A Multi-objective Optimization Concept for Risk-based Early-warning Monitoring Networks in Well Catchments

Abstract

Groundwater wells are often protected by restricted land use within wellhead protection zones. Unfortunately, one cannot restrict land use in the entire catchment (especially in urban areas), and there is uncertainty in wellhead delineation. Thus, nearly all well catchments have an entire inventory of risk sources. Each of these risk sources may fail at any time, release contamination and affect the well earlier or later. In fact, most catchments are equipped with some form of monitoring network. Such networks, however, often grow historically, follow various purposes that changed over time, and thus are often suboptimal (if not even inadequate) for rigorous risk control. In this work, we propose a concept to plan monitoring networks through multi-objective optimization. The different objectives are minimal costs, maximal probability to detect all possible contaminants once they entered the aquifer, and earliest possible detection. Also, risk sources that are classified as severe versus medium or tolerable should be treated with different priorities. Therefore, we propose to treat detection probability and early-warning time as separate objectives for each risk class. The concept will allow catchment managers to obtain optimal monitoring networks for risk control, and to gain insight into the costs of certainty, the costs of early warning, and the costs of covering top risks versus the luxury situation of controlling even minor risks.

Reconnecting Stochastic Methods With Hydrogeological Applications: A Utilitarian Uncertainty Analysis and Risk Assessment Approach for the Design of Optimal Monitoring Networks

“…For simplification reasons, all possible contamination sources are assumed to be point sources with a pulse release of their contaminant and we assume conservative tracer transport. The groundwater flow was simulated with the standard‐Galerkin finite element code already used in Nowak et al () (for the discretization of the domain we used a regular grid, cell size: ). As already stated in section 2.2, we used the particle‐tracking‐random‐walk method for all transport simulations.…”

“…To aid in this effort, we developed a problem-specific visual representation of multiobjective optimization (Figure 4; we already used this representation in Bode et al, 2016 andNowak et al, 2015. Different visualization techniques of Pareto fronts can be found in, e.g., Ibrahim et al, 2016).…”

Optimization for Early-Warning Monitoring Networks in Well Catchments Should Be Multi-objective, Risk-Prioritized and Robust Against Uncertainty

“…As already described in Nowak et al (2015), the considered three objectives are competing. Multi-objective optimization will reveal the competition and trade-offs between the goal attainment levels in these three objectives, and will provide a transparent decision basis.…”

“…Thus, to the best knowledge of the authors, there is no study to date that approaches the problem of robust optimal monitoring for multiple risk sources as an early-warning system in drinking water well catchments. In Nowak et al (2015), the authors already described the general concept behind our current approach, yet without any implementation, without application to a synthetic test case, and without the concept for entirely unknown risk sources. Also, we refine now the definition of our objective functions.…”