The following is the established format for referencing this article:Blachly, B., E. Uchida, and S. G. Roy. 2023. Integrating public preferences with biophysical production possibilities: an application to ecosystem services from dam removal. Ecology and Society 28(1):51.
Effective management of ecosystem services requires understanding the biophysical relationships governing the trade-offs, as well as stakeholder preferences for the trade-offs. However, useful tools to guide the complex decision-making process are often lacking. This study demonstrates an approach that combines biophysical and economic models to identify socially preferred solutions. We demonstrate in the context of dam-removal decisions across thousands of dams in Maine, U.S. The results demonstrate the practical usability of this framework for identifying key trade-offs, areas in which people are in agreement and conflicted, along with solutions that are more preferred by society overall. The results also reveal a 30–47% welfare gain from optimizing across all ecosystem services, compared to a more common, visual approach of optimizing two services at a time. This approach may be useful to identify restoration projects that are likely to garner broad public support, particularly when there are trade-offs between ecosystem services, numerous potential solutions, and communities with diverging preferences.
Ecosystem services are the benefits that humans derive from the environment. Because of the biophysical relationships governing their production, environmental management often implies increasing some services at the expense of others (Bennett et al. 2009). Significant progress has been made developing models to quantify these biophysical trade-offs using the economic concept of a production possibilities frontier (PPF; e.g., White et al. 2012, Ager et al. 2017, Roy et al. 2018). Production possibilities frontiers identify combinations of output that fully utilize productive capacity, e.g., the capacity of an ecosystem to provide services. However, these models do not inform which combinations are the most desirable for society. Methodologies for incorporating stakeholder preferences are needed to transform the models into viable tools for decision making, especially when stakeholder groups have different values (Cord et al. 2017).
One promising conceptual framework, firmly grounded in welfare economics, pairs PPFs with indifference curves to identify combinations of ecosystem services that increase society’s welfare (Cavender-Bares et al. 2015). Indifference curves represent the rate at which individuals are willing to trade-off one good for others (e.g., one might be indifferent between three apples and one orange). In traditional consumer theory, the budget constraint forces individuals to make trade-offs in consumption. Within the production-preference framework, consumption of ecosystem services is constrained by the productive capacity of the ecosystem, represented by the PPF. Theoretically, welfare is maximized at the PPF-indifference curve tangency, where preferences for trade-offs between ecosystem services maximizing utility match the ability of the ecosystem to generate those trade-offs (see Appendix 1). The theoretical framework has been used to analyze the trade-off between provisioning services (livestock and crop production) and regulating services (habitat, water quality, and carbon sequestration) using hypothetical PPFs and indifference curves (King et al. 2015, Stosch et al. 2019). However, few empirical applications exist within the context of ecosystem services. In the sole example we are aware of, Mastrangelo and Laterra (2015) paired an empirical PPF with preferences for an agriculture-biodiversity trade-off inferred from a Likert scale measure.
We demonstrate an empirical approach to the production-preference framework that is consistent with its theoretical underpinning. Specifically, we show how a common nonmarket valuation technique, i.e., choice experiments, can be used to estimate indifference curves for like-minded stakeholder groups, which can then be paired with PPFs to identify areas of conflict and consensus. We do so in the context of watershed-scale decisions on dam removal in the state of Maine, USA. The dichotomy presented by dams provides an ideal application for analyzing trade-offs, and Maine has a history of basin-scale decision making (Opperman et al. 2011). We focus on four ecosystem services identified by focus group participants: hydropower production, lake shoreline, spawning habitat for herring, and spawning habitat for Atlantic salmon (Salmo salar).
Generally, we find that the production-preference framework is helpful for combining biophysical and economic models, and it yields significant advantages for decision making compared to analyzing the PPF and choice experiment results individually. For example, the choice experiment data indicate significant variation across stakeholder groups in willingness to pay (WTP) for each service. Variations in WTP, however, do not necessarily result in large disagreements about desired outcomes. The magnitude of disagreement depends on the shape of the PPF, and some seemingly large disagreements become noncontentious in this application. Additionally, in contrast to WTP analysis, the production-preference framework allows inclusion of preferences for stakeholders who refuse to value ecosystem services monetarily. Finally, we find that simplifying multidimensional problems to two dimensions can lead to welfare losses.
Our study is designed in the context of dam removal decision making. Dams are a ubiquitous part of the modern American landscape, numbering over two million nationwide (Smith et al. 2002). Many, however, no longer serve their original purpose or are nearing the end of their engineered lifespan (Doyle et al. 2003). At the same time, there is growing appreciation for the ecosystem services provided by freely flowing rivers (Auerbach et al. 2014), and their role supporting the resilience of social-ecological systems in the face of climate change (Hammersley et al. 2018). As a result, over 1400 dams have been removed in the U.S., with the vast majority occurring in the past two decades (American Rivers 2020). Dam removals inherently upset the status quo, which can have negative ecological and social consequences. For example, dam removal can drive downstream aquatic community mortality (Stanley and Doyle 2003), release contaminated sediment (Bednarek 2001), or drive eutrophication (Gold et al. 2016). On the social side, the novel landscapes created by dams and the community heritage derived from dam-related industry may induce a strong sense of place among residents (Lejon et al. 2009, Fox et al. 2016). Additionally, dam removal may impact home values (Lewis et al. 2008, Provencher et al. 2008, Bohlen and Lewis 2009). The unavoidable and multidimensional nature of trade-offs draws support and ire from a variety of competing stakeholder groups, which often leads to a contentious decision space (Fox et al. 2016). For this reason, individual dams tend to be selected for removal opportunistically, i.e., where conflict is unlikely, Magilligan et al. 2016), despite well-understood efficiency gains from coordinating removals across larger spatial scales (Neeson et al. 2015, Roy et al. 2018, 2020, Kraft et al. 2019). Decision support tools are needed, therefore, to help navigate the trade-offs and mitigate conflicts arising from competing stakeholder interests.
For this study, we focus on dam removal decisions in the Penobscot Watershed in Maine, USA (Fig. 1). The watershed drains over 22,000 km², roughly one-third of the state, and is home to 112 dams spread across 11,200 river kilometers. The large number of dams translates to many possible dam removal scenarios, which ensures the PPF will identify meaningful trade-offs. Additionally, the Penobscot is an attractive candidate for watershed-scale policy coordination. Despite its size, it is located entirely within a single state jurisdiction, and there is a history of coordinating removal decisions across multiple dam sites.
The Penobscot River Restoration Project (PRRP), completed prior to data collection for our study, involved the removal of two main-stem dams coupled with improving fish passage and increasing hydropower generation at others (Opperman et al. 2011). The project featured a broad coalition of stakeholders with diverse interests: hydropower companies, the Penobscot Nation, state and federal agencies, and several non-governmental organizations. It was heralded as a technical and social success. Although our study is forward-looking, the PRRP may validate our approach. Specifically, viewing the project retrospectively within the empirical production-preference framework we develop may provide insight into how and why the project was successful.
We held focus groups to identify services on which to concentrate. Two rely on the presence of dams: hydropower and dammed lakes, and two benefit from dam removal: river herring and Atlantic salmon populations. Hydropower accounts for nearly one-third of net electricity production in Maine (EIA 2019). Within the Penobscot Watershed, 17 dams are licensed to generate hydropower for a total installed capacity of 167,469 kW, enough to supply about 155,000 homes with relatively clean energy. Dams also support and stabilize lakes, which have recreational and aesthetic value. There are approximately 4860.219 km of lake shoreline in the Penobscot Watershed, of which roughly 804.7 km depend on the presence of dams.
Focus group participants identified two fish species as particularly important: river herring (Alosa pseudoharengus) and salmon. Both are anadromous, meaning they spawn in rivers, migrate to the ocean as juveniles, and return to freshwater to spawn subsequent generations. Because dams physically block up and downstream migration, they are a known cause of declining fisheries (Marmulla 2001), and they have contributed to the listing of Atlantic salmon as an endemically endangered species (Limburg and Waldman 2009). Atlantic salmon hold intrinsic value, whereas river herring hold joint importance to the ecosystem as a keystone species and to the economy as a bait source for the state’s lobster fishery. The current configuration of dams provides accessible habitat to support about 5000 Atlantic salmon and 260,000 river herring (23% and 2% of the respective historic populations).
First, we model production functions for each ecosystem service to create two-dimensional PPFs analyzing the biophysical relationship between each dam-dependent service (hydropower and lakes) and each service that benefits from dam removal (herring and salmon). Each PPF identifies a set of Pareto efficient outcomes (i.e., combinations where production of one service cannot be increased without decreasing production of the other). We then administer a choice experiment survey to understand public preferences for the same four ecosystem services, using the PPF results in the experimental design to identify feasible levels of each service. The choice experiment data are analyzed using a latent class model to identify groups holding similar preferences. We estimate each group’s WTP for each service, as well as the rate at which they are willing to trade one service for another (i.e., the marginal rates of substitution). We use these results to construct indifference curves, which are combined with the PPF to identify which of the Pareto efficient outcomes each group finds most desirable.
In addition, we investigate the consequences of simplifying multidimensional trade-offs to two-way analysis. Although the general theory extends to higher dimensions, it is common in empirical applications to bundle correlated services into two groups or to consider a single, representative two-way trade-off (e.g., Nalle et al. 2004, Polasky et al. 2008). This facilitates visualization and comprehension, which can be desirable for decision-making applications (King et al. 2015). In a growing literature focused on empirical production relationships between ecosystem services (Lee and Lautenbach 2016, Cord et al. 2017, Obiang Ndong et al. 2020 provide reviews), much attention is given to identifying spatial correlations between services so that managers can easily comprehend the trade-offs and synergies of managing for a particular service. However, services are unlikely to be perfectly correlated in production, and there is no guarantee that preferences are correlated along the same lines. Two-way analyses, therefore, could lead to non-utility maximizing solutions, especially when supply and demand are considered jointly. We investigate by quantifying the difference in welfare between two- and four-dimensional solutions.
Production possibilities frontier
To generate PPFs, we used the New England Dams Database (https://ddc-nedams.sr.unh.edu/) combined with hydropower capacity licensing data from the Federal Energy Regulatory Commission (FERC 2019), hydrologic data (USGS 2017), and fish passage and habitat suitability models (Roy et al. 2018). Of the 112 dams in the Penobscot, 17 are licensed to generate hydropower, 58 have a lake or pond reservoir, and 62 block historic upstream habitat for salmon and herring. Although hydropower capacity is simply the sum of capacity across all generating dams, measuring carrying capacity for diadromous fish requires spatially explicit ecological models (see Appendix 2). Following Roy et al. (2018), we used a multiobjective genetic algorithm to identify combinations of dam removal that can produce each pair of ecosystem services efficiently.
To fully understand the interactions requires optimizing across all four ecosystem services jointly. The four-dimensional frontier was defined by over 15,000 unique combinations of dam removals. As in two dimensions, each combination represents a Pareto-efficient outcome, i.e., a situation in which an increase to one service comes only at the expense of at least one other. Although technically feasible, a four-dimensional plot would not be useful for understanding trade-offs or informing decisions. To narrow in on which of the 15,000+ combinations are more preferred by the society, we turn to our empirical measure of public preferences.
Choice experiment survey
The survey consisted of four parts. First, respondents were asked about their perceptions and usage of nearby rivers and their prior knowledge of dam removal. Next was an informational section that described the current situation in terms of each ecosystem service, and how the levels could be altered by different dam removal plans. To maintain the applicability of results to a basin-scale optimization, the survey did not identify any individual dams. The information section was followed by the choice experiment. We used feasible ranges for each ecosystem service from the PPF analysis to create an efficient experimental design (Rose and Bliemer 2009). Each respondent faced six choices (see Fig. 2 for an example). The choice questions were followed by demographic and attitudinal questions (Dunlap et al. 2000).
Based on reports of low broadband penetration, we opted for a mail survey using five points of contact (Dillman et al. 2014). Data collection took place between September and November of 2018 using a commercially available mailing list of randomly sampled addresses within the Penobscot Watershed. We received 204 responses with all 6 choice questions completed for a response rate of about 31%. From the choice experiment data, we estimated utility function parameters to obtain WTP for each service, slopes of indifference curves, monetary welfare estimates, as well as their confidence intervals (Krinsky and Robb 1986; see Appendix 3 for modeling details). Baseline results assuming homogeneous preferences are obtained using a multinomial logit (MNL) model. We then explore heterogeneity in preferences using a latent class model (LCM). The LCM identifies groups, or classes, of respondents expressing similar preferences. Each class can then be characterized using responses to demographic and attitudinal questions.
Combining the results to identify preferred outcomes
The choice experiment results contain a lot of information about values and preferences. However, without knowledge of the system’s ability to jointly produce the services, we are left to speculate about a feasible and efficient balance between trade-offs. As the final step, we combine the preference data obtained from the choice experiment with the biophysical data summarized in the PPF.
Theoretically, the utility-maximizing combination of services occurs at the point of tangency between the PPF and indifference curve. In an empirical application, however, equating the slopes is impractical because the PPF is unlikely to be smooth. Further, it becomes visually and computationally intractable in higher dimensions. It is straightforward, though, to obtain the utility provided by each point along the PPF from the parameterized choice experiment model and identify the tangency condition as the point where utility is maximized. For two-way analyses, visual depiction of the tangency condition could reveal useful insights for facilitating discussions and negotiations. We produce these visualizations by plotting a line segment, with slope equal to the marginal rate of substitution that intersects the PPF at the utility maximizing point. To obtain four-dimensional results, we use the same utility maximization approach, but do not produce visualizations.
Production possibilities frontiers in two and four dimensions
The trade-offs between services from dam removals are apparent in the shape of the two-dimensional PPFs (Fig. 3). In these figures, the status quo is represented by the upper leftmost endpoint, where hydropower or lake shoreline are maximized while both fish species are held at their lowest levels. The simulated figures imply that fish populations could be increased without sacrificing hydropower or lake shoreline up to certain levels, but eventually hydropower capacity or lake shorelines would need to be reduced to achieve more fish populations. Some of the PPFs exhibit thresholds, which can arise when dams produce different levels of ecosystem services. These thresholds often represent the removal of large mainstem dams that can increase large habitat areas while not producing a substantial amount of hydropower relative to other dams in the watershed.
Choice experiment results: willingness to pay and latent classes
Results from the choice experiment indicate that respondents value all four of the ecosystem services (Table 1). The coefficients from MNL results suggest that respondents prefer a new dam removal plan to the status quo. On average, they prefer plans that deliver more salmon, hydropower capacity, and herring, while maintaining lake shoreline (Fig. 4).
The LCM results suggest a significant degree of preference heterogeneity between groups (Table 1). Class 1 comprises the largest proportion of respondents (39%) and holds preferences qualitatively similar to those revealed by the MNL model. Members of this class derive positive utility from more hydropower and herring, and negative utility from losing lake shoreline. In contrast, Class 2 (28% of respondents) is the only group that derives positive utility from maintaining the status quo, whereas Class 3 (25% of respondents) is the only group that associates negative utility with increasing the herring population, possibly reflecting a perception that herring crowd out more highly valued species (FERC 1997). Class 4 (9% of the population) derives order-of-magnitude higher utility from increased salmon and lake shoreline compared to other groups. Each group can be characterized by analyzing the descriptive statistics (Table 2) and WTP estimates (Fig. 4). Class 1, for example, is more likely to be male, lower income, concerned for the environment, and regularly recreate on Penobscot rivers. Members of this class have higher WTP for hydropower and lower WTP for lake shoreline than any other class.
Combining the results to identify preferred outcomes
First, we present a simple example, combining the MNL results with an arbitrarily chosen hydropower-salmon two-way PPF (Fig. 5). Maximum utility is achieved by the combination of dam removals delivering habitat to support 15,820 Atlantic salmon and generating hydropower for about 130,000 homes. Uncertainty about the slope of the indifference curve, represented by the 95% confidence interval, indicates that anywhere along that flat portion of the PPF may be the most desirable.
The two-way analyses reveal differences and similarities among classes (Fig. 6; see Appendix 4 for full numerical results). To identify areas of agreement between classes, we look for statistically equivalent indifference curve slopes. For example, the salmon-hydropower results (Fig. 6a, inset table) indicate relative agreement between two classes (Classes 1 and 3). In contrast, Classes 2 and 4 have starkly different preferences for this trade-off. In particular, Class 4 favors a large decrease in hydropower capacity for a marginal increase in salmon population, indicated by the steep indifference curve slope. Similarities between classes, however, are service dependent. Classes 1, 2, and 3, for example, have statistically equivalent preferences for the salmon-lake shoreline trade-off (Fig. 6c, inset table), while no classes are in statistical agreement regarding the herring-hydropower trade-off (Fig. 6b, inset table).
Four dimensional results
The four-dimensional results allow straightforward identification of agreements and disagreements (Table 3). For example, all four classes agree on the importance of maintaining lake shoreline, indicated by the tight range (0-5%) in the utility-maximizing amount of lake shoreline lost. In addition, Classes 1, 2, and 4 are willing to trade off hydropower to benefit herring and salmon populations. The group posing the largest impediment to agreement appears to be Class 3, particularly because of their strong preference for hydropower over both fish species. Further, because Class 3 exhibits negative WTP for herring, it may be possible to move toward consensus by addressing that concern in particular.
We also find that the four-dimensional solution provides additional welfare gains compared to any of the two-way results (Table 4). Maximizing utility across all four dimensions (Column 5) generates a higher welfare gain compared to two dimensions (columns 1-4). The welfare gain estimates from a two-way analyses range from $49.23-$55.83 per person. In comparison, the four-way solution (Column 5) generates a welfare gain of $72.31 per person, which is between 30-47% higher.
This study demonstrated how the production-preference framework can be used to explore preference heterogeneity for ecosystem service outcomes resulting from dam removal. We combined empirical PPFs with results from a choice experiment survey to identify conflict and consensus between stakeholder groups. We find that analyzing a choice experiment and PPF jointly yields insights that are not apparent from individual analysis. For example, latent class analysis of the choice experiment revealed differences in WTP between groups. When represented as indifference curves and combined with the PPF, however, some of the differences become points of agreement. Generally, the magnitude of disagreements depends not only on preferences, but also on the biophysical relationships that determine the shape of the PPF. Also, in contrast to WTP analysis, indifference curves allow the inclusion of preferences from people who refuse to value ecosystem services monetarily.
Production possibilities frontiers can reveal important thresholds, which can help decision makers identify combinations of dams whose management decisions can make large differences to the society. In general, PPFs may exhibit thresholds when a large proportion of each service capacity is concentrated in relatively few dams distributed nonuniformly in a watershed. In our context, a marginal reduction in one service, such as hydropower, can lead to a dramatic increase in another service, such as Atlantic salmon habitat. For the Penobscot River, the dams with the greatest impact on limiting habitat also provide the lowest service capacities for hydropower and lake shoreline length. The resultant thresholds create natural points of emphasis on the frontier. This type of analysis allows decision makers to identify critical dams for the watershed.
Although a visual two-way analysis is easier to comprehend, we find that care should be taken when reducing trade-offs to two dimensions. In PPF models for ecosystem services, correlated services are sometimes grouped or proxied onto a single axis (e.g., Polasky et al. 2008, Mastrangelo and Laterra 2015) to ease visualization, comprehension, and computation. Results indicate that when analyzed individually, the two-dimensional results could be misleading because the socially preferred levels of services can be highly dependent on which services are omitted. For example, the lake shoreline results indicated that herring and salmon can be recovered to about 75% and 80% of their respective maximums with virtually no loss to lake shoreline. However, it is unclear from that analysis how hydropower capacity would be affected with the same changes. Considering instead the trade-off between fish and hydropower, restoring herring or salmon to the same levels required 13% or 32% reduction in hydropower, respectively. Even when two PPFs have a similar shape, indicating a high degree of correlation in production, results can diverge if preferences are not similarly correlated.
Optimizing across all ecosystem services jointly provides results that do not depend on a decision about which two services are represented visually. We find higher levels of welfare associated with solutions derived with all four ecosystem services compared to two. The improvement in welfare derives generally from the fact that the two-way analyses over- (under-) emphasize the importance of included (omitted) services. Though the overlapping confidence intervals suggest the differences are statistically insignificant, welfare improvement is consistent with expectation that jointly optimizing for each service improves the outcome. Moreover, when aggregated to a population (e.g., all residents in the study area), the total welfare gain is likely to be significant. Given the appeal of representing the trade-offs visually with two dimensions, one useful way for decision makers is to derive results using all relevant ecosystem services but use two-dimensional visualizations to convey findings and facilitate discussions.
Generally, we find that the results are consistent with ongoing debates around dams in Maine, suggesting the validity of our approach. Retrospectively, it provides a useful framework to understand the success of the Penobscot River Restoration Project. The project was completed prior to our study and involved removing some hydropower dams while simultaneously improving fish passage and increasing hydropower capacity at others (Opperman et al. 2011). Media coverage of the project focused on two metrics of success: technical success characterized by a win-win between fish habitat and hydropower generation, and social success characterized by consensus-building among stakeholder groups with competing interests (Quiring 2020). Within the production-preference framework, the social success was facilitated by the technical success. Increasing hydropower capacity at existing dams amounted to a shift in the PPF, likely creating a threshold that allowed the preferences of stakeholders on both sides to merge. Empirical applications like this study, therefore, may be especially useful for identifying similar cases in which competing interests join in support of a mutually beneficial outcome.
This study has a few limitations that can be addressed in future research. First, the choice experiment elicits preferences for all four ecosystem services jointly. Although advantageous for capturing preferences across multiple dimensions, it could be a driver of inconsistencies in the two-way analyses. If services were omitted from the choice experiment in a way that matched the two-way PPFs, preferences could differ and yield potentially more consistent results (e.g., if respondents proxy the omitted services when expressing their preferences). Future research could also test the welfare consequences of bundling services in both the PPF and choice experiment into two groups. Additionally, we ignore potentially important spatial factors. More research is needed to develop a model that is spatially explicit in both ecosystem service production and consumption, and to understand whether welfare gains warrant the additional complexity. Finally, it is important to note that every relevant social and ecological trade-off (e.g., spiritual values, equity concerns, sense of place) cannot be meaningfully quantified for inclusion in the model. Nonetheless, we think combining choice experiments with PPFs is a promising framework for policymakers navigating complex ecosystem service trade-offs.
Effective management of ecosystem services requires understanding the biophysical relationships governing these trade-offs, as well as stakeholder preferences for the trade-offs. However, useful tools to guide the complex decision making are often lacking. This study demonstrates an approach that combines biophysical and economic models to identify socially preferred options. Using the context of dam removal decisions across thousands of dams in Maine, U.S., the study combined empirical PPFs with results from a choice experiment survey to identify trade-offs between ecosystem services as well as conflict and consensus between stakeholder groups. The results also reveal a 30-47% welfare gain between optimizing across all ecosystem services compared to a more common, visual approach of optimizing two services at a time. This approach may be useful to identify restoration projects that are likely to garner broad community support, particularly when there are trade-offs between ecosystem services, numerous potential solutions, and communities with diverging preferences.
RESPONSES TO THIS ARTICLE
Responses to this article are invited. If accepted for publication, your response will be hyperlinked to the article. To submit a response, follow this link. To read responses already accepted, follow this link.
This research was supported by the National Science Foundation's Research Infrastructure Improvement NSF #IIA-1539071.
The data and code that support the findings of this study are openly available in Harvard Dataverse at https://doi.org/10.7910/DVN/4STVXI. Ethical approval for this research study was granted by University of Rhode Island IRB (Reference #778925).
Ager, A. A., K. C. Vogler, M. A. Day, and J. D. Bailey. 2017. Economic opportunities and trade-offs in collaborative forest landscape restoration. Ecological Economics 136:226-239. https://doi.org/10.1016/j.ecolecon.2017.01.001
American Rivers. 2020. American Rivers dam removal database. American Rivers, Washington, D.C., USA. https://figshare.com/articles/dataset/American_Rivers_Dam_Removal_Database/5234068
Auerbach, D. A., D. B. Deisenroth, R. R. McShane, K. E. McCluney, and N. LeRoy Poff. 2014. Beyond the concrete: accounting for ecosystem services from free-flowing rivers. Ecosystem Services 10:1-5. https://doi.org/10.1016/j.ecoser.2014.07.005
Bednarek, A. T. 2001. Undamming rivers: a review of the ecological impacts of dam removal. Environmental Management 27(6):803-814. https://doi.org/10.1007/s002670010189
Bennett, E. M., G. D. Peterson, and L. J. Gordon. 2009. Understanding relationships among multiple ecosystem services. Ecology Letters 12(12):1394-1404. https://doi.org/10.1111/j.1461-0248.2009.01387.x
Bohlen, C., and L. Y. Lewis. 2009. Examining the economic impacts of hydropower dams on property values using GIS. Journal of Environmental Management 90:S258-S269. https://doi.org/10.1016/j.jenvman.2008.07.026
Cavender-Bares, J., S. Polasky, E. King, and P. Balvanera. 2015. A sustainability framework for assessing trade-offs in ecosystem services. Ecology and Society 20(1):17. https://doi.org/10.5751/ES-06917-200117
Cord, A. F., B. Bartkowski, M. Beckmann, A. Dittrich, K. Hermans-Neumann, A. Kaim, N. Lienhoop, K. Locher-Krause, J. Priess, C. Schröter-Schlaack, N. Schwarz, R. Seppelt, M. Strauch, T. Václavík, and M. Volk. 2017. Towards systematic analyses of ecosystem service trade-offs and synergies: main concepts, methods and the road ahead. Ecosystem Services 28:264-272. https://doi.org/10.1016/j.ecoser.2017.07.012
Dillman, D. A., J. D. Smyth, and L. M. Christian. 2014. Internet, phone, mail, and mixed-mode surveys: the tailored design method. Fourth edition. John Wiley, Hoboken, New Jersey, USA.
Doyle, M. W., J. M. Harbor, and E. H. Stanley. 2003. Toward policies and decision-making for dam removal. Environmental Management 31(4):0453-0465. https://doi.org/10.1007/s00267-002-2819-z
Dunlap, R. E., K. D. Van Liere, A. G. Mertig, and R. E. Jones. 2000. Measuring endorsement of the new ecological paradigm: a revised NEP scale. Journal of Social Issues 56(3):425-442. https://doi.org/10.1111/0022-4537.00176
Federal Energy Regulatory Commission (FERC). 1997. Final environmental impact statement: license application for ten existing hydroelectric projects in the Kennebec River Basin and license surrender for one existing project. Federal Energy Regulatory Commission, Office of Hydropower Licensing, Washington, D.C., USA.
Federal Energy Regulatory Commission (FERC). 2019. Complete list of active and exempt licenses. Federal Energy Regulatory Commission, Washington, D.C., USA. https://www.ferc.gov/licensing
Fox, C. A., F. J. Magilligan, and C. S. Sneddon. 2016. “You kill the dam, you are killing a part of me”: dam removal and the environmental politics of river restoration. Geoforum 70:93-104. https://doi.org/10.1016/j.geoforum.2016.02.013
Gold, A. J., K. Addy, A. Morrison, and M. Simpson. 2016. Will dam removal increase nitrogen flux to estuaries? Water 8(11):522. https://doi.org/10.3390/w8110522
Hammersley, M. A., C. Scott, and R. Gimblett. 2018. Evolving conceptions of the role of large dams in social-ecological resilience. Ecology and Society 23(1):40. https://doi.org/10.5751/ES-09928-230140
King, E., J. Cavender-Bares, P. Balvanera, T. H. Mwampamba, and S. Polasky. 2015. Trade-offs in ecosystem services and varying stakeholder preferences: evaluating conflicts, obstacles, and opportunities. Ecology and Society 20(3):35. https://doi.org/10.5751/ES-07822-200325
Kraft, M., D. E. Rosenberg, and S. E. Null. 2019. Prioritizing stream barrier removal to maximize connected aquatic habitat and minimize water scarcity. JAWRA Journal of the American Water Resources Association 55(2):382-400. https://doi.org/10.1111/1752-1688.12718
Krinsky, I., and A. L. Robb. 1986. On approximating the statistical properties of elasticities. Review of Economics and Statistics 68(4):715-719. https://doi.org/10.2307/1924536
Lee, H., and S. Lautenbach. 2016. A quantitative review of relationships between ecosystem services. Ecological Indicators 66:340-351. https://doi.org/10.1016/j.ecolind.2016.02.004
Lejon, A., B. Malm Renöfält, and C. Nilsson. 2009. Conflicts associated with dam removal in Sweden. Ecology and Society 14(2):4. https://doi.org/10.5751/ES-02931-140204
Lewis, L. Y., C. Bohlen, and S. Wilson. 2008. Dams, dam removal, and river restoration: a hedonic property value analysis. Contemporary Economic Policy 26(2):175-186. https://doi.org/10.1111/j.1465-7287.2008.00100.x
Limburg, K. E., and J. R. Waldman. 2009. Dramatic declines in North Atlantic diadromous fishes. BioScience 59(11):955-965. https://doi.org/10.1525/bio.2009.59.11.7
Magilligan, F. J., B. E. Graber, K. H. Nislow, J. W. Chipman, C. S. Sneddon, and C. A. Fox. 2016. River restoration by dam removal: enhancing connectivity at watershed scales. Elementa: Science of the Anthropocene 4:000108. https://doi.org/10.12952/journal.elementa.000108
Marmulla, G. 2001. Dams, fish and fisheries: opportunities, challenges and conflict resolution. FAO technical paper no. 419. Food and Agriculture Organization of the United Nations, Rome, Italy. https://www.fao.org/3/y2785e/y2785e01a.htm#TopOfPage
Mastrangelo, M. E., and P. Laterra. 2015. From biophysical to social-ecological trade-offs: integrating biodiversity conservation and agricultural production in the Argentine Dry Chaco. Ecology and Society 20(1):20. https://doi.org/10.5751/ES-07186-200120
Nalle, D. J., C. A. Montgomery, J. L. Arthur, S. Polasky, and N. H. Schumaker. 2004. Modeling joint production of wildlife and timber. Journal of Environmental Economics and Management 48(3):997-1017. https://doi.org/10.1016/j.jeem.2004.01.001
Neeson, T. M., M. C. Ferris, M. W. Diebel, P. J. Doran, J. R. O’Hanley, and P. B. McIntyre. 2015. Enhancing ecosystem restoration efficiency through spatial and temporal coordination. Proceedings of the National Academy of Sciences 112(19):6236-6241. https://doi.org/10.1073/pnas.1423812112
Obiang Ndong, G., O. Therond, and I. Cousin. 2020. Analysis of relationships between ecosystem services: a generic classification and review of the literature. Ecosystem Services 43:101120. https://doi.org/10.1016/j.ecoser.2020.101120
Opperman, J. J., J. Royte, J. Banks, L. R. Day, and C. Apse. 2011. The Penobscot River, Maine, USA: a basin-scale approach to balancing power generation and ecosystem restoration. Ecology and Society 16(3):7. https://doi.org/10.5751/ES-04117-160307
Polasky, S., E. Nelson, J. Camm, B. Csuti, P. Fackler, E. Lonsdorf, C. Montgomery, D. White, J. Arthur, B. Garber-Yonts, R. Haight, J. Kagan, A. Starfield, and C. Tobalske. 2008. Where to put things? Spatial land management to sustain biodiversity and economic returns. Biological Conservation 141(6):1505-1524. https://doi.org/10.1016/j.biocon.2008.03.022
Provencher, B., H. Sarakinos, and T. Meyer. 2008. Does small dam removal affect local property values? An empirical analysis. Contemporary Economic Policy 26(2):187-197. https://doi.org/10.1111/j.1465-7287.2008.00107.x
Quiring, T. 2020. Using engaged rhetorical methods to understand and inform collaborative decision making about dams and restoration in the Penobscot River Watershed. Dissertation. University of Maine, Orono, Maine, USA. https://digitalcommons.library.umaine.edu/cgi/viewcontent.cgi?article=4234&context=etd
Rose, J. M., and M. C. J. Bliemer. 2009. Constructing efficient stated choice experimental designs. Transport Reviews 29(5):587-617. https://doi.org/10.1080/01441640902827623
Roy, S. G., A. Daigneault, J. Zydlewski, A. Truhlar, S. Smith, S. Jain, and D. Hart. 2020. Coordinated river infrastructure decisions improve net social-ecological benefits. Environmental Research Letters 15(10):104054. https://doi.org/10.1088/1748-9326/abad58
Roy, S. G., E. Uchida, S. P. de Souza, B. Blachly, E. Fox, K. Gardner, A. J. Gold, J. Jansujwicz, S. Klein, B. McGreavy, W. Mo, S. M. C. Smith, E. Vogler, K. Wilson, J. Zydlewski, and D. Hart. 2018. A multiscale approach to balance trade-offs among dam infrastructure, river restoration, and cost. Proceedings of the National Academy of Sciences 115(47):12069-12074. https://doi.org/10.1073/pnas.1807437115
Smith, S. V., W. H. Renwick, J. D. Bartley, and R. W. Buddemeier. 2002. Distribution and significance of small, artificial water bodies across the United States landscape. Science of the Total Environment 299(1-3):21-36. https://doi.org/10.1016/S0048-9697(02)00222-X
Stanley, E. H., and M. W. Doyle. 2003. Trading off: the ecological effects of dam removal. Frontiers in Ecology and the Environment 1(1):15-22. https://doi.org/10.1890/1540-9295(2003)001[0015:TOTEEO]2.0.CO;2
Stosch, K. C., R. S. Quilliam, N. Bunnefeld, and D. M. Oliver. 2019. Quantifying stakeholder understanding of an ecosystem service trade-off. Science of the Total Environment 651:2524-2534. https://doi.org/10.1016/j.scitotenv.2018.10.090
U.S. Energy Information Administration (EIA). 2019. Maine - state energy profile overview. U.S. Energy Information Administration, Washington, D.C., USA. https://www.eia.gov/state/?sid=ME.
U.S. Geological Society (USGS). 2017. National hydrography dataset (NHD) Plus. Version 2. USGS, Washington, D.C., USA. https://www.usgs.gov/national-hydrography/national-hydrography-dataset
White, C., B. S. Halpern, and C. V. Kappel. 2012. Ecosystem service tradeoff analysis reveals the value of marine spatial planning for multiple ocean uses. Proceedings of the National Academy of Sciences 109(12):4696-4701. https://doi.org/10.1073/pnas.1114215109
Table 1. Regression results from the multinomial logit (MNL) and latent class model (LCM). Coefficients represent the marginal utility of each attribute. Standard errors are in parentheses and ***, **, * indicate statistical significance at the 99, 95, and 90% level. Multinomial logit assumes homogeneous preferences. Latent class model allows for heterogeneity by identifying classes of respondents with similar preferences. The full sample includes 1205 choices made by 204 respondents.
|MNL||Latent Class Model|
|Class 1||Class 2||Class 3||Class 4|
|% of sample||100||38.7||27.9||24.5||8.8|
Table 2. Descriptive statistics for the four classes of respondents identified by the latent class model (LCM) as holding similar preferences. Reported statistics are proportions except where noted. Standard errors are in parentheses.
|Class 1||Class 2||Class 3||Class 4|
|Income $1000s (mean)||67.9||79.2||83.4||95.6|
|B.A. or higher||0.25||0.33||0.28||0.23|
|15+ yr resident||0.73||0.71||0.80||0.83|
|Own/visit a cabin||0.42||0.45||0.51||0.44|
|Identify as risk averse||0.24||0.37||0.35||0.17|
|Env. concern index (mean)||4.34||2.34||2.89||6.82|
|Recreational river use||0.63||0.50||0.53||0.41|
|High water quality rating||0.27||0.22||0.30||0.18|
|Aware of dam removals||0.74||0.65||0.68||0.76|
Table 3. Four-dimensional utility maximizing combinations by class.
|Class 1||Class 2||Class 3||Class 4|
|1000s of fish||15.6||17.9||5.1||18.1|
|% of max||71%||82%||23%||83%|
|1000s of homes||130.0||104.8||155.1||105.6|
|% of max||84%||68%||100%||68%|
|Millions of fish||12.2||14.7||0.3||12.7|
|% of max||74%||89%||2%||78%|
|% of max||4%||5%||0%||2%|
Table 4. Welfare estimates (average welfare gain per person in USD) for utility-maximizing outcomes derived from the pooled multinomial logit (MNL) model. Two-dimensional results use levels for the two omitted services implied by delivery of the two optimized services. Four-dimensional results optimize all four services jointly.
|1000s of fish||72%||65%||82%||82%||75%|
|% of max|
|1000s of homes||84%||87%||61%||61%||80%|
|% of max|
|millions of fish||73%||75%||66%||86%||78%|
|% of max|
|miles lost||21%||18%||< 1%||4%||4%|
|% of max|