Optimization of a groundwater quality campaign utilizing the NSGA-II with preference ordering algorithm, contamination risk maps and well availability.

In Paraguay, the Patiño aquifer (1173 km2) located under the Metropolitan Area of Asuncion, serves a total of 2,976,400 inhabitants (43% of the total population of the country). Within this area there are an estimated 8,000 potential sources of pollution while 93% of the households have no sanitary treatment and depend mostly on spetic tanks or cesspools.

This study first presents the development of 2 contamination risk maps of the aquifer for Total Nitrogen (TN) and Total Coliforms (TC) utilizing a modified version of the DRASTIC model. The modified model incorporates parameters such as land use, density of cesspools and major transport routes as indicators. These maps were calibrated and validated with existing measurements. In order to understand the calibration the 2 maps were compared with each other and only 20% of the area were found to be similar.

Given that the highest indices (70-100) were found in the country's capital and its surroundings cities that cover an area of 567 km2, and have around 2800 deep wells that extract water for residential use, with little or no treatment, a monitoring campaign where 19 parameters were analyzed was designed and implemented. The design of the campaign was developed by optimizing 4 parameters: (1) Well distribution to cover as much area as possible, (2 and 3) sites with with high indices of TN and TC contamination risks, and (4) the accessibility of wells in order to extract samples (i.e. public or privately owned wells).

Due to economic and practical limitations, only 70 of the 2800 wells were selected utilizing the the Non-dominated Sorting Genetic Algorithm II with Preference Ordering [5] was implemented to solve it. The proposed algorithm selected 60 wells which were publicly accessible, 49 wells that are located in places with the highest indices of contamination risk and cover 74% of the study area. The results of the water quality campaign are now being analyzed in order to characterize the actual situation of the aquifer and develop a monitoring scheme utilizing a combination of numerical models, citizen monitoring, and wireless water quality measurements.