GIS ‐ MCDA and Uncertainty Analysis for Underground Dam Candidate Zone Prioritization in the Beyşehir and Çumra Subbasins

ABSTRACT Underground dams can support subsurface water storage in water‐scarce basins, but preliminary siting requires the integration of heterogeneous hydrogeological, environmental, and planning factors together with an assessment of spatial decision uncertainty. This study develops a GIS‐based sequential spatial decision‐support framework for prioritizing underground‐dam candidate zones in the Beyşehir and Çumra subbasins of the Konya Closed Basin, Türkiye. Twenty‐one raster factor layers representing hydrological, hydrogeological, geological, topographic, soil, land‐use, climatic, accessibility, and settlement conditions were standardized and integrated using hierarchical AHP weights and weighted linear combination over 2,092,998 valid pixels of an 84.68‐m grid. Criterion redundancy and criterion influence were examined before aggregation: the largest variance inflation factor in the primary criterion set is 4.18, and decomposing the variance of the composite score shows that its spatial pattern is driven mainly by five soil and geology layers, which carry 31.6% of the total weight but produce 77.6% of the spatial variation. The continuous suitability surface was classified using quantile five‐class classification, with observed equal‐interval classification as a control; within‐one‐class agreement between the two classifications was 85.03%, with 45.51% exact agreement. AHP weight uncertainty was examined through 500 Monte Carlo perturbation scenarios, and cells exceeding a fixed reference score in at least 80% of scenarios, covering 254,887.07 ha, were carried forward. After protected land was removed, 1267 contiguous zones were extracted; 156 exceeded a 50 ha minimum‐area threshold and 16 were non‐dominated on suitability, exceedance frequency, and compactness, covering 19,461.56 ha. Groundwater depths at 33 wells showed no significant association with the suitability score and are used as a plausibility check. The framework links criterion redundancy and influence, classification stability, weight uncertainty, spatial contiguity, and planning constraints within a sequential decision‐support workflow.