Health-risk-based groundwater remediation system optimization through clusterwise linear regression

Performance evaluation of the ISMLR package for predicting the next day's influent wastewater flowrate at Kirie WRP

Soft-sensor applications for wastewater management can provide valuable information for intelligent monitoring and process control above and beyond what is available from conventional hard sensors and laboratory measurements. To realize these benefits, it is important to know how to manage gaps in the data time series, which could result from the failure of hard sensors, errors in laboratory measurements, or low-frequency monitoring schedules. A robust soft-sensor system needs to include a plan to address missing data and efficiently select variable(s) to make the most use of the available information. In this study, we developed and applied an enhanced iterated stepwise multiple linear regression (ISMLR) method through a MATLAB-based package to predict the next day's influent flowrate at the Kirie water reclamation plant (WRP). The method increased the data retention from 77% to 93% and achieved an adjusted R² up to 0.83 by integrating with a typical artificial neural network.

Environmental- and health-risk-induced remediation design for benzene-contaminated groundwater under parameter uncertainty: a case study in Western Canada

This study proposes an environmental- and health-risk-induced remediation design approach for benzene-contaminated groundwater. It involves exposure frequency and intake rates that are important but difficult to be exactly quantified as breakthrough point. Flexible health-risk control is considered in the simulation and optimization work. The proposed approach is then applied to a petroleum-contaminated site in western Canada. Different situations about remediation durations, public concerns, and satisfactory degrees are addressed by the approach. The relationship between environmental standards and health-risk limits is analyzed, in association with their effect on remediation costs. Insights of three uncertain factors (i.e. exposure frequency, intake rate and health-risk threshold) for the remediation system are also explored, on a basis of understanding their impacts on health risk as well as their importance order. The case study results show that (1) nature attenuation plays a more important role in long-term remediation scheme than the pump-and-treat system; (2) carcinogenic risks have greater impact on total pumping rates than environmental standards for long-term remediation; (3) intake rates are the second important factor affecting the remediation system's performance, followed by exposure frequency; (4) the 10-year remediation scheme is the most robust choice when environmental and health-risk concerns are not well quantified.

Stochastic goal programming based groundwater remediation management under human-health-risk uncertainty

An optimal design approach for groundwater remediation is developed through incorporating numerical simulation, health risk assessment, uncertainty analysis and nonlinear optimization within a general framework. Stochastic analysis and goal programming are introduced into the framework to handle uncertainties in real-world groundwater remediation systems. Carcinogenic risks associated with remediation actions are further evaluated at four confidence levels. The differences between ideal and predicted constraints are minimized by goal programming. The approach is then applied to a contaminated site in western Canada for creating a set of optimal remediation strategies. Results from the case study indicate that factors including environmental standards, health risks and technical requirements mutually affected and restricted themselves. Stochastic uncertainty existed in the entire process of remediation optimization, which should to be taken into consideration in groundwater remediation design.

Development of a clusterwise-linear-regression-based forecasting system for characterizing DNAPL dissolution behaviors in porous media

Groundwater contamination by dense non-aqueous phase liquids (DNAPLs) has become an issue of great concern in many industrialized countries due to their serious threat to human health. Dissolution and transport of DNAPLs in porous media are complicated, multidimensional and multiphase processes, which pose formidable challenges for investigation of their behaviors and implementation of effective remediation technologies. Numerical simulation models could help gain in-depth insight into complex mechanisms of DNAPLs dissolution and transport processes in the subsurface; however, they were computationally expensive, especially when a large number of runs were required, which was considered as a major obstacle for conducting further analysis. Therefore, proxy models that mimic key characteristics of a full simulation model were desired to save many orders of magnitude of computational cost. In this study, a clusterwise-linear-regression (CLR)-based forecasting system was developed for establishing a statistical relationship between DNAPL dissolution behaviors and system conditions under discrete and nonlinear complexities. The results indicated that the developed CLR-based forecasting system was capable not only of predicting DNAPL concentrations with acceptable error levels, but also of providing a significance level in each cutting/merging step such that the accuracies of the developed forecasting trees could be controlled. This study was a first attempt to apply the CLR model to characterize DNAPL dissolution and transport processes.

Quasi-Monte Carlo based global uncertainty and sensitivity analysis in modeling free product migration and recovery from petroleum-contaminated aquifers

This paper presents a global uncertainty and sensitivity analysis (GUSA) framework based on global sensitivity analysis (GSA) and generalized likelihood uncertainty estimation (GLUE) methods. Quasi-Monte Carlo (QMC) is employed by GUSA to obtain realizations of uncertain parameters, which are then input to the simulation model for analysis. Compared to GLUE, GUSA can not only evaluate global sensitivity and uncertainty of modeling parameter sets, but also quantify the uncertainty in modeling prediction sets. Moreover, GUSA's another advantage lies in alleviation of computational effort, since those globally-insensitive parameters can be identified and removed from the uncertain-parameter set. GUSA is applied to a practical petroleum-contaminated site in Canada to investigate free product migration and recovery processes under aquifer remediation operations. Results from global sensitivity analysis show that (1) initial free product thickness has the most significant impact on total recovery volume but least impact on residual free product thickness and recovery rate; (2) total recovery volume and recovery rate are sensitive to residual LNAPL phase saturations and soil porosity. Results from uncertainty predictions reveal that the residual thickness would remain high and almost unchanged after about half-year of skimmer-well scheme; the rather high residual thickness (0.73-1.56 m 20 years later) indicates that natural attenuation would not be suitable for the remediation. The largest total recovery volume would be from water pumping, followed by vacuum pumping, and then skimmer. The recovery rates of the three schemes would rapidly decrease after 2 years (less than 0.05 m(3)/day), thus short-term remediation is not suggested.

Quantitative effects of composting state variables on C/N ratio through GA-aided multivariate analysis

It is widely known that variation of the C/N ratio is dependent on many state variables during composting processes. This study attempted to develop a genetic algorithm aided stepwise cluster analysis (GASCA) method to describe the nonlinear relationships between the selected state variables and the C/N ratio in food waste composting. The experimental data from six bench-scale composting reactors were used to demonstrate the applicability of GASCA. Within the GASCA framework, GA searched optimal sets of both specified state variables and SCA's internal parameters; SCA established statistical nonlinear relationships between state variables and the C/N ratio; to avoid unnecessary and time-consuming calculation, a proxy table was introduced to save around 70% computational efforts. The obtained GASCA cluster trees had smaller sizes and higher prediction accuracy than the conventional SCA trees. Based on the optimal GASCA tree, the effects of the GA-selected state variables on the C/N ratio were ranged in a descending order as: NH₄+-N concentration>Moisture content>Ash Content>Mean Temperature>Mesophilic bacteria biomass. Such a rank implied that the variation of ammonium nitrogen concentration, the associated temperature and the moisture conditions, the total loss of both organic matters and available mineral constituents, and the mesophilic bacteria activity, were critical factors affecting the C/N ratio during the investigated food waste composting. This first application of GASCA to composting modelling indicated that more direct search algorithms could be coupled with SCA or other multivariate analysis methods to analyze complicated relationships during composting and many other environmental processes.

A coupled simulation-optimization approach for groundwater remediation design under uncertainty: an application to a petroleum-contaminated site

This study provides a coupled simulation-optimization approach for optimal design of petroleum-contaminated groundwater remediation under uncertainty. Compared to the previous approaches, it has the advantages of: (1) addressing the stochasticity of the modeling parameters in simulating the flow and transport of NAPLs in groundwater, (2) providing a direct and response-rapid bridge between remediation strategies (pumping rates) and remediation performance (contaminant concentrations) through the created proxy models, (3) alleviating the computational cost in searching for optimal solutions, and (4) giving confidence levels for the obtained optimal remediation strategies. The approach is applied to a practical site in Canada for demonstrating its performance. The results show that mitigating the effects of uncertainty on optimal remediation strategies (through enhancing the confidence level) would lead to the rise of remediation cost due to the increase in the total pumping rate.