Optimization of a sampling well network based on the geostatistical analysis of hydrochemical parameters

Settling decision support tools for water quality management is one of the most important goals in countries where water resources are scarce and polluted. The existing networks for monitoring the hydrochemical parameters in groundwater can be extensively used to achieve water quality information. Parameters estimation based on sampled data strongly influences managerial choices. Monitoring networks are normally designed considering the total number of available \vells drilled in an aquifer. This procedure often increases the sampling cost but does not reduce the uncertainty related to the estimation. A methodology has been developed useful to reduce to the minimum the number of wells in an existing monitoring network, keeping the estimation uncertainty below a desired threshold. In geostatistics, the spatial and temporal behavior of hydrochemical parameters in groundwater can be studied by means of coregionalization models. If the spatial structure can be assumed persistent in time, the characteristic parameters of the auto-and cross-variograms (model type, sill, and range) can be evaluated using data from previous sampling campaigns performed in different periods. The methodology strongly depends on the assumption of temporal persistence of the spatial behavior of the considered parameter. In order to take into account the natural variability of the considered phenomena the methodology has been investigated by increasing and decreasing the variogram parameters in an application to a real case, the monitoring network of the aquifer of the Lucca Plain, Central Italy.