Data-driven decision management of urban underground infrastructure through groundwater-level time-series cluster analysis: the case of Milan (Italy)

Abstract

The significant increase in urbanization has resulted in greater use of the subsurface in urban planning and, therefore, increased interaction between groundwater and underground infrastructure. Numerical models are the primary tool adopted to manage the resulting problems; however, their construction is time- and cost-consuming. Groundwater-level time-series analysis can be a complementary method, as this data-driven approach does not require an extensive understanding of the geological and boundary conditions, even if providing insights into the hydrogeologic behaviour. Thus, a data-driven approach was adopted to analyse groundwater time-series of the shallow aquifer, occupied by several underground structures, beneath Milan city (Northern Italy). Statistical (Mann-Kendall and Sen’s slope estimator, autocorrelation and cross-correlation, hierarchical cluster analysis) and geospatial techniques were used to detect the potential variables influencing the groundwater levels of 95 monitoring wells, covering the period 2005–2019. A general rising trend of the water table was identified, with local hydrogeologic differences in the western and southernmost areas. Based on time-series analysis results, four management areas have been identified. These areas could act as future geographic units with specific groundwater management strategies. In particular, subsurface public car parks can be classified with respect to groundwater flooding as (1) not submerged, (2) possibly critical, or (3) submerged at different groundwater conditions. According to these outcomes, targeted guidelines for constructing new car parks have been elaborated for each management area. The methodology proved to be efficient in improving the urban conceptual model and helping stakeholders design the planned underground development, considering groundwater aspects.