Combining high resolution input and stacking ensemble machine learning algorithms for developing robust groundwater potentiality models in Bisha watershed, Saudi Arabia

Abstract The present research aims to build a unique ensemble model based on high-resolution groundwater potentiality (GPM) by merging the random forest (RF) meta classifier-based stacking machine learning method with conditioning factors in Bisha watershed, Saudi Arabia. Using satellite images and other secondary sources, twenty-one parameters were derived this study. SVM, ANN, LR meta-classifiers used create new method. RF classifiers algorithm. Each of these three models was compared separately. GPMs then confirmed using ROC curves, such as empirical binormal ROC, both parametric non-parametric. Sensitivity analyses GPM carried out an RF-based approach. Predictions made six hybrid algorithms for very high (1835–2149 km 2 ) potential (3335–4585 regions. (ROCe-AUC: 0.856; ROCb-AUC: 0.921) beat ROC's area under curve (AUC). sensitivity study indicated that NDMI, NDVI, slope, distance water bodies, flow accumulation most sensitive parameters. This work will aid improving effectiveness developing sustainable management plans utilizing DEM-derived