Institutional Repository
Technical University of Crete
EN  |  EL



My Space

Determination of groundwater nitrate pollution in the extended area of Asopos River using ANNs

Stylianoudaki Christina

Full record

Year 2019
Type of Item Master Thesis
Bibliographic Citation Christina Stylianoudaki, "Determination of groundwater nitrate pollution in the extended area of Asopos River using ANNs", Master Thesis, School of Environmental Engineering, Technical University of Crete, Chania, Greece, 2019
Appears in Collections


The scope of the present study is the estimation of groundwater pollution by nitrates (NO3-), using artificial neural networks (ANNs).Nitrate pollution is caused through the introduction of excessive amounts of nitrogen to surface and groundwater, mainly as a result of agricultural practices because of the overuse of nitrogen based fertilizers. NO3- are particularly mobile through water and soil, so excess nitrates from sewage and agricultural fertilizers easily make their way into groundwater and surface waters. The guideline value for nitrate in drinking water is set by Greek and international legislation at a concentration of 50 mg/l, but even at lower concentrations (greater than 25 mg/l), concerns are being raised about long-term use of drinking water.Modeling of nitrate pollution with the commonly used mathematical methods can be a particularly difficult process and requires very good knowledge of the geomorphology of the study area, which in most cases is characterized by heterogeneity. In addition, techniques for detecting and measuring nitrate concentrations in water are characterized by high cost and high demands of time, while portable devices used are not of sufficient accuracy. Furthermore, in the various methods used for chemical analysis of water, the detection of nitrates is affected by the presence of other ions, especially Cl-. Based on the aforementioned issues, this thesis examines the possibility of using neural networks for the assessment of concentrations of nitrate in groundwater.ANNs do not require the knowledge of the geomorphology characteristics of the area, which is hard to obtain, while it is possible to use variables as inputs parameters without knowing the relationships that condition them in the system being studied.For the aforementioned purpose of this thesis, three neural networks were developed using Matlab toolbox. For the first network, easily measurable field data were used. The parameters initially studied were pH, electrical conductivity, water temperature, air temperature and aquifer level. This model achieved a fairly good simulation (R=0.92259, NSE=0.8406). Then, for the better description of the underground system, it was chosen to include in the model, the percentage of land uses in a radius of 1000 m from each well. The performance of the model increased significantly by the use of these variables (R=0.97412, NSE=0.9481). In the third model, the simulation was based on typical water quality measurements. In particular, pH, electrical conductivity, HCO3-, Cl-, Ca2+, Mg2+, Na+, K+, and SO42- were used. This model achieved the best simulation (R=0.96545, NSE=0.987838), possibly due to the larger number of available data. Finally, scenarios of climate change and land use changes were examined, in order to assess their impact on NO3- levels, using the second model, which is considered to have included factors that significantly affect nitrate transport into the geoenvironment. The results of the scenarios demonstrate the important contribution of agricultural and industrial activities to increased nitrate pollution.The models achieved a good simulation, which indicates that they are a possible useful tool for the estimation of groundwater pollution by nitrates and so, they may constitute the basis for the development of groundwater management plans.

Available Files