طراحی شبکه پایش کیفی آب‌های زیرزمینی با استفاده از روش بهینه‌سازی الگوریتم ژنتیک

Introduction:
A spatial distribution and accuracy of groundwater quality data is required for management of groundwater resources. These data are usually collected from monitoring wells which are spatially distributed in the studied aquifer. In the design of the monitoring network, the minimum number of monitoring wells with an optimum spatial distribution is necessary to ensure a cost efficiency. Therefore, the configuration of the wells distribution and their number in groundwater monitoring networks are an important problem for optimizing groundwater issues. This study was targeted to find an optimal monitoring network with minimum number of wells in the Guilan’s aquifer so that provides sufficient spatial distribution on groundwater quality. Salinity is one of the most important criteria for the quality of groundwater which is measured by using of parameters such as Total Soluble Solids (TDS), Chloride ion (Cl) and Electrical Conductivity (EC). Hence, EC was selected as a quality parameter in the design of the monitoring network in this study.
Materials and methods:
Genetic optimization algorithm (GA) was used to search for optimal quality monitoring network. In this method, a possible network of monitoring wells located in the aquifer considered for each “chromosome”. Then each monitoring well in this network is represented by a binary bit. Finally, they are coded by bit value equals to 1 for well that was selected for the network or by bit value equals to 0 for well that was not selected for the network. In this paper, two conflicting objective functions are simultaneously solved. The first objective function is the maximization of the match between the interpolated EC distributions obtained from data of the all available monitoring wells and the wells from the newly generated network. The match is evaluated by using of the Nash-Sutcliffe (NS) model efficiency. The second objective is the minimization of the number of the monitoring wells in the newly generated network by considering cost-related constraints. These two objectives are integrated in a single objective function where different combinations of both objectives are investigated by considering two cases.
Results and discussion:
The results showed that the relative importance of each objective is expressed using the weighting coefficient, w. It was found that the solution of the optimization is very dependent on the selection of w. Therefore, a w value that are resulted by the most balanced solution should be selected. Most balanced means that the trade-off between cost and spatial distribution is most acceptable. To choose the most solution, it is highly recommended to evaluate additional performance indicators besides NS coefficient such as RMSE, PBIAS, the regression coefficient and standard deviation. Additionally, mean values of EC observed in the optimized network are higher than those in all monitoring wells. Therefore, it could be clearly concluded that the optimized network provides groundwater quality data from more polluted areas.
Conclusion:
The results showed that the optimization approach significantly reduces the number of monitoring wells with spatial distribution of the EC values. Additionally, the monitoring network was optimized such a way that sampling points were removed from less polluted areas and were selected in areas with higher pollutant concentrations. The optimal design of the monitoring network should be performed periodically. Since monitoring efficiency is expected to change when the data of the new wells become available, a re-evaluation of the optimized monitoring network considering the addition of new wells every few years may help the determination of the long-term effectiveness of a groundwater quality monitoring program.