Institutional Repository
Technical University of Crete
EN  |  EL



My Space

Groundwater simulation using artificial neural networks (ANNs) and the Princeton Transport Code (PTC) – Performance Comparison

Chatzakis Alexandros

Full record

Year 2014
Type of Item Diploma Work
Bibliographic Citation Alexandros Chatzakis, "Groundwater simulation using artificial neural networks (ANNs) and the Princeton Transport Code (PTC) – Performance Comparison", School of Environmental Engineering, Technical University of Crete, Chania, Greece, 2014
Appears in Collections


The purpose of this study is to simulate the hydraulic head change in an aquifer using two different models. The first model is based on Artificial Neural Networks (ANNs), and was implemented using the ANN MATLAB toolbox. The simulation period is from November of 2008 until October 2012. The input to the ANN includes meteorological parameters, such as rainfall, snowfall, and temperature, as well as flow data of the river running through the study area. For rainfall, snowfall and flow data of the river, correlation analysis was conducted in order to define the appropriate time lags in which the parameter affects the hydraulic head and the use of multiple time lag was considered necessary. Two different cases were studied. One case included snowfall as input parameter, while the other case did not. For each case, two ANN architectures were considered. One of them was consisted by 1 hidden layer and the other architecture included 2 hidden layers. One ANN is trained for each one of the 30 observation wells, where data are available, and the hydraulic head change of a daily time step is simulated. The case that did not contain snowfall as input parameter proved to have significantly better performance than the case that contained snowfall.The second model is the groundwater flow simulator PTC (Princeton Transport Code), combined with the ArgusOne GIS program. The simulation period is one year starting in December of 2008. Digital maps of elevation, geology of the area and position of the observation wells are imported in the program.The model is calibrated using meteorological and geological data and by setting as initial conditions the observed field measurements. The output of the model is the simulated hydraulic heads. The performance of both models is evaluated and compared based on the observed field data in the Bayern region, Germany. As conclusion, ANNs were found to be more accurate in point simulation, while PTC although it is not very accurate in simulating the hydraulic head in points where data were available, gives a better general figure of the groundwater level in a large area (total water balance.The selection of the most appropriate modelfor the simulation at hand, depends on the further use of it, as well as the available data.

Available Files