QUANTITATIVE ASSESSMENT OF SINKHOLE SUSCEPTIBILITY AND HAZARD. THE CASE STUDY OF THE EBRO VALLEY EVAPORITE KARST (NE SPAIN)

A probabilistic approach to sinkhole susceptibility and hazard modelling has been developed and independently tested in a 50 km² sector of the Ebro valley alluvial evaporite karst. Two main morpho-hydrological domains can be differentiated in the study area: the floodplain (discharge area), and the lowest terrace of the Ebro River located on both flanks of the valley (perched aquifers). Three genetic types of sinkholes have been identified and analysed: large sagging sinkholes (Type 1) and large collapse sinkholes (Type 2) in the floodplain, and small cover collapse sinkholes (Type 3) in the southern terrace. A spatial database composed of a sinkhole inventory and 29 thematic layers related to potential conditioning factors has been implemented in a GIS. Multiple susceptibility models were generated analysing the statistical relationships between the sinkholes and combinations of factors using Favourability Functions (empirical likelihood ratio models). The application of different validation strategies to the models has allowed us to assess their prediction capability and to determine the variables that best explain the spatial distribution of each sinkhole type. In the best susceptibility models of sinkholes types 1, 2 and 3, 100% of the sinkholes in the validation sample occur within the 20, 10 and 25% of the most susceptible area. These results indicate that probabilistic models can provide good predictions on the future spatial and temporal distribution of sinkholes in the study area. In a subsequent step, the best susceptibility model of sinkhole type 3 has been transformed into a hazard map considering the mean size and number of sinkholes formed in a subsequent year. This hazard model, based on an incomplete inventory, provides a minimum probability for a pixel to be affected by a sinkhole in a given period (temporal-spatial probability). The obtained susceptibility and hazard models could be very useful for the effective application of preventive and corrective mitigation measures.