spatial distribution a critical factor development management drylands. This research aims to improve prediction using combined data collection consisting Sentinel-1 C radar Sentinel-2 optical acquired simultaneously via integrated sensor variables. modelling approach focuses feature selection strategies regression learning. Feature approaches that include filter, wrapper, embedded methods were used with 47 selected variables depending genetic algorithm scrutinize whether regions spectrum from indices SAR texture choose optimum combinations sub-setting resulting each train learners’ random forest (RF), linear (LR), backpropagation neural network (BPNN), support vector (SVR). Combining BPNN RF learner better predicted (RME 0.000246; = 18). Integrating different remote sensing machine learning provides an opportunity develop robust predict evaluated performances various models, overcame limitations conventional techniques, optimized variable input combinations. can assist farmers soil-salinization-affected areas planting procedures enhancing their lands.