A data driven model for spatio-temporal estimation of shallow water table depth in soils

Data from a monitoring network were used to develop a data driven model for predicting water table depth in space and time. Records of 160 piezometer sites available from 1997 to 2010 were analyzed to detect the overall temporal trend in water table depth in a relevant agricultural area in Northern Italy. Evolutionary Polynomial Regressions (EPR) were used to calibrate a predictive tool based on climatic data and the records from 47 selected sites between 2004 and 2009 (N= 5611). The model was validated against the WT depths observed in 15 independent sites between 2005 and 2010 (N= 2052). Validation resulted in a mean absolute error of 30.0 cm (R2 = 0.65). The general model was extended to the whole area, using the geostatistical estimates of the average water table depth as input, providing spatio-temporal maps of the water table depth at any give date.