Prioritizing restoration sites that improve connectivity in the Appalachian landscape, USA

Restoring landscape connectivity is a key strategy for sustaining biodiversity and ecosystem services. We developed a decision‐focused process that moves from opportunistic siting of restoration to strategic prioritization by incorporating connectivity enhancement in the Appalachian Mountains, USA. Our approach builds from a recent national‐scale assessment identifying a Resilient and Connected Network (RCN) to inform land protection priorities under climate change. In three high‐ranking study areas within the Appalachians, we demonstrated a circuit‐theory based approach simulating the connectivity value of restoring natural vegetation at sites with high human modification. Our methods were co‐developed by scientists and local decision‐makers. This emphasis on study area‐specific decisions led to differences in how model inputs were defined, for instance using feasible potential restoration sites instead of pixels to define opportunities. Similarly, landscape context influenced our process and in fragmented study areas we added a step that considered additional potential restoration sites outside of the current land protection‐based boundaries. To help interpretation of the results, and link back to the broader network of conservation priorities, we mapped connectivity values in current flow categories that matched the RCN product. Our innovative approach and the decision‐relevant framing, can inform a broad range of connectivity science applications.


| INTRODUCTION
Widespread conversion of natural lands and steep declines in biodiversity suggest an urgent need for actions to protect and improve the condition of ecological systems (Ceballos et al., 2021;Knight et al., 2021;Leadley et al., 2022).Ecological restoration projects, which we define here as reestablishment of natural plant communities, can play important roles in supporting biodiversity and enhancing ecosystem functions and services in degraded landscapes (Aronson et al., 2020;Mappin et al., 2019).At the same time, the costs and complexities of repairing ecosystems constrain widespread implementation (Brancalion et al., 2019;Lindenmayer, 2020), emphasizing the need for tools and approaches that help planners identify cost-effective options that promote desired conservation outcomes (Strassburg et al., 2019).Tools from connectivity modeling can support restoration decisions by identifying sites that help repair critical species movement networks that have been disrupted by land use change (Crooks & Sanjayan, 2006;Tucker et al., 2018;Wilcove et al., 1998).These connectivity values can help planners compare sites based on their role in supporting ecological processes like gene flow and species range shifts that can help biodiversity persist under climate change (Heller & Zavaleta, 2009;McRae et al., 2008).
In typical practice, conservation decision-making has been opportunity based, integrating assessments of biodiversity values and risks with social factors that reflect feasibility of implementation (Knight et al., 2010).Over the past decade, guidance to practitioners has addressed the limitations of this "opportunity mapping" approach, promoting more rigorous methods for setting priorities (Evans et al., 2015;Game et al., 2013;Withey et al., 2012), and demonstrating approaches for considering the conservation impacts of actions on site-specific levels (Tallis et al., 2021).Concurrently, the field of connectivity modeling, as well as practitioners' access to spatial data and computational power, have rapidly expanded (Hall et al., 2021;Hilty et al., 2019;Theobald et al., 2022).While there have been advances in both areas, practitioners prioritizing restoration investments have few tools and guiding processes to assess and incorporate connectivity value.
The United Nations' (2019) declaration of 2021-2030 as the "Decade on Ecosystem Restoration," which in the United States (US) has been followed by "generational" levels of federal funding (Executive Office of the President, 2022), suggests a tremendous opportunity to increase the pace and scope of restoration.Yet, conservation outcomes are only as good as decisions that are made, which are only as good as the methods behind them (Keeney, 2012).In this paper, we emphasize the use of structured decision-making approaches to help conservation planners integrate multiple areas of expertise, values, and information sources as they decide how to invest their time and resources (Gregory et al., 2012).We employed a decision-focused process in which we used a connectivity analysis workflow that evaluated potential benefits from restoration in three study areas in the Appalachian Mountain region of the U.S. In these study areas, local decision-makers are actively working on numerous potential high-investment restoration projects.Enhanced resilience through connectivity is a primary goal, and remediation is challenging.Therefore, prioritization is important to effective conservation outcomes.Our goal is to contribute a practical example of prioritization and support more robust, outcome-oriented restoration action decisions.
Our work was grounded in McRae et al. (2008) wellestablished circuit-theory methods, applications of which include assessments of "structural connectivity," generalized movement flow across landscapes (Dickson et al., 2019;Hall et al., 2021).In these models, "current" (representing moving organisms) flows across a surface representing resistance to movement that is created from spatial data on land cover, land use, and various barriers.Current flow patterns follow a random walk, changing direction and spreading out (diffuse flow) or becoming concentrated in response to the pattern of resistance values (e.g., Koen et al., 2019;McRae et al., 2016;Pelletier et al., 2014).Building upon these methods, which produce current flow values for every pixel in an assessment area, we developed a simulation-based approach for evaluating restoration options at the scale of feasible on-the-ground projects, which we call "sites."To our knowledge, this is the first study that simulated changes in structural connectivity using circuit theory models in a local-scale restoration planning context.

| Conservation context and approach
Our methods for evaluating the connectivity benefit of restoration projects were co-developed by a team of scientists and conservation planners from multiple U.S. business units of The Nature Conservancy (TNC), a global conservation organization.The case study areas were previously recognized as priorities in TNC's Resilient and Connected Network (RCN; Anderson et al., 2023).The RCN integrates multiple continent-scale analyses to identify a land protection blueprint that emphasizes connectivity as an indicator of long-term biodiversity value.We built from the RCN map of important conservation opportunities by crafting a flexible, decision-oriented workflow for evaluating the benefits of site-scale restoration actions that can contribute to the regional connectivity patterns embedded in the RCN (Figure 1).
To characterize benefits to the network from restoration actions, we evaluated current flow at two scales.At the site scale, current flow volume (Figure 1c, blue to brown) is responsive to nearby land use patterns (as translated to movement potential in the resistance surface) and the intactness of natural land cover (Figure 1e).Neighborhoods around sites (defined as areas within a 3 km radius) that are characterized by high current flow contribute to regional-scale flow (Figure 1b), which we classified in two broad categories.Diffuse flow, mapped in blue, occurs in more homogeneous, intact sections of the landscape, and suggests many options for movement.Concentrated flow (orange-brown) occurs where some natural areas have been converted, limiting options for movement, and promoting an accumulation of current in channels and pinch points in the remaining natural lands.Areas mapped in white are dominated by non-natural land cover (high resistance to movement), resulting in very low or blocked current flow.We evaluated current flow at both scales to highlight areas that increase movement potential locally and strengthen the regional network.When restored, a site may have high internal current flow, but be isolated from critical parts of the movement network.Conversely another restoration site may be in a location where the neighborhood is less intact thus having lower current values, but it may fill a key network gap, making a strong contribution to regionalscale flow.
The results of our simulations illustrated how various restoration options improved regional current flow patterns and described local-scale changes in current flow.Notably, we did not create a prioritization for restoration decisions, but instead provided connectivity information (i.e., diffuse or concentrated, amount of influence at local and regional scales) that can be assessed alongside other values and interests held by decision-makers.TNC's multiple conservation projects in the Appalachian region address freshwater resources, benefits of healthy ecosystems to livelihoods, carbon sequestration, and more (https://nature.org/Appalachians).As suggested by recent guidance (e.g., Game et al., 2013;Tallis et al., 2021), prior to being chosen for investment, our restoration connectivity values could be evaluated with other factors like cost and likelihood of restoration success.For connectivityrelated values, our collaboratively developed workflow allows conservation planners to go beyond the static current flow categories in the RCN and evaluate how various restoration opportunities vary in terms of potential to improve connectivity outcomes, and to support related strategies addressing threats across multiple scales.
Given the cost and complexity of restoration, siting decisions often seek to identify one or a few sites at a time, or may arise as the need to say "yes" or "no" to an opportunity to acquire a land parcel from a willing landowner.We simulated possible connectivity changes resulting from restoring individual sites, with site characteristics defined by specific conservation objectives of planners in each location (Miller et al., 2017;Nicolson et al., 2002).As with the RCN, we used a circuit-theory approach for evaluating structural connectivity, which allowed us to consider how restoration could contribute to complex networks that could support a variety of species and ecological processes.We focused on three locations, yet our intent was to develop a flexible workflow that was adaptable to different connectivity contexts (i.e., degree and pattern of land conversion and types of barriers).In our experience, we have observed that the outputs produced by structural connectivity models can be hard to understand and share with decision-makers.We emphasized the relationships between site-focused simulations and regional-scale patterns, and used the same connectivity (current flow) terminology and mapping methods as in the RCN.Through linking these assessments both conceptually and visually, we developed a broader understanding of connectivity model outputs, assumptions, and caveats among TNC staff and conservation partners.

| Study areas
The Appalachian Mountains span over 3000 km, 16 U.S. States, and 3 Canadian Provinces; they are one of the oldest mountain ranges on Earth.This long history, in addition to rugged terrain and varied habitat conditions, has promoted a spectacular diversity of life (Eastern Resource Office, 2001;Ricketts et al., 1999).As the climate changes, elevational gradients and south-north orientation of the links in the Appalachian network provide critical pathways for species range shifts at local and regional scales (Jenkins et al., 2015;Lawler et al., 2013).These attributes, and the relative intactness of habitats within much of the Appalachians led to their identification as a key continental scale corridor in the Connectivity and Climate Flow model that underlies the RCN (Anderson et al., 2023).Natural land extents are high in the Appalachians relative to much of eastern North America, but there is strong heterogeneity, with fertile valleys experiencing widespread conversion to agriculture, towns, and cities.This land use pattern constrains movement options, directing current flow, and presumed wildlife movement, along the many mountain ridge corridors (Figure 2).On some ridges, coal mining has modified topography, and created persistent patches of degraded lands and roads (Maigret et al., 2019;Ruggles et al., 2021;Wickham et al., 2013).Sustaining pathways for range shifts and other wildlife movements across this vast region will require effective land protection, management, and restoration in a variety of landscape and social contexts.Here, we focused on three study areas that vary in degree and pattern of fragmentation (Figure 2), providing an opportunity to demonstrate framing modeling decisions with study area-specific conservation decisions (Table 1), while also evaluating the strengths and limitations of our novel method.
The Nature Conservancy's goals for the Catskills to Adirondacks Linkage focus on connecting core forest blocks separated by an expansive agricultural valley in the state of New York (NY), and rebuilding connections between the central and northern Appalachians (Figure 2 and Table 1b).The second study area, the Allegheny Front, is a 265 km escarpment that extends from Pennsylvania (PA) through Maryland (MD) and West Virginia (WV), where our regional model suggested a movement network persists, but is constrained by agriculture to a series of forested ridgelines, many of which have a history of mining (Figure 2 and Table 1b).For these two study areas, many opportunities for restoration are on privately-owned lands (Table 1c).The southern-most study area, the Cumberland Forest occupies adjacent sections of Kentucky (KY), Virginia (VA), and Tennessee (TN) and is a largely forested landscape under conservation ownership (Figure 2 and Table 1b).This study area is fragmented by former mining sites and related road development, but retains high movement potential, suggesting opportunities to re-purpose some sites for renewable energy (Table 1c).Additional study area details are provided in Supplement 1.2.

| METHODS
Our methods were co-developed with connectivity scientists, decision scientists, and the conservation planners who implement conservation actions in each study area.Following a series of conversations about the context for How do we compare the potential connectivity value of sites that could be restored from human-modified to "natural" condition?
b. What is the regional context for the project area?
North-south corridor spanning 100 km between extensive forests in the Catskill and Adirondack Mountain regions of NY.Characterized by sharp gradients in forest cover from the ends of the linkage to the center in the Mohawk River valley, an agricultural and developed area with many barriers and pinch points.
North-south corridor that extends 265 km along an escarpment that connects intact forests in WV and PA.The study area boundaries had been defined previously in the context of land protection (rather than restoration) decisions, and closely followed the most intact regions, and highest current flow areas mapped in the RCN.We expanded the boundaries to consider a wider range of restoration site options using a least-cost corridor analysis.
The pattern of fragmentation and current flow is broadly similar across the study area.
We focused on 3 watersheds, delineated by 12-digit USGS Hydrologic Units (HUCs).S1).The study areas described here included one or more areas designated as being of high value for biodiversity under climate change.
The "Connectivity and Climate Flow" structural connectivity model of the RCN analysis (referred to here as regional flow, e.g., in Figure 2, or as regional connectivity), for the eastern US. was co-developed with TNC staff and other collaborators (Anderson et al., 2016).Broad consistency in our connectivity modeling and cartographic choices between this work and these earlier works aimed to reduce the learning curve for partners and other decision-makers.As we explored how to model restoration connectivity benefits, we continued the co-development process, allowing definitions of study area boundaries and potential restoration opportunities to vary across the three study areas (Table 1d,e).Connectivity modeling outputs, including the RCN's regional flow map, typically focus on describing how the extent and configuration of various land uses influence movement potential.A key strength of circuit-theory models is that they produce a surface of current flow values that includes the least-cost path, and longer, more random paths between the same focal patches or nodes (McRae et al., 2008; Supplement 2).A connectivity map representing variation and redundancy in movement potential (rather than one path) facilitates the integration of connectivity values with other decision criteria.While studies using circuit-theory tools that characterize movement networks have been increasing rapidly (Hall et al., 2021;Hilty et al., 2019), relatively little effort has focused on locations that impede movement.Focusing on areas of low or blocked current (e.g., the white areas in Figure 2) presents an important opportunity for conservation scientists to compare potential benefits from restoring higher resistance land cover types such as intensive agriculture or former mining sites to natural habitats.
Our methods were informed by McRae (2012) and McRae et al. (2012) Barrier Mapper tool, which identifies areas of high resistance that, if restored to a low resistance cover type, would create the shortest least-cost distances between focal patches (e.g., Dutta et al., 2018;Jones, 2015;Lechner et al., 2015;Torrubia et al., 2014).Identifying key constraints to connectivity has also been addressed with patch-focused circuit-theory methods (e.g., pairwise Circuitscape), including work to highlight "pinch points" (e.g., Gantchoff & Belant, 2017) and potential dispersal barriers (Jarchow et al., 2016;Zi ołkowska et al., 2016).Software, data sources, and computational power have improved, and researchers have developed more dynamic assessments that emphasize connectivity barriers (Anantharaman et al., 2020;Hall et al., 2021), for example, considering climate and land use as drivers of changes in species invasions (Di Febbraro et al., 2019) and droughtflood dynamics (Bishop-Taylor et al., 2018).
Recent advances in circuit-theory tools have eliminated the constraint of having to define focal patches to use as sources and grounds for current flow (Supplement 2), allowing modelers to emphasize "structural" impediments to movement, and to map current flow patterns across entire landscapes.Specific approaches include wall-to-wall or omnidirectional Circuitscape (Anderson et al., 2016;Koen et al., 2019;Pelletier et al., 2014), and the Omniscape tool, which deploys Circuitscape in a circular moving window (Landau et al., 2021;McRae et al., 2016).Maps from these tools allow decision-makers to consider movement networks more generally, with fewer assumptions about start and endpoints.While structural connectivity applications often identify areas of constrained current flow (McClure et al., 2017;Pelletier et al., 2014) to our knowledge, this study is the first to systematically compare changes in movement potential for a pre-selected set of sites that are suitable for restoration.
Evaluating areas for restoration of impeded current flow is challenging because of the dynamic and cumulative nature of connectivity (Figure 1d,e).To model restoration actions on a site-by-site basis, we employed the same wallto-wall Circuitscape approach used in the RCN.To identify where restoration could most reduce impedance, we ran the model iteratively, changing the resistance value for hundreds or thousands of individual restoration sites (Table 1e).For each site, we quantified changes in site-scale current values, as well as changes in current flow patterns classified from variation at the neighborhood (3 km radius) scale.The sub-sections below describe the steps in our workflow.Unless otherwise noted, all spatial data manipulation was done in ArcGIS Pro (ESRI, 2023).Modeling steps, including runs of Circuitscape 5.0 (Anantharaman et al., 2020), were scripted for replication.

| Creating resistance grids
A resistance grid is a cost surface for evaluating movement potential and is a key input for many types of connectivity models.A resistance grid for current land use condition was built from the same spatial data inputs as the anthropogenic flow components of the resistance grid in the RCN Connectivity and Climate Flow product (Supplement 2; Anderson et al., 2016Anderson et al., , 2023)).We began by updating this data layer to incorporate a more recent (2019) version of the National Land Cover dataset (NLCD; Dewitz & USGS, 2021).We then worked with the conservation planners from each study area to identify additional local changes to current land use condition that helped align these depictions of the landscape with their local knowledge.We aggregated the resulting 30 m resolution resistance grid by mean to 90 m to reduce computational demand during thousands of model runs, and to better match the resolution of the previous regional analyses (180 m).
We ranked the "cost" of movement through a pixel on a scale from 1 to 20 using the same land use condition-based methods as in the RCN (Anderson et al., 2016).A value of "1" indicated natural habitat with high movement potential, and "20" was high density development associated with many movement barriers.The full cost-weighting scheme, along with details on the treatment of water, are in Supplement 2.2.

| Evaluating study area boundaries
Defining boundaries for the set of restorable sites at each study area was a key starting point.The conservation planners for the Catskills to Adirondacks Linkage and the Allegheny Front envisioned these study areas as movement corridors that link forest blocks with high land protection value (Table 1b).Initial boundaries for these study areas followed the RCN's high current flow areas.After group discussions, these were determined to be overly restrictive, in that restoration outside of these high flow areas might create new movement options.We used least-cost corridors modeled in the Linkage Pathway tool in Linkage Mapper with the resistance grid described in 2.1 to explore how to expand the boundary (McRae & Kavanagh, 2017; Supplement 2.2).This method is based on a least-cost path model that produces a corridor (rather than just a least-cost line) between focal patches based on a user-defined threshold in cost-weighted distances from the endpoints of the corridor.
The Cumberland Forest study area and the surrounding area are relatively intact, with many possible movement pathways.The goals here included enhancing movement potential in all directions (Table 1b), rather than enhancing movement potential along a corridor.For this large area, we chose to use boundaries that aligned with other planning efforts, focusing on three HUC 12 watersheds that captured variation in land use patterns (Table 1d,e).

| Defining restorable lands
We defined the subset of land within the study area boundaries that could potentially be chosen as restoration sites.The criteria for defining restorable lands varied based on the decisions planners sought to inform (Table 1c) and data availability.Urban areas, low density development, and roads were not considered "restorable" in any of our models.For the Adirondacks to Catskills Linkage, decisions focused on where to restore hay/pasture areas to forested systems.The hay/pasture from the NLCD were split into sites that the local team considered to be of actionable size (4-81 ha, Table 1e).In the Allegheny Front, we focused on historical mining sites, which were identified from state-based datasets (Aronhalt, 2018;Pennsylvania Department of Environmental Protection, 2022;West Virginia DEP, 1996).Local knowledge indicated these datasets were likely incomplete, so we expanded the set of restoration polygons by including areas defined as hay/pasture in the NLCD as "potential" former mining sites.The Allegheny Front team's interest was to restore larger parcels, so we set the minimum size for known historical mining sites at 10 ha (25 acres) and at 20 ha (50 acres) for hay/pasture, and included any polygon above those thresholds (Table 1e).The Cumberland Forest area also has a long history of mining, and a comprehensive dataset of historical mined lands had already been created (Beaty, 2022).We used this dataset to define 2068 restorable polygons ranging in size from 4 to 52 ha (Table 1e).These restoration polygons were the units of analysis and were also used to parameterize the resistance grid.In the Catskills to Adirondack linkage, the restoration target was pasture/hay areas that were given a pre-restoration resistance of five.The other two study areas' restoration targets were former mined areas that were given a resistance of nine representing the highly modified status of these lands.

| Estimating connectivity values for restoration sites
We modeled the changes in connectivity/current flow with restoration for each restorable site.For each polygon site, we created a new resistance surface where we changed the resistance of that land unit to 1 (natural), and then ran wall-to-wall Circuitscape.We used previously established methods (Anderson et al., 2016(Anderson et al., , 2023)), with some modifications to reflect the smaller areas of assessment (Supplement 2.4).This iterative method ensured that the change in connectivity for each restoration site was considered in isolation, but required thousands of runs to test all of the sites across the three study areas.The resulting continuous grids were then classified into regional current flow categories of diffuse, concentrated, and blocked based on total current values and the standard deviation of values in a local neighborhood of 3 km (Anderson et al., 2023;Cameron et al., 2022).
For each potential restoration site, we calculated the total amount of current flow, change of flow relative to a baseline model where that location retained its original resistance values, and contribution to regional networks.To evaluate regional scale contributions, we calculated the classified flow categories for the whole surface F I G U R E 3 Expansion of the study area boundaries.In the Catskills to Adirondacks and the Allegheny Front linkages we used least-cost corridors to inform expansion of the study area boundaries.These two study areas were seeking to restore connectivity between conservation areas (dark green areas in all maps).The middle column of maps show the original planning boundaries (maroon), which closely follow land protection priorities defined in the Resilient and Connected Network (Anderson et al., 2023).The right maps show the expanded boundary where the color ramp and edges reflect gradients and a threshold in movement potential, providing a wider range of opportunities for considering restoration values.
(diffuse, concentrated, or blocked) and determined the category applied to the restored site based on the neighborhood analysis described above.

| Evaluating study area boundaries
The least-cost corridors for the Catskills to Adirondacks and Allegheny Front Linkages provided a useful, repeatable way to create study area boundaries that reflected the underlying resistance surface (Figure 3).They served the intended purpose of helping the teams consider areas on the landscape that, if restored, might strengthen the existing current flow patterns, and provided guidance on areas that could represent additional movement routes.

| Estimating connectivity values for restoration sites
We evaluated changes in connectivity values associated with 8209 potential restoration sites totaling over 200,00 ha.For each restoration unit, we evaluated local and regional impacts.For each restoration site, we calculated the absolute amount of current and the change in current flow as the site was restored from current land use to natural cover (Figures 4b and 5a).For regionalscale flow, we used a neighborhood analysis to determine if restoration of the site would contribute to regional flow as part of a concentrated flow corridor or a diffuse flow area (Figures 4c and 5a).
With restoration, current flow improved by an average of 3.5 (the regional mean current flow is 12, ranging from 2.08 to 39.72, Supplement 2).The Allegheny Front had the highest flow amount in the restoration sites and the greatest change in flow (11.07, 4.11).These results indicate that restoration sites would add to the high flow that characterized the study area.The Cumberland Forest had the second highest average current flow of the study areas, but the smallest average change in flow (10.75, 3.06).While restoration improved flow, the current conditions of this area were already high.
Regional-scale flow uses a neighborhood analysis to look beyond the site scale to quantify how the current flow contributes to the network.Restoration sites may have high flow on the site scale, but be isolated from flow corridors and therefore not contribute to the regional connectivity network.At the regional scale, if restored, 29% of the restoration sites would contribute to the network as concentrated flow (n = 960) or diffuse flow (n = 1435).The Cumberland Forest study area had the greatest percentage of restoration sites contributing to regional flow (53%, n = 1106), with most restoration areas adding to diffuse flow (n = 700).The Allegheny Front had a smaller percentage contributing to regional flow (25%, n = 556), but a much higher proportion (70%, n = 389) contributing to the concentrated flow network.The Catskills to Adirondacks linkage had the smallest percentage of restored areas contributing to the network (19%, n = 733).Most of the Catskills to Adirondack restoration sites contributed to the diffuse areas (78%, n = 571) in the southern portion of the study area.Despite a high number of restoration sites in the central portion of the study area, very few restoration sites helped restore connectivity through the large region of blocked flow.In the Catskills to Adirondack and the Allegheny Front 43% of the restoration sites that contributed to regional flow were outside of the original (RCN-derived) study area boundary.

| DISCUSSION
This collaboration integrated connectivity modeling, decision science, and conservation planning to develop a workflow for comparing projected connectivity enhancements of potential restoration actions.The results from the thousands of sites that we evaluated allow decision-makers to move from comparing static structural connectivity maps toward an outcome-oriented approach, asking "which sites, if restored, could add the most value to the network?"These results are being used in combination with data on other key site attributes (other values, costs, and willing partners) by TNC and partners to prioritize on-the-ground restoration in these study areas.The ability to bring in connectivity values to these prioritizations builds from a decade of innovation (e.g., Anderson et al., 2023;Cameron et al., 2022;McRae et al., 2016) on how to model, and then communicate, the complexity of landscape connectivity.Yet strategic planning of restoration actions with a connectivity focus remains an emerging area of research that has not been included in many restoration plans (America's Longleaf Restoration Initiative, 2009;Doherty et al., 2022).We urge restoration practitioners to use methods such as presented in this study to prioritize engagement and outreach toward site-scale restoration actions that contribute to landscape level benefits.The RCN land protection blueprint (Anderson et al., 2016(Anderson et al., , 2023) ) provided a critical foundation for this work, and identified the three study areas as important places for conservation.Conservation strategies in the study areas involve many approaches, from land protection, to improved management, to restoration.Given that the RCN only identified areas currently in natural land cover, the biodiversity safety net that it delineates emphasizes where actions can help maintain, rather than improve, the strength and support of that safety net.Our results provided the information needed to target non-natural or degraded, restorable properties in key flow areas of the landscape to improve conservation value, and potentially pair land protection and restoration in the same project.
In the sections below, we describe key insights from this case study.

| Scales of connectivity
Our findings indicate that connectivity modeling at multiple scales can improve relevant information to decision-makers.In the continental U.S., over 90% of the land surface is within 3 km of non-natural land use (evaluated by M.C. from Dewitz & USGS, 2021), but knowing which of those modified places to address through restoration requires multiple perspectives.Decades of research in landscape ecology have shown that variations at local and regional scales influence the condition of habitat networks (Hilty et al., 2019).Similarly, these factors help decision-makers understand the level of investment it could take to improve connectivity.
We measured flow at the site and regional scale to identify restoration units that have high potential for increases in flow at both scales.Considering the influence of potential actions at both scales emerged as a key area of insight to the team.We began by measuring just the site scale change in current flow, which is relatively easy to quantify, but can be misleading on its own if the goal of the assessment is to strengthen regional connectivity networks.Through multiple iterations of methods testing and discussion, we added the regional flow analysis.Our findings suggest that regional flow results are a key to measuring restoration projects that have regional connectivity impact.
We evaluated connectivity for restoration units individually, with the goal of identifying options within a bounded study area identified by conservation planners for land protection.While our approach provided useful comparisons, it also revealed evidence of scale mismatches.Most notably, in the Catskills to Adirondacks linkage, only a few sites in the middle section of the corridor were identified as contributing to the network if restored.This zone is where the least natural land cover remains, and conservation planners believe that restoration here is essential to supporting the overall connectivity of the Appalachian system.This finding indicates that a conservation strategy of individual "feasible" size restorations is not the appropriate scale for restoring connectivity of this system.In a modeling sense, the model simulation does not yield important insights on potential benefits in this highly converted zone, because there are so few pathways for current to flow into those locations relative to intact areas at the northern and southern ends of the linkage.Future research that simulates restoration of multiple units at F I G U R E 5 Results at the site scale and the regional scale.At the site scale in the first column, the x-axis values are the original current flow and the y-axis values are the current flow values with restoration.The colors represent the regional flow values of the site.The second column of graphs show the regional-scale flow.The x-axis is the mean flow in the neighborhood, the y-axis is the standard deviation in the neighborhood, and the size of the point relates to the mean flow of the restoration unit.Sites that have high flow, but low standard deviation in their neighborhood (diffuse flow areas) are areas with many movement options.Sites that have high flow, but high standard deviation in their neighborhood (concentrated flow areas) are areas with fewer options for movement, and current flow accumulates in channels and pinch points in the remaining natural lands.
the same time, modeling a suite of conservation actions, would allow the evaluation of larger restoration actions that could effectively leverage multiple investments.

| Restoration in a diffuse current flow area
In the Cumberland Forest, coal mining is the primary disturbance.This area, which is dominated by diffuse flow patterns in the RCN, showed less specificity in terms of where investments could improve connectivity.While restoration often led to high increases in current flow in the restored area, contributions to regional flow tended to be low.The highest gains in flow were in the westernmost watershed, which had the highest human modification (Figure 4b).Within each watershed, the gains for connectivity among the different restoration options were similar.Other benefits (e.g., freshwater resources, benefits of healthy ecosystems to people and livelihoods, carbon sequestration) can be used to prioritize investments.In effect, the case study helped clarify situations where decision-makers could simplify decisions by eliminating the connectivity criterion-not because it was unimportant, but because it did not help them distinguish value among the restoration options due to its limited variability.These types of decisionrelevant judgments are rarely justified in real-world applications (Gregory et al., 2012).

| Collaborative decision-making and promoting implementation
Targeting areas where conservation planners were committed to and involved with making decisions on how to invest in partnerships and on-the-ground action was essential.Modeling restoration opportunities in existing priority areas helped us to frame relevant decisions and allocate effectively within the busy schedules of practitioners that engaged in the process (Cargo & Mercer, 2008;Wyborn & Leith, 2018).
There are several key differences between a "co-development" research approach, where scientific questions are formed in collaboration with stakeholders, and a more siloed approach.First and foremost, framing an appropriate decision context is useful.Many types of restoration decisions can be analyzed that cover various values, options, and scales.Equally critical is framing an analysis that directly responds to problems posed by decision-makers where they have the authority to make decisions (Gregory et al., 2012).Patience on the part of decision-makers is key to the co-development process, as information may be requested at times when the research team is showing limited progress, perhaps mired in a technical challenge evaluating multiple approaches to discuss with the full team.The many different approaches, parameter options, and sensitivities associated with connectivity modeling can easily contribute to slow progress.Knowing each team member's perspective and priorities, especially the decision-makers helped us consider where various conservation approaches (protection, management, and restoration) and evaluation methods were most important.While tailored to these focal decisions, our methods and results can be valuable to many decision contexts in addition to the ones presented in the case study.Through this collaborative process, it became clear that our goal was not to complete "a final map," but as suggested by Chaplin-Kramer et al. (2022), to co-develop a process of linking national, regional, and local scales and adding to the toolkit of conservation approaches to see what places can most add connectivity value, while also respecting there will be other criteria identified as valuable by future decision-makers across the landscape.

| Study limitations and next steps
This paper presents a novel method to modeling connectivity with land use restoration enabled by a decade's worth of advances in computer hardware and software (Anantharaman et al., 2020).With these advances, iteratively running the model thousands of times was time intensive.For our purposes this work provided a useful proof of concept and an opportunity to explore new ideas with colleagues.In the future, we could advance the methods through additional scripting and automation of restoration simulations.
We acknowledge the persistent challenges of validating "generalized structural connectivity" with animal movement data, and for determining "how much change is enough?" to merit restoration (Sowa et al., 2016;Tear et al., 2005).However, this study aimed to seek patterns that indicated relative value and overall patterns of variation among the choices that could be made if resources were available for restoration in the study areas.Our intent was to produce connectivity information that could be combined with other decision-relevant factors to develop a portfolio of promising restoration projects (Hermoso et al., 2015) or to directly rank restoration projects (Martin et al., 2016).While incomplete, we see this method for incorporating connectivity values into multi-criteria restoration decisions as an important advance.
In recent reviews, Keeley et al. (2018) and ( 2019) found that primary success factors for connectivity-focused conservation programs included corridor prioritization, identifying co-benefits, and engaging broad coalitions of stakeholders.We acknowledge that there could be additional benefits of restoration that could yield different results such as carbon sequestration, regulation of water supply, pollination of crops, and cultural and recreational opportunities (Mooney et al., 2009;Pecl et al., 2017).Importantly, in much of this region, the boom and bust of the mining industry has promoted economic and public health disparities (Surber, 2021;Surber & Simonton, 2017).Addressing these additional values of land restoration options is critical, but beyond the scope of this study.

| CONCLUSION
Increasing support for ecosystem protection and restoration (Executive Office of the President, 2022; United Nation, 2019) and declarations of the importance of biodiversity (Manfredo et al., 2021) offer opportunities for science-based decision-making.Implementing ecological restoration actions to produce benefits at regional and continental scales requires strong buy-in from decisionmakers and careful planning and coordination among scientists and other conservation professionals.Our findings indicate that restoring connectivity strategically can provide benefits to the system that can extend far beyond just the area restored.The Appalachians are an exemplary study area for exploring collaborative and innovative approaches, as regional conservation organizations are currently making investments in restoration across diverse landscapes.The methods described herein are currently being used to look into potential future scenarios with improved land use, with the goal of measuring how those potential futures change this exceptional landscape.

F
I G U R E 2 Location of the three study areas in the continental-scale Appalachian movement corridor, as mapped in the Resilient and Connected Network's (Anderson et al., 2023) regional connectivity analysis.Blue areas indicate diffuse current flow, an attribute of highly intact landscapes that suggests many options for movement.Orange-brown areas represent concentrated flow, where neighboring natural areas have been converted leading to accumulation of current in the remaining natural lands, and reduction of movement options to channels and pinch points.Areas of very low or blocked current flow, typically associated with highly modified land cover classes (e.g., intensive agriculture, urban areas), are shown in white.
T A B L E 1 Aligning connectivity modeling approaches with decision opportunities at three study areas in the Appalachians.Catskills to Adirondacks NY TNC chapterAllegheny front WV, MD, and PA TNC chapters Cumberland forest TN, KY, and VA TNC chapters a.What is the connectivity question?

F
I G U R E 4 Restoration area results.The analysis identified potential restoration areas where conservation programs can achieve the greatest connectivity improvement, either concentrated or diffuse.Column (a) shows the landuse of the study areas.Column (b) shows the site scale results.Column (c) shows the regional scale scale results.This information is valuable for decision-making when funding and other resources are limited.