Electrical Resistivity Tomography for collapse susceptibility screening in karst cenote environments: A case study in the Yucatán Peninsula, Mexico

Sudden collapses in karst terrains pose a significant disaster risk, threatening people, infrastructure, and culturally and or economically important sites. The Yucatán Peninsula, one of the largest karst aquifer systems in the world, is particularly exposed to dissolution processes that can culminate in sudden collapse; its limestone platform hosts thousands of cenotes whose roofs and walls can fail with little or no warning at the surface. This study develops a rapid, non-invasive framework for evaluating collapse susceptibility in cenote environments using Electrical Resistivity Tomography (ERT) method. Two contrasting study sites near Valladolid town in the Yucatán state were analyzed: the newly exposed Xocén cenote, revealed after a natural roof collapse in a Maya community, and the mature Saamal cenote, where localized cliff failures indicate continuing instability. A total of three profiles at the Xocén cenote and five profiles at the Saamal cenote were processed and inverted, yielding final RMS misfits of 1.97%–2.54% and 1.45%–2.89%, respectively. The results of both cenotes were interpreted in terms of mechanically relevant subsurface domains such as competent limestone, dissolution-weakened and saturated zones, fracture-controlled infiltration pathways, and cavity-prone rock masses, these domains were translated into qualitative collapse-susceptibility classes and monitoring-priority areas, the main emerged features associated with present or incipient instability are shallow weathered veneers, infiltration corridors and sharp resistivity contrasts. At Xocén, the data indicate subsurface weakening linked to altered limestone and conductive infiltration pathways beneath a more competent shallow cover; at Saamal, they show that even long-exposed cenotes can retain structurally vulnerable interfaces beneath apparently stable ground. The novelty of this work lies in the comparative assessment of two contrasting karst-failure settings, and in translating the resulting resistivity domains into mechanically relevant classes and monitoring-priority sectors. At Saamal, the independently inverted sections were additionally interpolated into a three-dimensional volume using a feed-forward neural network, which is used here as a visualization aid for the continuity of the resistivity domains between profiles rather than as an independent source of subsurface information. The obtained results show that ERT can serve as a practical, screening-level tool for post-collapse assessment, hazard zoning, monitoring prioritization, and risk-informed land-use management in karst communities and tourist areas.