Studying the Reliability of Estimating Groundwater Remediation Cost Based on Qualitative Time-Series Data for an Aquifer

Nowadays, limited availability and pollution of water resources have necessitated the use of lower-quality water resources requiring greater spending on remediation. In this research, the aquifer zone of the Dehloran Plain in western Iran was investigated. The reliability level and risk were studied in a combined approach with a Monte Carlo method, including the determination of water resource quality using artificial neural networks (ANN) and estimating the water remediation cost by reverse osmosis system analysis (ROSA) software. Accordingly, 400 categories of qualitative water resource data over a continuous 18-year period, including sulfate (SO), fluorine (F), magnesium (Mg), calcium (Ca), potassium (K), nitrate (NO), hydrogen carbonate (HCO), chlorine (Cl) and sodium (Na), were considered as effective parameters of water resource quality. Using these parameters, a qualitative prediction model of the aquifer zone was generated by ROSA and a three-layered ANN (19 neurons), with 0.8% error. Finally, the confidence level of remediation cost estimation was determined using the mathematical model extracted from the ANN (as the activator function) in the Monte Carlo method. The remediation cost was estimated in the range of [54.31 $/m, 54.47 $/m] with 95% reliability level by the Latin hypercube algorithm. This reduces the risk of investing in management decision-making and covers unexpected changes in water resource quality which affects the remediation cost. Furthermore, the correlation ratio of polluting parameters can be calculated based on sensitivity analysis, and can be used as a criterion to measure the effect of the variables on the reliability level range. It shows a 67% effect of calcium as the most effective ion.