Cumulative Effects of Environmental Change on Culturally Significant Ecosystems in the Inuvialuit Settlement Region

The Inuvialuit Settlement Region (ISR) in the western Canadian Arctic is experiencing environmental changes that affect subsistence harvesting practices and are of concern to local communities. In order to assess the impacts of multiple disturbances on culturally important ecosystems in the ISR, we created a cumulative disturbance map that represents relative intensity of terrestrial disturbances across the study region. We then assessed the relative level of environmental disturbance in important harvesting areas and management zones. Subsequently, we modeled nine future disturbance scenarios that included combinations of increased human impacts and more frequent and widespread wildfires. Using the conservation planning software Marxan, we assessed the potential to conserve large, contiguous areas of unaffected harvesting lands across all scenarios. Our results show that important management zones, wildlife harvesting areas, and community planning zones are all affected by environmental disturbances. Marxan optimizations show that existing disturbance levels create thresholds for current conservation potential and indicate that future disturbances will further limit conservation potential. These results suggest that conservation planners in the region must take steps to anticipate more widespread natural and human-caused disturbance in the ISR and work to maintain large contiguous landscapes that can support wildlife harvesting in the face of ongoing environmental disturbance.


INTRODUCTION
Intensifying human impacts on the environment, combined with a changing climate, are dramatically altering ecosystems worldwide (Steffen et al., 2015).Habitat loss and fragmentation due to human development are well-established drivers of biodiversity loss (Noss et al., 1996;Debinski and Holt, 2000), and the interactions between these disturbances and a changing climate are accelerating ecological transformations (Brooke et al., 2008;Garcia et al., 2014).This pattern is particularly relevant in the Arctic, where increases in air and ground temperatures are well above the global average (Serreze et al., 2000;ACIA, 2005;Burn and Kokelj, 2009), and human development is occurring in previously unaffected ecosystems (Johnson et al., 2005;Kiggiak -EBA Consulting Ltd., 2011).On the surface, individual changes may seem small and insignificant, but when these are combined with other disturbances, the cumulative effects of environmental perturbations can significantly alter ecosystem function (Spaling, 1994).Cumulative environmental impacts are often measured over large spatial and temporal scales and refer to the accumulation of current, previous, or near future disturbances that affect valued ecosystem components (Hegmann et al., 1999).
Cumulative landscape change has the potential to affect communities that are linked to their local environment through subsistence harvesting (Berkes and Jolly, 2001;Parlee et al., 2012;Shanley et al., 2013), particularly in Arctic indigenous communities, where a high reliance on local landscapes for food security intensifies the impact of environmental change on human health and community well-being (Corell, 2006;Furgal and Seguin, 2006).An emerging sub-field of cumulative effects research seeks to understand the impacts of environmental change on culturally valued ecosystem components (Ehrlich and Sian, 2004;Mitchell and Parkins, 2011;Parlee et al., 2012;Spyce et al., 2012).However, to date very little research has explored the overlap between cumulative environmental change and landscape-scale patterns of subsistence use (Mitchell and Parkins, 2011).
To address this gap, we explored the cumulative effects of multiple environmental disturbances on culturally important ecosystems in the Inuvialuit Settlement Region (ISR) in the western Canadian Arctic.Ecosystems in the ISR provide critical habitat for a suite of marine and terrestrial species (Yukon Ecoregions Working Group, 2004;Ecosystem Classification Group, 2009, 2012).This region is also the traditional territory of the Inuvialuit, who rely on the land for hunting, trapping, whaling, and fishing (Usher, 2002;Alunik et al., 2003;Furgal and Seguin, 2006).The ISR has also felt the impact of industrial development associated with hydrocarbon exploration and is experiencing environmental transformations associated with climate change (Burn and Kokelj, 2009;Pearce et al., 2011).The impacts of these perturbations have raised concern among residents, many of whom depend on the land for subsistence use, about the ecological and cultural effects of landscape change (Bennett and Lantz, 2014).However, we are not aware of research that quantifies the cumulative impact of environmental change on culturally significant landscapes in the ISR.
To investigate the cumulative effects of environmental change on wildlife harvesting areas in the ISR, as well as the vulnerability of these areas to future disturbance, we quantified the amount of environmental change that has occurred in culturally significant ecosystems across the mainland ISR over the past 50 years.We also assessed future impacts by developing nine scenarios of increased disturbance and used Marxan software (Ball et al., 2009) to explore the impact of increasing environmental disturbance on the amount of contiguous habitat and the spatial configuration of intact wildlife harvesting areas.This research was one component of a larger initiative exploring the cumulative impacts on culturally important landscapes, which also involved interviews with knowledgeable hunters (Tyson, 2015) to examine the effects of a changing Arctic on hunting and trapping in the region.While the research presented here did not take place in direct consultation with the Inuvialuit Government, it has the potential to inform ongoing research, monitoring, and management efforts in the ISR.

Study Area
This study focuses on the southern ISR, which we define as the mainland portion of the region (Fig. 1).Vegetation structure in this region changes with increasing latitude and can be divided into four broad zones: high boreal forest, low subarctic, high subarctic, and low Arctic tundra (Timoney et al., 1992).The northern portion of the ISR is characterized largely by shrub tundra, while subarctic boreal forest extends through the southern portion of the Mackenzie Delta and southeastern ISR (Burn and Kokelj, 2009;Ecosystem Classification Group, 2012).Alpine tundra dominates the Richardson Mountains to the west (Yukon Ecoregions Working Group, 2004).There are four small communities in the study area: Inuvik (Pop.3463), Aklavik (Pop.633), Tuktoyaktuk (Pop.854), and Paulatuk (Pop.313) (Statistics Canada, 2011).Beyond the municipal boundaries of these communities, human impacts to the land stem largely from a history of hydrocarbon exploration in the region (Burn and Kokelj, 2009).To quantify the cumulative impact of natural and human-caused disturbance in the region, we applied a grid of 25 km 2 cells over the entire planning region (Fig. 1).This grid divided the 131 331 km 2 area into 5815 unique planning units (PUs).While most units were 25 km 2 , instances where the grid overlapped the edge of the study area created some planning units of irregular shape and size.This irregularity was accounted for in all calculations.

Current Disturbances
To assess cumulative impacts to terrestrial ecosystems in the region, we obtained spatial data on disturbances from a variety of sources and used them to estimate the proportion of each PU directly affected by the disturbance.Seismic lines were mapped using polyline coverage available for both the Yukon (Yukon Highways and Public Works, 2014) and Northwest Territories (WWF, 2002).We used air photos with a resolution of ~0.5 m from the Tuktoyaktuk Coastlands (NWT Geomatics, 2004) to estimate the width of a typical seismic line.Subsequently, we buffered polylines to create shapefiles that extended 3.5 m on either side of the line features, the average width of seismic lines we measured in air photos.Similarly, to map the Ikhill Pipeline, which extends from Inuvik to a gas field approximately 49 km to the north, we used aerial imagery to determine that the right of way typically extends 7.5 m on either side of the pipeline (NWT Geomatics, 2004).We combined aerial imagery of Inuvik, Tuktoyaktuk, Aklavik, and Paulatuk (NWT Geomatics, 2004) with data on municipal boundaries (Government of Canada, 2010) to estimate the spatial footprint of each settlement and used these data to delimit the footprints of each settlement as the maximum north, east, south, and west extent of community infrastructure.Point locations of drilling mud sumps (locations of buried drilling fluids and other waste from resource exploration) were obtained from the Environmental Studies Research Fund sumps database (INAC, 2005).To estimate the total area of each PU affected by sumps, we multiplied the number of sumps per planning unit by the mean sump area visible in aerial imagery (22.3 ha) (NWT Geomatics, 2004).
The area of each PU affected by natural disturbances was also estimated using GIS data.Historic wildfires were mapped using the Yukon and Northwest Territories historic fire databases (WWF, 2002;Department of Community Services, 2014).We mapped the area affected by a severe storm surge along the Beaufort coast by using data on vegetation change presented in Lantz et al. (2015).The footprint of retrogressive thaw slumps (areas of ground subsidence and erosion due to permafrost thaw) in each PU was estimated using a broad-scale map of slump density in the NWT (Segal et al., 2016).This dataset portrays the density of slumps in 225 km 2 cells as no slumps, low density (1 -5), medium density (6 -14), or high density (15 or more).To use these data to estimate slump coverage in each PU, we assumed that, on average, low-density grid cells contained three slumps, medium-density grid cells contained 10 FIG. 1. Study area map.We defined our study area as the mainland ISR, which covers an area of 131 331 km 2 .This area includes the communities of Inuvik, Aklavik, Tuktoyaktuk, and Paulatuk.We applied a grid of 25 km 2 cells to the region, creating 5815 unique planning units (PUs), which were used to tabulate levels of environmental disturbance.
slumps, and high-density grid cells contained 20 slumps.We then multiplied the number of slumps in each cell by the mean slump size in the region (3.02 ha) (Segal et al., 2016).This produced an estimate of the total area disturbed by slumps in each 225 km 2 grid cell.The percentage of each 225 km 2 grid cell affected by slumps was then attributed to every 25 km 2 PU that occurred within its boundary.In instances where a 25 km 2 PU was split by the boundary of multiple 225 km 2 grid cells, the 225 km 2 grid cell that contained the largest portion of the PU area was used to determine the percentage of PU affected by slumps.

Future Disturbances
To explore the impact that more frequent wildfires and increasing industrial development might have on the footprint of disturbances in the region, we generated spatial data representing scenarios of increased industrial activity and wildfire over the next 50 years.We restricted the modeling of future human impacts to either development that is already in progress or potential development that has publicly available plans.This approach limited our modeling to an all-season road that is currently being built from Inuvik to Tuktoyaktuk (Kiggiak -EBA Consulting Ltd., 2011); the proposed route of the Mackenzie Valley Pipeline, which enters the ISR near Inuvik and runs northwest to the Beaufort Sea (Joint Review Panel, 2010); and an area of existing mineral claims near Paulatuk (WWF, 2002).The future road was mapped at a width of 20 m on the assumption that it will be similar in size to the Dempster Highway (Gill et al., 2014).The pipeline was mapped by applying a right of way with the same width as the Ikhill Pipeline.To simulate the impacts of future mineral exploration in the Paulatuk area, we used the boundaries of existing mineral claims in the region (WWF, 2002).In the absence of data on the level of planned development, we modeled a scenario in which mineral extraction had direct impact on approximately 20% of the area in each PU affected.
To simulate future natural disturbances, we focused on wildfire because it can be modeled in a systematic fashion on the basis of historical fire rates (WWF, 2002; Department of Community Services, 2014) and known vegetation zones (Timoney et al., 1992).Although other types of ecological disturbances are likely to increase across the ISR (Fraser et al., 2014;Lantz et al., 2015), we chose to limit the scope of our modeling to fire because simulation was relatively straightforward and yielded scenarios ranging from low to high disturbance.The spatial extent of future wildfire was estimated by generating disturbances using the Geospatial Modeling Environment (GME) software (Beyer, 2014).The first step in this process involved parameterizing GME to simulate fires with a size and frequency that were consistent with historical wildfires in each of the vegetation zones in the region (WWF, 2002;Department of Community Services, 2014).We first calculated the size and density of historic wildfires in the high boreal, low subarctic, high subarctic, and low Arctic vegetation zones described by Timoney et al. (1992).Using the spatial boundaries for each of these vegetation zones (Timoney et al., 1992) and data on historical fire frequency (WWF, 2002;Department of Community Services, 2014), we adjusted the frequency of ignition, the rate of spread, and the time that a fire was active on the landscape until GME yielded outputs that mimicked the percentage of area disturbed by fire over the past 50 years in each vegetation zone (online Appendix 1: Table S1).
We then created three "future fire" scenarios intended to reflect the impacts on fire frequency of rising air temperatures (Serreze et al., 2000;ACIA, 2005) and increasing fuel accumulation (Lantz et al., 2013;Fraser et al., 2014).Scenarios were based on the assumption that the future size and density of fires in a given zone would be similar to the patterns now common in the vegetation zones immediately to the south (Table 1).The Mackenzie Delta was excluded from fire simulations because the high density of rivers and lakes limits the potential for large or frequent fires (Burn and Kokelj, 2009).To simplify our model, we chose not to increase rates of fire in the boreal forest zone of our study area, which is at the northern limit of the biome and comprises only a small percentage of our study area, because we were primarily concerned with changes to Arctic ecosystems.Consequently, given the forecasted changes in fire frequency in this biome (de Groot et al., 2013), our scenarios in the boreal portion of the study area should be viewed as conservative.
Using these data layers, we constructed nine future disturbance scenarios that involved combinations of fire and anthropogenic disturbance (Table 2).In each future scenario, current disturbances were combined with potential future disturbances to represent a range of possible disturbance levels over the next 50 years.The modeled intensity of each disturbance and its persistence on the landscape are described in the following section.

Weighting
The intensity of a disturbance's environmental impact varies according to the ecological variable being measured, the nature of the disturbance, the ecosystem component(s) TABLE 1. Percent of the different landscape zones affected by fires in three scenarios created to model wildfire frequency over the next 50 years.Simulation 1 is the baseline scenario, in which fire rates in each zone remain constant.Simulations 2 and 3 assume that increasing fuel loads, warming temperatures, and more frequent lightning will yield disturbance regimes similar to those in lower-latitude vegetation zones, with fire rates increasing in a stepwise manner.it affects, and the conditions of the landscape on which it occurs (Duinker et al., 2013).There is no standard method for weighting disturbances on the basis of their intensity and frequency, and cumulative effects research typically weights disturbances differently depending on the ecosystem component in question (Johnson et al., 2005;Gunn et al., 2011;Raynolds et al., 2014).We developed a weighting scheme that accounts for differences in 1) the impact that disturbances have on vegetation structure, soils, and ground temperature (disturbance severity) and 2) the time it takes to recover following disturbance (recovery time).This relative scheme was developed using existing data on the impacts of disturbances on vegetation, soils, and permafrost conditions (Table 3).Disturbances were weighted in relation to each other by multiplying a severity score by a recovery score for each disturbance type (Table 3).Severity scores ranged from 1 (minimal ecological alteration) to 10 (total land transformation).Recovery time was ranked using a scale ranging from 0 to 1 to represent the length of time a disturbance persists on the land.If a disturbance, such as a community development, is likely to persist over a 50-year period, it received a score of 1.If a disturbance, such as seismic lines, is likely to show significant recovery of vegetation structure and ecological processes over a 50-year period, it received a score between 0.1 and 0.9.Lower scores represented a less persistent disturbance that is likely to exhibit significant recovery over a 50-year period.To calculate cumulative disturbance scores for each PU, we multiplied the percentage of area affected by each disturbance by the disturbance weight and summed these scores in each PU.
In simulated future scenarios, all existing disturbances were included, but we recalculated the original disturbance score by multiplying again by the recovery score (Table 3).This approach allowed us to simulate the cumulative impacts of disturbances over time, while also acknowledging the decreasing impact of current disturbances in the future.

ANALYSIS ZONES
We examined the spatial pattern of current landscape disturbance by mapping disturbance scores across the study area (Fig. 1) and comparing disturbance levels between different analysis zones (Table 4).We defined our analysis zones using areas of importance that were outlined in community conservation plans for the region (IJS, 2008a -d).These analysis zones consist of individual planning areas (PAs) that are designated for each community; priority management zones, which are areas of high cultural or ecological significance that are managed to avoid environmental disturbance (IJS, 2008a -d); and caribou harvesting zones.While many other types of harvesting are common in the region, we chose to analyze caribou harvesting zones as an example of species-specific use areas because caribou are harvested in multiple communities (Usher, 2002;Alunik et al., 2003;Joint Secretariat, 2003), are the focal point of research and management efforts (Adamczewski et al., 2009;Environment and Natural Resources, 2011;Gunn et al., 2011), and because conservation plans provided consistent data across the study area.In every analysis zone, we measured the percentage of PUs containing environmental disturbance and the number of unique disturbance types occurring in PUs.We also measured the percentage of PUs containing high levels of environmental disturbance, which we defined as a disturbance score ≥ 80 (the equivalent of half the PU being affected by wildfire).The impact of current and future disturbances on multiple harvesting values was examined in our Marxan analysis.

Marxan Analysis
The spatial prioritization software Marxan (Ball et al., 2009) was used to analyze the impact of each disturbance scenario on the area and contiguity of undisturbed PUs in the study area.Marxan is spatial planning software designed to find a near-optimal solution to a conservation problem by maximizing the acquisition of valued habitat while minimizing the cost associated with protecting these lands.In our analysis, we used the cumulative disturbance score as Subsequent iterations targeted a higher percentage of each harvesting area.Optimizations were run until the threshold for each disturbance scenario was reached, and Marxan failed to achieve the targeted percentage of wildlife harvesting areas.Each of these optimizations comprised 100 runs, in which Marxan attempted to select the targeted percentages for each use value while incurring the lowest possible cost (i.e., minimizing the disturbed terrain selected) and maintaining contiguity.Marxan parameters were set so that optimizations prioritized maintaining a low overall cost and avoided areas of high disturbance intensity (online 1 Since we estimated that active slumps will continue to occupy a similar area, the future weight of thaw slumps was not adjusted.Appendix 1: Table S2).For a full list of Marxan parameters and the 40 terrestrial harvesting areas used in this analysis, see online Appendices 2 and 3. We compared Marxan outputs across all simulations to investigate the impact of changing disturbance levels on the availability of community-defined harvesting areas.In order to measure the connectivity in each output, we compared the average number of PU edges per PU for all Marxan solutions in each disturbance scenario.To measure the level of disturbance in Marxan outputs, we measured the average cost per PU.Patterns were analyzed between all 10 disturbance scenarios and across each conservation target of use values.

Current Disturbance Footprint
Our compilation of GIS data shows that disturbance has impacts on most areas of the southern ISR, but it also reveals substantial spatial variation in the intensity of these impacts (Fig. 2).Over half of PUs contained at least one disturbance type, while 14.8% of PUs contained two or more disturbances (Table 4).Slumping and historic seismic lines were the most widespread disturbance types (Table 5), and disturbance intensity was low to moderate in most of the study region (Fig. 2).Large expanses of undisturbed areas existed in the Richardson Mountains, Amundsen Gulf Lowlands, Anderson River Plain, Bluenose Lake Plain, the eastern Dease Arm Plain, and Coronation Hills (Fig. 2).Well-defined hotspots of high-intensity disturbances were found in the Tuktoyaktuk Coastal Plain, the western Dease Arm Plain, and the Great Bear Lake Plain (Fig. 2), which represented a low percentage of total PUs (Table 4).
The percent of PUs disturbed also varied among community PAs, management zones, and key wildlife harvesting areas.The percentage of disturbed PUs in Aklavik, Inuvik, and Tuktoyaktuk PAs was higher than the disturbance rate for the entire study area, while the Paulatuk PA contained a much lower percentage of disturbed PUs (Table 4).Management zones that have been identified as most significant, particularly significant, or seasonally significant in the community conservation plans each contained disturbance levels similar to those of the study area as a whole, while specific caribou harvesting zones contained a greater percentage of disturbed cells than the whole study area (Table 4).

Future Disturbances and Marxan Simulations
Simulations of future fires and increased human activity created nine distinct future disturbance scenarios (Fig. 3).Comparing the spatial distribution of disturbances in future scenarios to the baseline scenario revealed that the frequency and intensity of environmental disturbances in the region increased throughout all 10 scenarios (Fig. 3).Scenarios 2, 5, and 8 involved major shifts in fire occurrence and displayed the greatest increase in disturbance levels across the study area (Fig. 3).Increasing human disturbance alone (Scenarios 3,4,6,7,9,and 10) resulted in a much smaller increase in regional impacts (Fig. 3).
When disturbance scenarios were used to modify the cost layer in Marxan optimizations, we observed two distinct thresholds where Marxan could not achieve conservation targets.In disturbance scenarios 1 to 7, failure rates were 100% when optimizations attempted to conserve 82% of all use values.In Scenarios 8 to 10, the failure threshold was 76% (Table 6).At targets of 50% and 75%, Marxan also encountered a significant failure rate in disturbance scenarios 8 to 10 (Table 6).
The average measure of solution edges per PU and the average cost per PU increased in scenarios with greater disturbance.Increases in cost and edge ratios were much larger in scenarios that included shifts in fire frequency, compared to scenarios that simulated increased human activity (Figs. 4 and 5).As conservation targets increased, the magnitude of differences between cost scores and the edge to PU ratios among simulations also increased (Figs. 4 and 5).The impact of disturbance on habitat contiguity was also evident in mapped Marxan outputs, where scenarios with a higher level of disturbance produced outputs containing fewer intact solutions and much longer edges (Fig. 6).

Current Disturbance Levels
This study highlights the importance of assessing cumulative effects in regions with a strong reliance on subsistence harvesting.Our analysis shows that the Inuvialuit Settlement Region (ISR) is more affected by natural and human disturbances than is suggested in Community Conservation Plans that call for the avoidance of disturbances across large management areas (IJS, 2008a -d).This finding underscores the potential to overlook cumulative impacts that occur in large regions.Our mapping shows significant overlap between widespread disturbances and wildlife harvesting areas, important management zones, and community planning areas.We found that this overlap is substantial enough to limit the availability of undisturbed harvesting zones, and reduces conservation potential in the region.To our knowledge, this is the first comprehensive assessment of cumulative effects in the ISR, and it suggests that a wide range of perturbations have already affected Inuvialuit land use.
Our findings also highlight the importance of broadscale cumulative effects assessments for conservation planning in the region.Numerous co-management organizations in the ISR have the mandate to protect biodiversity and traditional harvesting (INAC, 1984), and monitoring of cumulative effects has been incorporated into regional management across the Northwest Territories and Canada (Government of Canada, 1998;MVEIRB, 2004;Duinker and Greig, 2006;Ehrlich, 2010).To date, however, few initiatives have effectively translated the general consensus regarding the significance of cumulative effects into effective monitoring and management practices (Duinker and Greig, 2006;Gunn et al., 2011).This gap is alarming because the combination of climate change impacts and industrial development represents a significant threat to northern food security and biodiversity (Corell, 2006;Furgal and Seguin, 2006;Fuller et al., 2008).To avoid crossing thresholds where subsistence harvesting is severely impaired or no longer possible, strategies for monitoring and managing cumulative environmental impacts are needed (Parlee et al., 2012).The method for weighting disturbances presented here provides a means to include both widespread, low-intensity disturbances (e.g., seismic lines), and high-intensity impacts (e.g., salt-water vegetation kill) in a cumulative effects assessment.The maps produced can be used to evaluate conservation potential across a gradient of disturbed to undisturbed landscapes and as inputs for spatial planning activity (Ball et al., 2009;Moilanen et al., 2009Moilanen et al., , 2011)).
Arriving at clear definitions of thresholds remains a challenge in cumulative effects management, and approaches range from restrictions on any development in pristine landscapes (Ehrlich, 2010) to explicit levels of allowable impacts on specific valued ecosystem components (Gunn et al., 2011).Regardless of the methods for defining a threshold, if land-use planning evaluates projects on a caseby-case basis, it risks exceeding acceptable disturbance  levels by ignoring the combined effect of multiple stressors (Duinker and Greig, 2006).The approach used in this analysis makes it possible to include a range of disturbances in the evaluation of multiple stressors across a region.Combined with data on proposed developments, this approach may be particularly useful in identifying when additional impacts will exceed thresholds for acceptable disturbance levels.

Future Disturbance Scenarios
Our analysis based on future disturbance scenarios showed that additional perturbations reduce the potential to conserve undisturbed harvesting areas in the ISR.Scenarios 2 to 6 maintain the same 82% failure threshold that exists under current disturbance levels, while scenarios with large increases in fire frequency (7 to 10) have a higher failure rate at all targets and a lower failure threshold of 76%.This result emphasizes that predicted increases in natural disturbance frequency and intensity (ACIA, 2005;Jandt et al., 2008;de Groot et al., 2013) are likely to limit flexibility in meeting conservation targets and alter the potential to maintain undisturbed harvesting areas in the ISR.Global climate change and increasing pressure from human activity require conservation strategies that anticipate these types of change (Pressey et al., 2007;Trombulak and Baldwin, 2010;Groves et al., 2012;Brodie et al., 2013).Cumulative effects management has proven effective when it considers reasonably foreseeable impacts in addition to those that are most likely to occur (Ehrlich, 2010).Our Marxan simulations of future disturbance scenarios TABLE 6. Percent of Marxan runs in which the solution failed to conserve the targeted percentage for at least one use value for lack of available planning units (PUs) with a low enough disturbance score for inclusion.Two distinct thresholds exist at which Marxan solutions are unable to meet conservation targets for all use areas: the failure threshold is 82% of use values conserved in scenarios 1 to 7, but 76% in scenarios 8 to 10. Scenario 1: current disturbance levels, 2: baseline future fire rates, 3: baseline future fire rates and road and pipeline development, 4: baseline future fire rates and road, pipeline, and mineral development, 5: moderate increase in future fire rates, 6: moderate increase in future fire rates and road and pipeline development, 7: moderate increase in future fire rates and road, pipeline, and mineral development, 8: high future fire rates, 9: high future fire rates and road and pipeline development, 10: high future fire rates and road, pipeline, and mineral development.demonstrate that climate change will reduce the availability of high-quality harvesting areas in the ISR.They also highlight the importance of conservation planning efforts that limit direct human disturbance in order to allow ecosystems to absorb the impacts of climate change (Doak et al., 2013).
Our results indicate that an increase in disturbance will reduce the quality of conservation outcomes by increasing fragmentation.Even if optimizations are able to meet conservation targets, increased disturbance levels in our scenarios limit the potential for large configurations of contiguous harvesting areas.This result is concerning because habitat fragmentation is directly correlated with species decline (Lindenmayer and Fischer, 2006;Collinge, 2009).Even large protected areas have struggled to meet conservation goals as surrounding landscapes become increasingly disturbed (Trombulak and Baldwin, 2010), which raises the possibility that the conservation solutions in our scenarios may not represent adequate protection of harvesting areas if they are highly fragmented or surrounded by significantly disturbed PUs.
Our scenarios coupled modest industrial development in a region with a range of natural disturbance intensities.These scenarios provided a simplified method for exploring the conservation implications of development in a future affected by climate change.These simulations were based on the likelihood that wildfire will become more common with increases in temperature (ACIA, 2005;Jandt et al., 2008;de Groot et al., 2013) and are used as estimations, not predictions.Given the likelihood of additional increases in permafrost degradation (Kokelj et al., 2015) and storm surges (Lantz et al., 2015), our future scenarios are conservative generalizations of potential impacts.
The large range of variation in fire frequency and narrow range in human disturbance used in our simulations do not represent the relative significance of human impacts in the region.Human disturbance modeling was restricted to potential development that is either in progress (Kiggiak -EBA Consulting Ltd., 2011) or has publicly available plans (WWF, 2002), which limited the spatial extent of human disturbances relative to simulated wildfire.Roads, pipelines, and infrastructure all significantly affect Arctic wildlife (Nellemann and Cameron, 1998;Johnson et al., 2005;FIG. 6. Maps showing the "best output" from Marxan runs, using (A) disturbance scenario 1, 50% conserved; (B) scenario 10, 50% conserved; (C) scenario 1, 75% conserved; (D) scenario 10, 75% conserved; (E) scenario 1, 82% conserved; and (F) scenario 10, 90% conserved.The shading on the base maps represents disturbance intensity from low (blue) to high (red).Areas selected are shown in green.As disturbance levels and conservation targets increase, the contiguity of Marxan outputs decreases.Gunn et al., 2011), ecological processes (Myers-Smith et al., 2006;Kokelj et al., 2010;Gill et al., 2014;Raynolds et al., 2014), and land users' ability to hunt and trap (Tyson, 2015).Their impacts should not be discounted because of their relatively small footprint in this analysis.Even development projects with a small spatial footprint may be undesirable in particularly sensitive or culturally important ecosystems (Ehrlich, 2010).To assess the full range of potential human disturbance, future research should simulate human disturbance in the region by projecting a number of scenarios that represent more widespread potential resource extraction (Holroyd and Retzer, 2005).

CONCLUSION
The results of this analysis emphasize three main points: 1) the ISR is already widely affected by environmental disturbance; 2) the potential to conserve large contiguous areas of habitat is limited by existing disturbances; and 3) future environmental disturbances, particularly those associated with climate change, will further reduce the potential to conserve large, contiguous undisturbed harvesting areas.These findings indicate that land-use planning in the ISR needs to account for increasing environmental change in order to achieve conservation objectives in culturally important ecosystems.The mapping and weighting approach described in this paper can be used to quantify the impacts of environmental change on subsistence land use, particularly where local communities are concerned about changes across a large landscape (Ehrlich, 2010).This approach creates tractable representations of the impacts to culturally significant ecosystems and encourages greater consideration of local land users in cumulative effects analysis.APPENDIX 1: TABLE S1.Parameters for fire scenario generation Geospatial Modeling Environment (GME) software (Beyer, 2014).TABLE S2.Parameters used in this Marxan analysis and their treatment across all simulations.APPENDIX 2: Parameters imported from Marxan input file (input.dat).APPENDIX 3: Wildlife harvesting areas included in Marxan analysis (imported from Marxan spec.datfile).

FIG. 3 .
FIG.3.Spatial output of each disturbance scenario.Scenario 1: current disturbance levels, 2: baseline future fire rates, 3: baseline future fire rates and road and pipeline development, 4: baseline future fire rates and road, pipeline, and mineral development, 5: moderate increase in future fire rates, 6: moderate increase in future fire rates and road and pipeline development, 7; moderate increase in future fire rates and road, pipeline, and mineral development, 8: high future fire rates, 9: high future fire rates and road and pipeline development, 10: high future fire rates and road, pipeline, and mineral development.Inset in the bottom left corner shows the study area location in black and the entire ISR boundary in red.

Scenario
FIG. 4. Average edge score per planning unit (PU) for all Marxan analyses.Three sets of simulations were run for each disturbance scenario, attempting to reach conservation targets of 50%, 75%, and 82%.We averaged the Marxan edge score per PU to assess the contiguity of solutions.Symbols show the mean connectivity score and 95% confidence intervals around the mean.Scenarios that attempted to conserve 82% of use values all failed to meet targets for at least one value.Connectivity scores for these outputs represent the mean score of unsuccessful solutions.

TABLE 2 .
Scenarios based on combinations of current and future disturbances.All future disturbance scenarios included current disturbances and the simulated impacts of more widespread fire or anthropogenic disturbance.Disturbance intensity increases in each scenario with the introduction of either increased fire occurrence or increased human activity in the study area.
vation Plans(IJS, 2008a -d)were selected as the use values that Marxan simulations attempted to conserve.These use areas varied in size, and significant overlap occurred among many values.To identify near-optimum spatial configurations of wildlife harvesting areas, multiple Marxan optimizations were run for each disturbance scenario.Marxan selections began by identifying the near-optimal spatial output that selected at least 50% of PUs in each harvesting area.

TABLE 3 .
Disturbances mapped in the study area and their recovery score, severity score, weight, and future weight, which were used to calculate the disturbance score in each planning unit.To represent continued recovery in future disturbance scenarios, existing disturbance weights were multiplied by the recovery score.

TABLE 4
. Impacts to planning units (PUs) across multiple analysis zones.We calculated the percentage of disturbed PUs and the percentage of PUs containing high disturbance levels (disturbance score ≥ 80) in every analysis zone.We also assessed the degree to which disturbance types overlapped by measuring the percentage of PUs in each analysis zone affected by 1, 2, 3, 4, and 5 different disturbance types.PA = planning area.

TABLE 5 .
Percentage of planning units (PUs) in the study area affected by each disturbance type.