Dolines are important features strongly influencing the outcomes of groundwater vulnerability maps, subsidence risk and land use studies. Their relationship with subsurface features like epikarst, stresses the importance of doline mapping for environmental and hydrological management strategies. Current methodologies to map dolines from elevation models apply morphometric attributes on depressions, including a depth threshold, to filter depressed areas and to define dolines. However, the use of a single threshold tends to overlook dolines located in already depressed areas. In this work a new geographic information systems (GIS)-based methodology is proposed to identify karst depressions within digital elevation models, applying a multidepth threshold approach. The method statistically classifies depression intervals to identify dolines at variable depths. The method was tested in the Yucatan karst, displaying a final accuracy of 63% after testing different parameters. The results are affected by false positives due to the impossibility of verifying by imagery 190 possible dolines in areas of dense vegetation. Nevertheless, out of 655 estimated dolines, 464 match those located by imagery giving sensitivity and precision values of 85% and 71%, respectively. Comparing this methodology against single threshold outcomes, improvement is evident in doline mapping. Notwithstanding, its application and performance with lower and higher resolution elevation models must be investigated.