Artificial Neural Network simulation of hourly 3 groundwater levels in a coastal aquifer system of the 4 Venice lagoon

20 21 Artificial Neural Networks (ANNs) have been successfully employed for predicting and 22 forecasting groundwater levels up to some time steps ahead. In this paper, we present an 23 application of feed forward neural networks (FFNs) for long period simulations of hourly 24 groundwater levels in a coastal unconfined aquifer sited in the Lagoon of Venice, Italy. After 25 initializing the model with groundwater elevations observed at a given time, the developed 26 FNN should able to reproduce water level variations using only the external input variables, 27 which have been identified as rainfall and evapotranspiration. To achieve this purpose, the 28 models are first calibrated on a training dataset to perform 1-hour ahead predictions of future 29 groundwater levels using past observed groundwater levels and external inputs. Simulations 30 are then produced on another data set by iteratively feeding back the predicted groundwater 31 levels, along with real external data. The results show that the developed FNN can accurately 32 reproduce groundwater depths of the shallow aquifer for several months. The study suggests 33 that such network can be used as a viable alternative to physical-based models to simulate the 34 responses of the aquifer under plausible future scenarios or to reconstruct long periods of 35 missing observations provided past data for the influencing variables is available. 36