You are here:
A Decision Analysis Framework for Groundwater Remediation
Overview: A policy analyst for a contaminated groundwater site faces a formidable list of issues on social values. The choice of a remedial strategy involves tradeoffs between the large expenses of action and the large and uncertain consequences of inaction -- consequences for health, for ecological quality, and for the availability of drinking water. The choice of a strategy also involves attitudes toward risk, the discounting of outcomes in the distant future, concerns for equity between the outcomes at different times, and concerns for equity between the outcomes for different groups. The proposed research will gain synergy by bringing together ideas that are emerging in the fields of the principal investigators. Groundwater hydrologists are recognizing that slow mobile/immobile mass-transfer processes impose constraints on aquifer remediation and that slow degradation processes can reduce organic contaminant levels over the long-term. Decision analysts are recognizing the importance of modeling social values that give significance to future generations and include concerns for equity. Objectives and Approach: The proposed research will provide a comprehensive framework for the evaluation of alternative strategies for the cleanup and containment of pollution at a contaminated groundwater site. The framework will include both the identified physical processes and the identified issues of social values. Decision analysis models will be added to groundwater hydrology models so that a policy analyst can relate the information on a contaminated site to a policy choice on remediation for the site. Issues of social values will be modeled such that the policy analyst can perform what-if analyses on the implications of different social values for the evaluation of the alternative strategies. Such an inclusive framework will require the development of new decision analysis methods: models for intertemporal equity, procedures that separate different issues of social values, and parametric functions for social values to provide for what-if analyses on the implications of social values for a policy choice. These new methods will be developed and will be combined with available methods such as models of risk aversion for the highly uncertain health effects of groundwater pollution and models of discounting that can assign significant value to outcomes for future generations. Work Plan and Expected Results: The work will be in three stages. First, we will develop a decision analysis framework that includes assessment procedures and parametric functions for social values on the issues of (1) discounting future outcomes, (2) risk aversion for health and environmental effects, (3) concern for equity between the outcomes at different times, and (4) concern for equity between different groups. Second, we will apply the framework to the contamination site at Lawrence Livermore National Labs. The hydrology and the behavior of the contaminants have already been extensively studied at this site, and detailed models have been developed for both regional transport and remediation. The first and second stages will proceed interactively. In the third stage, we will apply the framework to the general context of groundwater remediation. We will examine how social values and information on a specific site affect the choice of a remedial strategy, and how social values and regulatory constraints imply the distribution of limited monetary resources among different contaminant sites. Thus, we intend that the proposed framework will be a coherent means for studying national policy questions of groundwater remediation.
R825825Principal Investigators: Harvey, Charles F.
Harvey, Charles MTechnical Liaison:Research Organization:
Houston, University ofFunding Agency/Program:
EPA/ORD/ValuationGrant Year: 1997Project Period: October 1. 1997 to September 30, 2000Cost to Funding Agency: $205,000
- Project Reports
Objective: An analyst who is evaluating policy options on the remediation of a contaminated groundwater site or region needs methods from a range of disciplines. First, the analyst needs methods from groundwater hydrology. For instance, he may need such methods to evaluate the hydrology of the site and the extent (and future dispersal) of the contaminant. To evaluate a remedial strategy, the analyst also must use methods from other disciplines such as epidemiology, economics, and decision analysis. For instance, he may need methods to evaluate the time stream of costs, health effects, and environmental effects due to a remediation strategy. Moreover, the analyst may need to examine relevant issues of social values, for example, tradeoffs between the expenses of action and the uncertain consequences of inaction, attitudes toward risk, the discounting of outcomes in the distant future, concerns for equity between costs and benefits at different times, and concerns for equity between the outcomes for different stakeholder groups.
The objectives and results of this research project focused on the interdisciplinary needs described above. Our results are reported in the sections below. The objectives were of three types: (1) developing models in decision analysis that can aid environmental policy studies?in particular, models that are applicable to groundwater management and other environmental problems involving long-term and highly uncertain outcomes; (2) developing models in groundwater hydrology that can aid environmental policy studies?in particular, models that describe the slow mobilization of contaminants into groundwater and, thus, may describe the time constraints to clean up the contaminants; and (3) performing a comprehensive policy study that would utilize these models.
Summary/Accomplishments: Our work brought together ideas that are emerging in the fields of the two principal investigators. Groundwater hydrologists are developing models of slow mobile/immobile mass transfer processes and using these models to estimate constraints on aquifer remediation and to show how slow degradation processes can reduce organic contaminant levels over the long term. Decision analysts are developing models of social values that give importance to future generations and that include concerns for social risk, intertemporal equity, and equity between stakeholders. As described below, we contributed to these developments and we applied our ideas as appropriate in a comprehensive study of contaminated groundwater.
In the following sections, we describe the principle results of our research organized according to the objectives stated above.
Decision Analysis Models
Our work on decision analysis models has resulted in two published papers and three papers in preparation. The research is described below.
Aggregation of Individuals' Preference Intensities into Social Preference Intensity. This research has been published in a paper that develops a method for measuring distributions of health outcomes on a single scale. A decision between policies that affect people's health depends on their preferences regarding their own health and on social values regarding overall health. We construct a model for using people's preference intensities for health outcomes to aggregate measures of their health states (which range from perfect health to death) into an evaluation of the overall outcome. The model is analogous to that developed by J. Harsanyi (Cardinal welfare, individualistic ethics, and interpersonal comparisons of utility, Journal of Political Economy, 1955) that uses people's preferences among probability distributions to aggregate personal expected utilities into social expected utility. With the model based on people's preference intensities, a policy analyst can separate the two value issues that are directly involved?preferences among personal health states and social values for overall health states?from other issues such as beliefs about uncertain health outcomes.
Equity, Efficiency, Uncertainty, and the Mitigation of Global Climate Change. This research has been published in a paper that examines the discount factor assigned to the future outcomes of a policy option (e.g., a choice to mitigate damages from global climate change). Cost-benefit studies often analyze proposed projects by the "opportunity-cost criterion" that states that society should not spend resources on a proposed project if the resources could be spent on an investment project that would produce a greater benefit. The opportunity-cost criterion implies that the discount factor used to evaluate the proposed project should equal the ratio of the cost of the resources to the benefit of the investment project. We argue that the opportunity-cost criterion must be severely modified to evaluate a proposed policy to mitigate future environmental harms that extend over generations and are highly uncertain. The reason is that the hypothetical investment policy provides returns that are highly uncertain and add to the uncertainty of the future harms, whereas the mitigation policy reduces the uncertainty of the future harms. If society is risk averse, the investment policy's effect of increasing uncertainty and the mitigation policy's effect of decreasing uncertainty provide an influence in favor of the mitigation policy. The influence is further strengthened by uncertainty (e.g., disagreement among experts) as to the base welfare of future generations.
Social Value of Health: Evaluation Based on Preference Intensity. This research will be published in a paper that further develops the model in the first paper described above. It considers time streams of health states rather than fixed health states, and uses the model in that paper to develop a scale for measuring such time streams. The scale is sufficiently general so that with different assessments of individual and social values it becomes equivalent to well-known scales such as quality adjusted life years or it becomes a new scale. In either case, the foundation in terms of individuals' preference intensities and social preference intensities provides procedures for assessing a health scale.
Optimizing Groundwater Management With Hyperbolic Discounting. This dissertation examines how long-term groundwater management strategies depend on the choice of how to discount the future costs and benefits of groundwater. We find optimal time streams of groundwater withdrawal for two scenarios: an aquifer with finite recharge that is subject to dewatering, and an aquifer that suffers from salinization due to irrigated agriculture. For both problems, we show that the optimal pumping rate over time is qualitatively different for different types of discounting. So-called hyperbolic discounting, whose discount weights are of the form a(t) = b/(b + t) for some parameter amount b > 0, implies short-term pumping rates similar to those implied by constant-rate discounting and long-term pumping rates similar to those implied by no discounting. Furthermore, we find that when future costs and benefits are uncertain, constant-rate discounting results in optimal pumping rates that vary over time in a manner similar to the optimal pumping rates found from hyperbolic discounting applied to deterministic models.
Slow-Discounting for Natural Resource Policy Studies. This research differs from the dissertation described above in that it is far more general, focusing on time streams of the usage of a renewable resource (such as groundwater) and time streams of the usage of a nonrenewable resource (such as oil or metals). The limiting feature of the models is that the costs and benefits of the resource usage and the resource stock must be quadratic functions of usage rate and stock level. We find optimal usage and stock functions under a variety of assumptions on costs and benefits, and we compare the optimal functions implied by constant-rate discounting with those implied by a type of slow-discounting in which the discount weights are of the form a(t) = b/(b + t)2 for some parameter amount b > 0.
Models of Contaminant Mass Transfer and Time to Cleanup
Our work in groundwater hydrology and geostatistics models has resulted in two published papers and one paper in preparation. The research is described below.
Rate-Limited Mass Transfer or Macrodispersion: Which Dominates Plume Evolution at the Macrodispersion Experiment (MADE) Site? This research has been published in a paper that argues that slow diffusion of contaminants in and out of immobile regions is an important process that must be accounted for to predict contaminant behavior over long time periods. It contains an exhaustive analysis of recent solute transport experiments at the MacoDispersion Experiment (MADE) field site. The MADE aquifer has very heterogeneous hydraulic conductivity, unlike aquifers in earlier extensively instrumented solute transport experiments. We show that a rate-limited mass transfer model better explains the large-scale behavior of solute plumes, and we argue that this result may be generally applicable to other heterogeneous aquifers.
A Method for Estimating Distributions of Mass Transfer Rate Coefficients With Application to Purging and Batch Experiments. This research has been published in a paper that provides models for describing the time scales of contaminant mobilization. We develop a model that accounts for multiple time scales of mass transfer and describe a method for estimating model parameters from data. We show that by accounting for multiple time scales of mass transfer, this model can reproduce observed contaminant tailing, a behavior that often limits the rate of contaminant cleanup.
Optimal Management Schemes for Groundwater Pumping: The Effects of Valuing the Future. This research will be published in a paper that describes how contaminant mass transfer controls the time to cleanup of groundwater contamination. We consider the choice between an expensive option to remediate an aquifer as quickly as possible and a less costly option to hydraulically contain the contaminated groundwater that can delay cleanup by years and perhaps decades. Specifically, we focus on how contaminant mass transfer effects this choice. We consider the present value cost of remediation for a range of pumping rates bounded by the minimum rate to contain a contaminant plume from migrating offsite and the maximum obtainable without dewatering the aquifer. We combine realistic treatment cost models with models of solute transport towards a pumping well for contaminants subjected to rate-limited mass transfer and first-order degradation. Without rate-limited mass transfer, costs increase with increased pumping rate (i.e., low pumping rates [containment schemes] are the least-cost strategies). However, our model simulations indicate that initial nonequilibrium between mobile and immobile states may cause: (1) present value costs that monotonically decrease with increased pumping rates, or (2) present value costs that increase to a maximum at some intermediate pumping rate, then drop to a lower cost at higher pumping rates. Thus, for all the cases we considered, the optimal pumping rate is restricted to either the minimum rate that prevents the contaminant plume from migrating offsite, or the maximum possible rate, but never an intermediate rate between these two bounds. Simulations further show that: (1) in the presence of rate-limited mass transfer, the time since contamination is the primary determinant of whether containment or aggressive remediation is the least PVC strategy; (2) multiple time scales of mass transfer increase the expense of aggressive remediation, improving the relative attractiveness of containment strategies; and (3) these results are not qualitatively changed for the treatment cost functions of either granular activated carbon filtration or air stripping.
Our primary field application has been an evaluation of arsenic-contaminated groundwater and its health effects in Bangladesh. Shortly after we received the grant for this project, we became aware of a potential choice for a comprehensive policy study?an evaluation of the health effects of arsenic contamination in the groundwater wells in Bangladesh. We were attracted to this problem as our primary application because: (1) the magnitude of the problem is very large, possibly the most widespread case of natural poisoning in human history; (2) the health effects of arsenic, at the high levels found in Bangladesh, are relatively easy to measure; and (3) we have had experience in modeling the behavior of arsenic in groundwater.
However, we viewed the application as a departure from the original objectives of the project because certain groundwater hydrology models and decision analysis models that we intended to employ were not needed, and we wished to select methods according to their relevance rather than according to their origin in our research. We conferred with the appropriate officials in the EPA and gained their approval for this change from the project as originally proposed. We then developed and applied models as needed in the course of the Bangladesh study to address issues that we had not envisioned in the project proposal. The result has been that certain hydrology and decision analysis models that we have developed as part of objectives (1) and (2) were not used in objective (3). We believe, however, that these models will be useful in other environmental policy studies.
This work has resulted in the paper that brings together data and models from groundwater hydrology, geostatistics, demographics, epidemiology, and decision analysis. In the paper, we model the geographic distribution of arsenic concentration in groundwater throughout Bangladesh where about 50 percent of the four million wells drilled for household use produce water with arsenic concentrations in excess of 10 ppb (the WHO limit), and 30 percent of the wells produce water with arsenic concentrations in excess of 50 ppb (the current U.S. limit). Using available field surveys, we combine geologic mapping with classical variogram analysis. Regional mapping of the geology and geomorphology of Bangladesh is shown to explain much of the large-scale ( >10 km) spatial variability. Minimal spatial structure exists at scales less than 1 km, and a significant component of the small-scale spatial variability is explained by differences in well depths. We also estimate for each defined region a depth trend in arsenic concentration, and typically find concentration decreasing with depth.
We combine the geographic distributions of arsenic concentration in regions with census data to estimate population distributions of concentration in regions. Then, we use epidemiological data from Taiwan and Bangladesh to estimate dose-response functions for the health effects of certain types of arsenicosis and cancer, and we combine the dose-response functions with the population distributions to estimate the health effects due to groundwater arsenic, namely, the prevalences of arsenicosis and skin cancer and the incidences of internal cancers. We predict that with long-term exposure to present concentrations of arsenic, there will be prevalences of approximately 1,200,000 cases of hyperpigmentation, 600,000 cases of keratosis, and 150,000 cases of skin cancers, and an overall incidence of 3,000 fatalities per year for internal cancers of various types. As a remedial strategy, we consider the option of drilling deeper groundwater wells in selected regions of Bangladesh. We estimate that such a strategy could significantly reduce the health effects of drinking arsenic-contaminated groundwater, provided that arsenic concentrations at fixed depths remain constant over time.
A second field application has been a study of long-term groundwater management in the Gaza Strip. In contrast to the arsenic study, this study focuses on economic issues of water supply and combines economic analysis with groundwater modeling. Groundwater consumption on the Gaza strip now exceeds recharge, preventing high salinity agricultural runoff from being flushed to the sea and causing seawater intrusion at the coast. We construct an optimization model that couples the hydrology and water supply of Gaza with the economics of water?both agricultural and municipal. The model maximizes social welfare (the integrated difference between demand and supply) based on the assumption of partial equilibrium, as calculated using: (1) a finite difference model of groundwater flow and salt transport, (2) approximations of agricultural yield and profit as a function of water quantity and salinity, (3) municipal demand curves or willingness-to-pay curves, and, (4) the costs of pumping, transport, treatment, and potential environmental damage. We then employ this model to evaluate various water supply and agricultural strategies, and to consider which strategies may offer long-term, sustainable, solutions. We contrast the optimal solutions?found with conventional economic models?with the steady-state sustainable solutions, and show that the water management solutions are radically different. The steady-state solution, without added infrastructure, would impose severe consequences to the population's social welfare, but by definition does not degrade water resources for future generations. The conventional economic model would provide much greater short-term benefit, but would lead to further salinization and overpumping of the aquifer, degrading the aquifer beyond use for future generations.
Duarte K. Optimizing groundwater management with hyperbolic discounting. Ph.D. dissertation, 2001.
Hollenbeck K.J., Harvey C.F., Haggerty R. and Werth C.J. A method for estimating distributions of mass transfer rate coefficients with application to purging and batch experiments. Journal of Contaminant Hydrology, Volume 37, Issues 3-4, 15 April 1999, Pages 367-388.
Harvey CM. Aggregation of individuals' preference intensities into social preference intensity. Social Choice and Welfare 1999.
Harvey CM, Hammitt J. Equity, efficiency, uncertainty, and the mitigation of global climate change. To appear in Risk Analysis.
Publications and Presentations: Total Count: 20
Type Citation Journal Searches not available Journal Article Hammitt JK, Harvey CM. Equity, efficiency, uncertainty, and the mitigation of global climate change. Risk Analysis 2000;20(6):851-860. Journal Article Harvey, CF, Gorelick SM. Rate-limited mass transfer or macrodisperion: which dominates plume evolution at the macrodispersion experiment (MADE) site? Water Resources Research March 2000. not available Journal Article Huber A, Harvey CF. Sustainable water management in the Gaza Strip: a combined economic hydrologic mode. Water Resources Research. not available Presentation Harvey CF. A geostatistical and geochemical analysis of groundwater arsenic concentrations in Bangladesh: what should be done? Presented at the Lamont-Doherty Earth Observatory, Columbia University, February 2001. not available Presentation Yu W, Harvey C. Estimating the health effects of arsenic contaminated groundwater in Bangladesh: a combined geostatistical and epidemiological model. Presented at the American Geophysical Union Fall Annual Meeting, December 2000. not available Presentation Huber A, Harvey C. Possibilities for a sustainable water supply for the Gaza Strip. Presented at the American Geophysical Union Fall Annual Meeting, December 2000. not available Presentation Huber CF, Harvey, Rogers. Strategies for sustainable water management in the Gaza Strip. Presented to the World Water Foundation, April 2000. not available Presentation Huber A, Harvey C, Rogers P. Strategies for sustainable water management in the Gaza Strip. Presented at the Second World Water Forum, The Hague, March 17-22, 2000. not available Presentation Harvey CF. The arsenic crises in Bangladesh. Presented at the Interfaculty Initiative on Water and Health, Harvard School of Public Health, October 2000. not available Presentation Harvey CF. The arsenic crises in Bangladesh. Presented at the Alliance for Global Sustainability Annual Meeting, Lausanne, Switzerland, January 14-17, 2001. not available Presentation Swartz C, Keon N, Badruzzman B, Ali A, Brabander D, Yu W, Islam S, Rahman M, Ahmed F, Rahman H, Hug S, Mustafa M, Polz M, Kuai L, Hemond H, Harvey C. The arsenic crisis in Bangladesh: a geochemical analysis. Presented at the American Geophysical Union Fall Annual Meeting, December 2000. not available Presentation Yu W, Harvey CF. To remediate or to contain? The effects of rate-limited mass transfer and degradation on remedial decisions. Presented to the European Geophysical Union, April 1999. not available Presentation Yu W, Harvey CF. To remediate or to contain? The effects of rate-limited mass transfer and degradation on remedial decisions. Presented to the European Geophysical Union, March 1999. not available Presentation Yu W, Harvey C. To remediate or to contain? The effects of rate-limited mass transfer and degradation on remedial decisions. Presented at the American Geophysical Union Spring Annual Meeting, May 1999. not available Presentation Yu W, Harvey C. To remediate or to contain? The effects of rate-limited mass transfer and degradation on remedial decisions. Presented to the European Geophysical Union, October 1999. not available Presentation Zinn B, Harvey C. When is solute spreading described by dispersion, and when is it described by mass transfer? A comparison of upscaled solute transport behavior in multigaussian and well-connected conductivity fields. Presented at the American Geophysical Union Fall Annual Meeting, December 2000. not available
Supplemental Keywords: groundwater, hydrology, decision analysis, health effects.
- Project Status Reports
Objective(s) of the Research Project:
The objective of this research project is to provide a framework for the evaluation of alternative strategies to manage contaminated groundwater. The framework will include models of physical processes and social values so that a policy analyst can relate the information on a contaminated site to a policy choice on remediation for the site. Issues of social values will be modeled such that the policy analyst can perform what-if analyses on the implications of different social values for the evaluation of the alternative strategies.
Two papers have been completed that describe methods for evaluating environmental decisions. The first paper considers the discount factor that should be used for future outcomes of environmental decision. Cost-benefit studies often analyze proposed projects by the "opportunity-cost criterion," which states that society should not spend resources on a proposed project if the resources could be spent on an investment project that would produce a greater benefit. The opportunity-cost criterion implies that the discount factor used to evaluate the proposed project should equal the ratio of the cost of the resources to the benefit of the investment project. In this paper, we argue in detail that the opportunity cost criterion must be severely modified to evaluate a proposed policy to mitigate future environmental harms that extend over generations and are highly uncertain. The reason is that the hypothetical investment policy provides returns that are highly uncertain and thus add to the uncertainty of the future harms, whereas the mitigation policy reduces the uncertainty of the future harms. If society is risk averse, the policies' opposite effects on uncertainty provide a strong influence in favor of the mitigation policy over the investment policy. The influence is further strengthened by uncertainty (e.g., disagreement among experts) as to the base welfare of future generations.
The second paper describes a method for aggregating individual health outcomes. A decision between policies that affect people's health depends on their preferences regarding their own health and on social values regarding overall health. To evaluate a health policy, its effects on individuals' health must be aggregated into a single measure of its overall effect on health. In this paper, we construct a foundation for using people's preference intensities for health outcomes to aggregate measures of their health outcomes into an evaluation of the overall outcome. The foundation depends on a model by C.M. Harvey (Aggregation of individuals' preference intensities into social preference intensity, Social Choice and Welfare, 1999) that is analogous to the model developed by J. Harsanyi (Cardinal welfare, individualistic ethics, and interpersonal comparisons of utility, Journal of Political Economy, 1955) that uses people's preferences among probability distributions to aggregate personal expected utilities into social expected utility. With the foundation based on people's preference intensities, a policy analyst can separate the two value issues that are directly involved—preferences among personal health outcomes and social values for overall health outcomes—from other issues such as beliefs about uncertain health outcomes.
Much of our work during the last year has been directed towards characterizing the site for our case study in Bangladesh. Two graduate students (one Harvard, and one MIT) and one post-doc are in Bangladesh to collect data during the months of March and April. Through a variety of sources (The Bangladesh University of Engineering and Technology (BUET), the British Geological Survey, UNICEF, the Department of Public Health and Engineering, and the Center for Diarrheal Disease) we have put together a database containing arsenic concentrations in the environment and drinking water, the health effects of this arsenic, and the health effects (infections disease) of drinking surface water. This data will be used to construct the framework for evaluating options for dealing with arsenic contaminated groundwater and providing drinking water.
We received funding from the Alliance for Global Sustainability (AGS) to support the field component of our research. This funding enables us to conduct detailed field measurements of the chemical characteristics that control arsenic mobilization and transport. This project also ensures collaboration with health scientists and water treatment experts. To construct a comprehensive decision framework, we must quantify the effects on health outcomes of different strategies, and quantify the efficacy of different treatment options. Through the AGS, our project includes public health researchers from the University of Tokyo and water treatment researchers from ETH in Switzerland.
Harvey CM, Hammitt J. Equity, efficiency, uncertainty, and the mitigation of global climate change. To appear in Risk Analysis: An International Journal.
Harvey, CF, Gorelick SM. Rate-limited mass transfer or macrodisperion: which dominates plume evolution at the macrodispersion experiment (MADE) site? Water Resources Research March 2000.
Harvey CM. Aggregation of individuals' preference intensities into social preference intensity. Social Choice Welfare 1999.
Hollenbeck, Harvey, Haggerty, Werth. A method for estimating distributions of mass transfer rate coefficients with application to purging and batch experiments. Journal of Contaminant Hydrology 1999.
Yu W, Harvey CF. To remediate or to contain? the effects of rate-limited mass transfer and degradation on remedial decisions. Presented to the European Geophysical Union, April 1999.
Huber CF, Harvey, Rogers. Strategies for sustainable water management in the Gaza Strip. Presented to the World Water Foundation, April 2000.
During the next 12 months, we will work to construct a framework for evaluating options to manage arsenic contaminated groundwater and provide drinking water in Bangladesh, our case study. We also will continue to develop general methods for choosing among options for managing groundwater, and will continue to focus on the problem of evaluating the tradeoffs of present costs with safe water supply and environmental quality in the future.
Supplemental Keywords: groundwater hydrology, decision analysis, health effects.