Uncertainty-aware MAR planning with spatially explicit data-driven weighting

This study focuses on the Sado and Ribeira do Alentejo River Basins (southern Portugal), where managed aquifer recharge (MAR) is considered a promising strategy for increasing water availability while preserving groundwater quality through low-maintenance and cost-effective techniques such as infiltration basins and trenches. Key factors influencing MAR suitability were jointly selected with stakeholders and include aquifer properties (geochemistry, geometry, lithology, storage capacity, and specific yield), vadose zone thickness, land slope, land use, and topsoil texture. These criteria were integrated within a GIS-based multicriteria decision analysis (MCDA) framework. Although GIS-MCDA approaches are widely used for MAR screening, most studies rely on deterministic criterion weights and single suitability estimates. To address this limitation, an uncertainty-aware GIS-MCDA framework is proposed that explicitly accounts for spatial heterogeneity and autocorrelation in the weighting process. Stratified stochastic sampling guided by Moran's I correlogram was used to generate multiple weight realizations derived from correlation-based and principal component analysis (PCA)-based approaches. This procedure reduces subjectivity in weight assignment and enables quantification of uncertainty associated with suitability assessment. The PCA-based approach produced wider weight distributions (standard deviation: 0.033-0.073) than the correlation-based approach (0.0007-0.0023), indicating greater variability in criterion importance. Propagation of this uncertainty through the GIS-MCDA framework generated suitability-map ensembles yielding statistical summaries, probability-of-exceedance maps, and agreement diagnostics. Most of the study area exhibits low-to-moderate suitability (SI < 0.6), representing 78% and 65% of the area under the PCA- and correlation-based approaches, respectively. Combining the outputs of both weighting approaches yielded a final classification that distinguishes robust MAR candidate areas, conditionally suitable zones, and areas of low suitability or high uncertainty. The proposed framework provides a transferable screening-level methodology for evidence-based MAR planning and prioritization under uncertainty.