A Classification Tree for Modeling Ground Fractures from Subsidence

Abstract

This article presents a classification tree with predictive and prescriptive capabilities for the management of ground fractures in a crowded suburb in Tláhuac, a municipality of Mexico City. The tree is trained with observations of fractures parametrized with basic geotechnical and geological variables and specifications of the urban environment where they manifest. With the trained tree, the complexity of the scenario affected by the subsidence phenomenon is cleared because the relations between parameters can be viewed easily. With the trained tree, the influence of stratigraphic arrangements (geotechnical properties), geological conditions, size of roads and inhabited units, and location of water-pumping infrastructure in the appearance and exacerbation of cracks in soils is recognized, and this offers citizens and government administrators the possibility of anticipating damages and working on programs for improving structures and foundations, including relocation programs for communities at risk.