Formulating Convolutional Neural Network for mapping total aquifer vulnerability to pollution

Aquifer vulnerability mapping to pollution is topical research activity, and common frameworks such as the basic DRASTIC framework (BDF) suffer from the inherent subjectivity. This paper formulates an artificial intelligence modeling strategy based on Convolutional Neural Network (CNN) to decrease subjectivity. This formulation considers three definitions of intrinsic, specific, and total vulnerabilities. Accordingly, three CNN models are trained and tested to calculate IVI, SVI, and TVI, respectively referring to the intrinsic, specific, and total vulnerability indices. The formulation is applied in an unconfined aquifer northwest of Iran and delineates hotspots within the aquifer. The area under curve (AUC) values derived by the receiver operating curves evaluate the vulnerability indices versus nitrate concentrations. The AUC values for BDF, IVI, SVI, and TVI are 0.81, 0.91, 0.95, and 0.95, respectively. Therefore, CNNs significantly improve the results compared to BDF, but IVI, SVI, and TVI have approximately identical performances. However, the visual comparison between their results provides evidence that significant differences exist between the spatial patterns despite identical AUC values.

A modified SINTACS method for groundwater vulnerability and pollution risk assessment in highly anthropized regions based on NO3- and SO42- concentrations

Groundwater vulnerability and risk assessment are worldwide tools in supporting groundwater protection and land planning. In this study, we used three of these different methodologies applied to the Campanian Plain located in southern Italy: SINTACS, AVI and LOS. However, their capability to describe the observed chemical pollution of the area has resulted quite poor. For such a reason, a modified SINTACS method has been then implemented in the area in order to get a more reliable view of groundwater vulnerability. NO₃^- and SO₄^2- from more than 400 monitoring wells were used for specific vulnerability assessment. Land use was chosen as key parameter to infer the risk of groundwater pollution in our area. The new methodology seems to show a higher correlation with observed NO₃^- concentrations and a more reliable identification of aquifer's pollution hot spots. The main sources of NO₃^- were found in sub-urban areas, where vulnerability and risk are higher than in other areas. Otherwise due to reducing conditions triggered by the presence of elevated sedimentary organic matter and peat, concentrations below agricultural areas were lower than in sub-urban areas. The SO₄^2- specific vulnerability map showed a positive correlation with observed concentrations, due to geogenic and anthropogenic SO₄^2- sources present in the area. The combination of both NO₃^- and SO₄^2- derived risk maps becomes essential to improve the conceptual model of aquifer pollution in this severely anthropized area. The application of this new and original approach shed light on the strengths and weaknesses of each of the described previous methods and clearly showed how anthropogenic activities have to be taken into account in the assessment process.

Quantitative evaluation of specific vulnerability to nitrate for groundwater resource protection based on process-based simulation model

It has been proved that groundwater vulnerability assessment is an effective tool for groundwater protection. Nowadays, quantitative assessment methods for specific vulnerability are scarce due to limited cognition of complicated contaminant fate and transport processes in the groundwater system. In this paper, process-based simulation model for specific vulnerability to nitrate using 1D flow and solute transport model in the unsaturated vadose zone is presented for groundwater resource protection. For this case study in Jilin City of northeast China, rate constants of denitrification and nitrification as well as adsorption constants of ammonium and nitrate in the vadose zone were acquired by laboratory experiments. The transfer time at the groundwater table t50 was taken as the specific vulnerability indicator. Finally, overall vulnerability was assessed by establishing the relationship between groundwater net recharge, layer thickness and t50. The results suggested that the most vulnerable regions of Jilin City were mainly distributed in the floodplain of Songhua River and Mangniu River. The least vulnerable areas mostly appear in the second terrace and back of the first terrace. The overall area of low, relatively low and moderate vulnerability accounted for 76% of the study area, suggesting the relatively low possibility of suffering nitrate contamination. In addition, the sensitivity analysis showed that the most sensitive factors of specific vulnerability in the vadose zone included the groundwater net recharge rate, physical properties of soil medium and rate constants of nitrate denitrification. By validating the suitability of the process-based simulation model for specific vulnerability and comparing with index-based method by a group of integrated indicators, more realistic and accurate specific vulnerability mapping could be acquired by the process-based simulation model acquiring. In addition, the advantages, disadvantages, constraint conditions and applying prospects of the quantitative approach for specific vulnerability assessment were discussed.

Assessing the vulnerability of groundwater to pollution in Ireland based on the COST-620 Pan-European approach

The aim of the analysis was to assess the intrinsic and specific vulnerability of groundwater to pollution from pesticides in Ireland at the national scale. A methodology to incorporate the effect of groundwater recharge in vulnerability assessment is described which can be particularly useful for the evaluation of dilution of groundwater pollutants. A sensitivity analysis using Monte-Carlo simulation revealed that the most important parameters of the model were subsoil (ρ = 0.79) and topsoil (ρ = 0.72), which is in agreement with the current knowledge of the parameters that have a significant effect on groundwater vulnerability in Ireland. The intrinsic vulnerability assessment was verified using total organic carbon (TOC) concentration in groundwater, a novel approach for the validation of groundwater vulnerability methods at regional scales. A statistical analysis showed that TOC concentration was significantly different (p < 0.001) between watersheds classified as highly vulnerable and watersheds classified as less vulnerable, providing evidence that the developed method can effectively classify karst areas in terms of groundwater vulnerability.