Abstract:
This paper evaluates the feasibility of using an Artificial Neural Networks (ANNs) methodology for estimating the groundwater level in some piezometers placed in an aquifer in northwestern Iran. This aquifer is complex and has a high water level in urban areas. Spatiotemporal groundwater level simulation in a multilayer aquifer is regarded as a difficult subject in hydrogeology due to complexity and different aquifer materials. In the present research the performance of different neural networks for groundwater level forecasting is examined, in order to identify an optimal ANN architecture that can simulate selected piezometer water levels and provide acceptable predictions up to 24 months ahead. Six different types of network architectures and training algorithms are investigated and compared in terms of model prediction efficiency and accuracy. The results of different experiments show that accurate predictions can be achieved with a standard feedforward neural network trained with the Levenberg—Marquardt algorithm. Obtained structure and spatial regression relations of the ANN parameters (weights and biases) are used for spatiotemporal model presentation. It was found in this study that ANNs provided accurate predictions when an optimum number of spatial and temporal inputs were included into the network, and that the network with lower lag consistently produced better performance.
