The National Wildlife Refuge System protects extensive salt marsh acreage in the northeastern United States. Much of this habitat has been degraded by a succession of human activities since the time of European settlement (Gedan and others, 2009), and accelerated rates of sea-level rise exacerbate these effects (Gedan and others, 2011; Kirwan and Megonigal, 2013). Therefore, strategies to restore and enhance the ecological integrity of national wildlife refuge (NWR) salt marshes are regularly considered. Management may include such activities as reestablishing natural hydrology, augmenting or excavating sediments to restore marsh elevation, controlling invasive species, planting native vegetation, minimizing shoreline erosion, and remediating contaminant problems. Uncertainty stemming from incomplete knowledge of system status and imperfect understanding of ecosystem dynamics commonly hinders management predictions and consequent selection of the most effective management options. Consequently, tools for identifying appropriate assessment variables and evaluating tradeoffs among management objectives are valuable to inform marsh management decisions.

Structured decision making is a systematic approach to improving the quality of complex decisions that integrates assessment metrics into the decision process (Gregory and Keeney, 2002). This approach involves identifying measurable management objectives and potential management actions, predicting management outcomes, and evaluating tradeoffs to choose a preferred alternative. From 2008 to 2012, the U.S. Geological Survey (USGS) and U.S. Fish and Wildlife Service (FWS) used structured decision making to develop a framework for optimizing management decisions for NWR salt marshes in the FWS Northeast Region (that is, salt marshes in the coastal region from Maine through Virginia). The structured decision-making steps were applied through successive “rapid prototyping” workshops, an iterative process in which relatively short periods of time are invested to continually improve the decision structure (Blomquist and others, 2010; Garrard and others, 2017). The decision framework includes regional management objectives addressing critical components of salt marsh ecosystems, and associated performance metrics for determining whether objectives are achieved (Neckles and others, 2015). The regional objectives structure served as the foundation for a consistent protocol for monitoring salt marsh integrity at these northeastern coastal refuges, in which the monitoring variables are linked explicitly to management goals (Neckles and others, 2013). From 2012 to 2016, this protocol was used to conduct a baseline assessment of salt marsh integrity at all 17 refuges or refuge complexes in the FWS Northeast Region with salt marsh habitat (fig. 1).

The Maine Coastal Islands National Wildlife Refuge Complex consists of five refuges along the coast of Maine supporting diverse habitats, including coastal islands, forested headlands, estuaries, and wetlands. One of the complex’s refuges, the Petit Manan National Wildlife Refuge, includes 73 islands and four mainland divisions. This report focuses on two divisions of the Petit Manan National Wildlife Refuge; the Gouldsboro Bay Division in Gouldsboro, Maine, and the Sawyers Marsh Division in Milbridge, Maine. These areas each protect salt marsh habitat used by birds identified as conservation priorities, including Anas rubripes (American black ducks) and Ammodramus nelsoni (Nelson’s sparrows), in the New England and mid-Atlantic coast bird conservation region of the U.S. North American Bird Conservation Initiative (Steinkamp, 2008; U.S. North American Bird Conservation Initiative, 2020). This region of the Maine coast is characterized by extreme coastal relief, limited availability of glacial sediments, a large tidal range (3.1 meters), and a rapid rate of local sea-level rise (Kelley and others, 1988). Salt marshes in this region are generally restricted to the borders of protected coves (Jacobson and others, 1987). Management goals set by the FWS for the refuge complex include sustaining the natural functions of the salt marsh ecosystem and maintaining and enhancing salt marsh habitat for migratory birds (FWS, 2005). In this study, the regional structured decision-making framework was used to help prioritize salt marsh management options for two mainland parcels of the refuge complex.

A regional framework for assessing and managing salt marsh integrity at northeastern NWRs was developed through collaborative efforts of FWS regional and refuge managers and biologists, salt marsh research scientists, and structured decision-making experts. This process followed the discrete steps outlined by Hammond and others (1999) and Gregory and Keeney (2002):

This sequence of steps was applied through successive workshops to refine the decision structure and incorporate newly available information. Initial development of the structured decision-making framework occurred during a week-long workshop in 2008 to define the decision problem, specify management objectives, and explore potential strategies available to restore and enhance salt marsh integrity. During 2008 and 2009, workshop results were used to guide field tests of salt marsh monitoring variables (Neckles and others, 2013). Subsequently, in 2012, data and insights gained from these field tests were used in a two-part workshop to refine management objectives and develop the means for evaluating management outcomes (Neckles and others, 2015).

From the outset, FWS goals included development of an approach for consistent assessment of salt marsh integrity across all northeastern NWRs (fig. 1). Within this regional context, staff at a given refuge must periodically determine the best approaches for managing salt marshes to maximize habitat value while considering financial and other constraints. The salt marsh decision problem was thus defined as applying to individual NWRs over a 5-year planning horizon. The objectives for complex decisions can be organized into a hierarchy to help clarify what is most important to decision makers (Gregory and others, 2012). The hierarchy of objectives for salt marsh management decisions (table 1) was based explicitly on the conservation mission of the National Wildlife Refuge System, which is upheld through FWS management to “ensure that the biological integrity, diversity, and environmental health of the System are maintained for the benefit of present and future generations of Americans,” as mandated in the National Wildlife Refuge System Improvement Act of 1997 (16 U.S.C. §668dd note). Two fundamental objectives, or the overall goals for salt marsh management decisions, were drawn from this policy to maximize (1) biological integrity and diversity, and (2) environmental health, of salt marsh ecosystems. Participants in the prototyping workshops deconstructed these overall goals into lower-level objectives relating to salt marsh structure and function and identified performance metrics to evaluate whether objectives are achieved (table 1). In addition, performance metrics were weighted to reflect the relative importance of each objective (Neckles and others, 2015).

The hierarchy of objectives for salt marsh management (table 1) provides the foundation for identifying possible management actions at individual NWRs and predicting management outcomes. Workshop participants developed preliminary influence diagrams (app. 1), or conceptual models relating management actions to responses by each performance metric (Conroy and Peterson, 2013), to guide this process. To allow metric responses to be aggregated into a single, overall performance score, participants also defined value functions relating salt marsh integrity metric scores to perceived management benefit on a common, unitless “utility” scale (Keeney and Raiffa, 1993). Stakeholder elicitation was used to determine the form of each value function relating the original metric scale to the utility scale, ranging from 0, representing the lowest management benefit, to 1, representing the highest benefit (app. 2). Neckles and others (2015) provided details regarding development of the structured decision-making framework and a case-study application to Prime Hook National Wildlife Refuge in Delaware.

In January 2018, FWS regional biologists, biologists and managers from seven northeastern NWR administrative units and USGS and Yale University research scientists (table 2) participated in a 1.5-day rapid-prototyping workshop to apply the regional structured decision-making framework to the Maine Coastal Islands National Wildlife Refuge Complex and the Monomoy, Moosehorn, Parker River, Rachel Carson, and Stewart B. McKinney National Wildlife Refuges. Participants worked within refuge-specific small groups to focus on management issues at individual refuges. Plenary discussions of common patterns of salt marsh degradation, potential management strategies, and mechanisms of ecosystem response offered additional insights to enhance refuge-specific discussions.

Participants identified a range of possible management actions for achieving objectives within both marsh management units at the Maine Coastal Islands National Wildlife Refuge Complex and estimated the total cost of implementation over a 5-year period; the specific years of implementation were not identified in this prototype. Potential strategies to enhance salt marsh integrity included restoring marsh hydrology or increasing marsh elevation (table 3). Participants predicted the outcomes of each management action 5 years after initial implementation in terms of salt marsh integrity performance metrics. For most metrics, baseline conditions within each unit measured during the 2012–16 salt marsh integrity assessment (FWS, 2016) were used to predict the outcomes of a “no-action” alternative. Baseline conditions were estimated by using expert judgement for three metrics that lacked assessment data (abundance of American black ducks, density of spiders, change in marsh surface elevation relative to sea-level rise). Regional influence diagrams relating management strategies to outcomes aided in predicting consequences of management actions (app. 1). Although the influence diagrams incorporated the potential effects of stochastic processes, including weather, sea-level rise, herbivory, contaminant inputs, and disease, on management outcomes, no attempt was made to quantify these sources of uncertainty during rapid prototyping. Management predictions also inherently included considerable uncertainty surrounding the complex interactions among controlling factors and salt marsh ecosystem components.

Following the workshop, the potential management benefit of each salt marsh integrity performance metric was calculated by converting salt marsh integrity metric scores (table 3, workshop output) to weighted utilities (table 4) using regional value functions (app. 2). Weighted utilities were summed across all salt marsh integrity metrics for each action; this overall utility therefore represented the total management benefit, across all objectives, expected to accrue from a given management action (table 4). Constrained optimization (Conroy and Peterson, 2013) was used to find the management portfolio (the combination of actions, one action per marsh management unit) that maximizes the total management benefit across all units under varying cost scenarios for the entire refuge complex. Constrained optimization using integer linear programming was implemented in the Solver tool in Microsoft Excel (Kirkwood, 1997).

Budget constraints were increased in $1,000 increments up to $10,000; in $2,500 increments up to $20,000; in $15,000 increments up to $50,000; in $25,000 increments up to $100,000; in $50,000 increments up to $300,000; and in $100,000 increments thereafter. The upper limit to potential costs was not determined in advance; rather, it reflected the total estimated costs of the proposed management actions. A cost-benefit plot of the portfolios identified through the optimization analysis was used to identify the efficient frontier for resource allocation (Keeney and Raiffa, 1993), which is the set of portfolios that are not dominated by other portfolios at similar costs (or the set of portfolios with maximum total benefit for a similar cost). The cost-benefit plot also revealed the cost above which further expenditures would yield diminishing returns on investment. To exemplify use of the decision-making framework to understand how a given portfolio could affect specific management objectives, the refuge-scale management benefits for individual performance metrics were compared between one optimal portfolio and those predicted with no management action taken.

Potential management actions identified to improve marsh integrity at the Petit Manan National Wildlife Refuge of the Maine Coastal Islands National Wildlife Refuge Complex included removing dikes to restore tidal flow; creating runnels to enhance drainage of water from the marsh surfaces; trapping or applying sediment to increase marsh elevation; or removing drainage tiles or filling ditches to restore natural hydrology (table 3). For costs ranging from $0 to $380,000, the estimated management benefits for individual actions across all metrics, measured as weighted utilities, ranged from 0.429 (for implementing no action, ditch remediation, or sediment trapping in the Sawyers Marsh management unit) to 0.975 (for removing all dikes, installing sediment traps, installing runnels, and filling pools in the Gouldsboro Bay management unit), out of a maximum possible total management benefit of 1.0 (table 3, table 4). In both marsh management units, the action with both the lowest management benefit and lowest cost was the “no action” alternative (management action A).

Constrained optimization was applied to identify the optimal management portfolios over 5 years for a range of total costs to the refuge complex. As total cost increased from $0 (no action in either unit) to about $139,000, the total management benefit at the refuge scale increased from 0.881 to 1.787 (a 103-percent increase; table 5) out of a possible maximum of 2.0 (the maximum possible management benefit of 1.0 for any management action, summed across the two marsh management units). Graphical analysis showed a fairly consistent increase in management benefit as costs increased to $9,545 (fig. 3, portfolio 5). Portfolio 5 represented the turning point in the cost-benefit analysis; as expenditures increased beyond the cost of portfolio 5, total management benefit continued to increase, but at a lower rate.

The portfolios that yielded the greatest total management benefit per unit cost (table 5, portfolios 2 through 5) consistently included removing dikes to restore tidal flow in Gouldsboro Bay management unit and installing runnels to improve surface-water drainage in Sawyers Marsh management unit. As costs increased from $10,000 to $139,000, portfolios included combinations of multiple actions to improve marsh hydrology and retain sediments in both marsh management units (table 5, portfolios 6 through 12). In contrast, some management actions were not included in any portfolio. For example, none of the management portfolios included thin layer deposition of sediments in Gouldsboro Bay management unit, or ditch remediation or sediment trapping in Sawyers Marsh management unit.

Examination of the refuge-scale metric responses to actions included in portfolio 5, which is the turning point in the cost-benefit plot (fig. 3), revealed how implementation could affect specific management objectives. The actions included were predicted to achieve large gains in the overall management benefits derived from increased numbers of tidal marsh obligate breeding birds and density of spiders (as an indicator of trophic health), reduced duration of flooding, and increased capacity of marsh elevation to keep pace with sea-level rise, and modest gains in the benefits derived from changes in density and species richness of nekton (fig. 4). Ecologically, the combination of actions in portfolio 5 may result in an average 200-percent increase in tidal marsh obligate bird counts (averaged across both marsh management units), a 111-percent increase in nekton density, a 13-percent increase in nekton species richness, a 1,450-percent increase in spider density, and a 35-percent decrease in the duration of surface flooding (derived as the average difference between the predicted metric scores for the actions implemented in portfolio 5 and the “no-action” alternative; table 3). Implementation of actions in this portfolio was also predicted to improve the capacity for marsh elevation to keep pace with sea-level rise in one of the marsh management units. The management benefits predicted for portfolios 2 through 4, at total costs up to $4,650, were derived primarily from expected improvements in surface-water drainage and capacity for marsh elevation to keep pace with sea-level rise, and presumed increases in densities of spiders and numbers of tidal marsh obligate birds (table 3, table 4).

A regional structured decision-making framework for salt marshes in NWRs in the northeastern United States was applied by the USGS, in cooperation with the FWS, to develop a tool for optimizing management decisions at the Petit Manan National Wildlife Refuge of the Maine Coastal Islands National Wildlife Refuge Complex. Use of the existing regional framework and a rapid-prototyping approach permitted NWR biologists and managers, FWS regional authorities, and research scientists to construct a decision model for the refuge complex within the confines of a 1.5-day workshop. This preliminary prototype provides a local framework for decision making while revealing information needs for future iterations. Insights from this process may also be useful to inform future habitat management planning at the refuge complex.

The suite of potential management actions and predicted outcomes included in this prototype (table 3) were based on current understanding of the salt marshes in the Petit Manan National Wildlife Refuge of the Maine Coastal Islands National Wildlife Refuge Complex and hypothesized process-response pathways (app. 1). Tidal flooding is the predominant physical control on the structure and function of salt marsh ecosystems (Pennings and Bertness, 2001), and there is widespread scientific effort to elucidate how salt marshes may respond to accelerating rates of sea-level rise and consequent increased marsh inundation (Kirwan and Megonigal, 2013; Roman, 2017). Installation of runnels, or shallow creeks, to increase marsh drainage has shown promise as a climate change adaptation technique in some Rhode Island salt marshes (Perry and others, in press). In this prototype, installing runnels in Sawyers Marsh management unit was included in optimal portfolios as a low-cost method to improve marsh integrity. Many salt marshes throughout the northeastern United States are degraded by roads, dikes, railroads, or other obstructions to tidal flow, and salt marsh restoration frequently focuses on reestablishing tidal flow (Konisky and others, 2006; Roman and Burdick, 2012). Benefits of dike removal often include increases in abundance of nekton in the upper reaches of the marsh (Burdick and others, 1997; Roman and others, 2002), and delivery of sediment to the marsh surface may also be enhanced (Roman, 2017). In this prototype, increasing tidal exchange in the upper marsh areas through dike removal was predicted to improve overall management benefit for relatively low (Gouldsboro Bay management unit, actions B and C) or moderate (Sawyers Marsh management unit, action B) cost (table 3, table 4).

Multiple, interacting factors influence the long-term success of restoration actions in prolonging marsh integrity and improving marsh resilience, and responses to management actions are complex and site specific (Roman, 2017). Due to the large tidal range and rapid rate of sea-level rise in this region of the Maine coast, large volumes of sediment are required for marsh maintenance and vertical growth (Kelley and others, 1988). Management actions tested in other coastal regions for increasing marsh surface elevation include installing low-cost sediment fences made from recycled Christmas trees to enhance natural sediment trapping (Boumans and others, 1997), and placing a thin layer of sediment on the marsh surface to build elevation artificially (Wigand and others, 2017; Raposa and others, 2020). Both of these approaches were suggested as possible mechanisms to improve salt marsh sustainability at Maine Coastal Islands National Wildlife Refuge Complex, but they were not included in optimal portfolios. This may have been due in part to a lack of information to inform the local predicted consequences. Future iterations of this decision model can incorporate improved understanding of marsh responses to management actions. Also, during construction of the regional decision model, lack of widely available data on rates of vertical marsh growth led to the adoption of a very coarse scale of measurement for change in marsh surface elevation relative to sea-level rise (table 1). In 2012, surface elevation tables (Lynch and others, 2015) were installed in each marsh management unit to obtain high-resolution measurements of change in marsh surface elevation. Incorporating this information into subsequent iterations of this structured decision-making framework would likely improve predictions related to the potential for marsh surface elevation to keep pace with sea-level rise.

The prototype model for the Petit Manan National Wildlife Refuge of the Maine Coastal Islands National Wildlife Refuge Complex provides a useful tool for decision making that can be updated in the future with new data and information. The spatial and temporal variability inherent in parameter estimates were not quantified during rapid prototyping. Previously, preliminary sensitivity analysis revealed little effect of incorporating ecological variation in abundance of marsh-obligate breeding birds on the optimal solutions for Prime Hook National Wildlife Refuge (Neckles and others, 2015). This lends confidence to use of this framework for decision making; however, including probability distributions for each performance metric in the decision model could be a high priority for future prototypes. Future monitoring of salt marsh integrity performance metrics will be useful to refine baseline parameter estimates and to determine the background rate of change in the absence of management actions; feedback from measured responses to management actions around the region will help reduce uncertainties surrounding management predictions. In addition, the constrained optimizations analyzed in this report were based on approximations of management costs. As salt marsh management is undertaken around the region, a detailed list of actual expenses can be compiled, including staff time for project planning as well as materials, equipment, contracts, and staff time for implementation. This will allow future iterations of the decision model to include more accurate cost estimates.

The structured decision-making framework applied here to the Maine Coastal Islands National Wildlife Refuge Complex is based on a hierarchy of regional objectives and regional value functions relating performance metrics to perceived management benefits. It will be important to ensure that subsequent iterations reflect evolving management objectives and desired outcomes. Elements of the decision model could be further adapted, for example through differential weighting of objectives or altered value functions, to reflect specific, local management goals and mandates. Future optimization analyses that use this framework could also incorporate additional constraints on action selection, such as ensuring that particular actions within individual marsh management units are included in optimal management portfolios, to further tailor the model to refuge-specific needs.