Possibilities of Applying Multivariable Regression in Groundwater Data Series along a Riverbank Area

Abstract Groundwater modelling needs a large number of piezometer well data, which is unfortunately not always available as earlier registered data series are not always continuous. The reconstruction of data series with gaps was previously examined on a small riverside pilot area with the help of multiple regression, this method has now been tested at a 50-times larger environment. The results show that the reconstruction of the data series with this method works with average monthly groundwater levels and that applying multiple regression with the independent variables being one of the wells and the river optimizes the accuracy of the calculated data series, even if the relationships between the river and the wells are weak. The effect of the multiple regression on the accuracy is greater if the data series is sparser.