Vulnerability to ground deformation phenomena in the city of Tuzla (BiH): a GIS and multicriteria approach

Ground deformation phenomena affecting the town of Tuzla (Bosnia Herzegovina), as consequence of underground salt deposits exploitation, have been surveyed since 60 years ago. The growing of cavities beneath the town and successive surface deformations have induced several hazard factors such as ground subsidence, deep and shallow fracturing and intense groundwater table oscillations. Fortunately, this processes had no catastrophic consequences such as sinkholes developing. Previous studies analyzed the historical database of topographic and piezometric data and highlighted a cumulative subsidence rate up to 12 meters over a period spanning from the 1956 to the 2003. Fractures arose as obvious result of ground deformation and caused damages and demolitions to thousands of buildings, with almost 15,000 people evacuated. Nowadays, in order to diminish the deformation processes, a strong reduction in brine withdrawal has been introduced by the local authorities. This caused an uplift of the water table which actually makes the spatial pattern of deformation more complex. In fact, in the most urbanized area, the sinking phenomena are now turning into uplift displacements. The presence of such a hazard required an urban vulnerability assessment able to support local authorities in the forthcoming planning procedure. The vulnerability assessment of urban elements over a wide area, potentially affected by deformation phenomena, has been accomplished by an integration between GIS solutions and a multicriteria approach based on ELECTRE-TRI methodology. The vulnerability procedure is able to define a discrete map where the 50 9 50 m width elementary cells report the most vulnerable parts of the town. In particular, the ELECTRE-TRI methodology is used to combine each vulnerable urban element with the others, expressed as vector data in a geographical dataset. The multicriteria approach is performed outside the GIS environment by the ELECTRE-TRI 2.0 package and requires a preliminary rasterization of the used dataset.