Identifying Potential Habitat for the Endangered Aleutian Shield Fern Using Topographical Characteristics

The historical and current distribution of the endangered (U.S. Endangered Species Act [ESA] 1973, as amended) Aleutian shield fern Polystichum aleuticum (hereafter, shield fern; Figure 1) was summarized by Byrd and Williams (2007) as follows. Shield ferns have only been documented on two islands (Atka and Adak) in the Andreanof group in the Aleutian archipelago of Alaska (hereafter, Aleutians). The Atka population, however, has not been located again following its initial detection in 1932, despite numerous attempts. Therefore, only four extant populations are currently documented, all of which are located on Mount Reed in the Alaska Maritime National Wildlife Refuge portion of Adak. Despite numerous surveys, shield ferns have not been detected on other islands in the Aleutians. Hence, the species is considered one of the rarest and most restricted (in range) ferns in North America.

A number of factors likely contribute to the rarity of the species. The steep and treacherous mountainous terrain occupied by shield ferns limits the amount of area that can be safely traversed during field surveys (Talbot and Talbot 2002). Furthermore, shield ferns are solitary-growing ferns that only grow to approximately 15 cm tall (Anderson 1992) and occur where visibility is regularly hindered by dense fog banks (A. Duarte and M.A. Ricca, personal observation). Therefore, the species' rarity may be an artifact of an observer's ability to detect populations. Also, the seismically active nature of the Aleutians causes large-scale soil mat slumping and rock fallouts in steep areas (Byrd and Williams 2007). This geological process can create new habitat for shield ferns while simultaneously destroying existing populations. Additionally, introductions of nonindigenous land mammals to numerous islands in the Aleutians (Murie 1959; Black 1984; Swanson and Barker 1992; Ebbert and Byrd 2002; Major and Jones 2005) have altered ecosystem processes that affect plant species distributions (Croll et al. 2005; Maron et al. 2006).

The restricted known range of the shield fern raises concern for the preservation of the species. The U.S. Fish and Wildlife Service management plan for the shield fern outlined several management and outreach actions that included, “surveys for new populations of ferns by refuge staff and others will continue opportunistically on Adak and on other refuge islands” (Byrd and Williams 2007:9). The remote and rugged nature of the Aleutians renders surveys difficult. Habitat suitability maps would assist with directing survey efforts, by targeting areas more likely to have shield fern populations, and identify sites of high conservation value for the species. Herein, we present a framework to model potential habitat suitability for the shield fern in the Andreanof Islands of the Aleutians using topographical habitat variables found at known extant locations of shield fern populations.

The Andreanof Islands are a group of islands in the Aleutians (Figure 2). The islands are located at the confluence between the North Pacific Ocean and Bering Sea, extend about 430 km across exterior coastlines, and consist of roughly 4,000 km² of mountainous terrain with steep slopes. The climate and floristic community of the Aleutians are characteristically maritime. Summers are cool (5–10°C), wet, and foggy, whereas cyclonic storms occur frequently during winter but temperatures hover near 0°C. The tundra vegetation comprises community types such as coastal and alpine grass–forb meadows, dwarf shrub and lichen-dominated heaths, and windswept fellfields. The growing season typically occurs from June to September.

We derived three topographic habitat variables (elevation, slope, aspect) that have been suggested to affect habitat suitability of shield ferns based on previous literature and current shield fern population locations. First, all extant shield fern populations reside between 338.0 m to 525.8 m in elevation (Talbot and Talbot 2002; Byrd and Williams 2007). This relationship between occupancy and elevation may not be driven by the actual elevation but rather the environmental characteristics (such as temperature, precipitation, amount of bare ground, competition, etc.) found within a specific range in elevation. Thus, we considered elevation as an important factor influencing shield fern occupancy, and classified habitats between 275 m and 600 m elevation as suitable.

Second, all known shield fern populations are reported to occupy northeast- (Talbot and Talbot 2002) or southeast-facing slopes (Byrd and Williams 2007). Talbot and Talbot (2002) suggested that the consistent association with aspect may be a result of the protection provided against the predominantly strong westerly winds during the snow-free summer growing season. Therefore, we only classified eastward-facing (22.5–157.7°) slopes as suitable habitat.

Third, shield fern populations are associated with horizontal ledges of exposed and weathered rock on the steep mountain slopes of Mount Reed (Anderson 1992). As the slope of the mountainside increases, the likelihood of soil mat slumping (thus increasing the exposure of rock) should increase. Therefore, we considered slopes of 25–90° as suitable habitat in the model.

Because higher resolution imagery data are not available for much of the Aleutians, a 60 m × 60 m National Elevation Dataset layer and a road network layer for the Andreanof Islands were retrieved from the U.S. Geological Survey's Seamless Data Warehouse (http://seamless.usgs.gov). Also, a 1∶63,000 land coverage and coastline shapefile for the state of Alaska was downloaded from the Alaska State Geo-Spatial Data Clearinghouse (http://www.asgdc.state.ak.us). Data were preprocessed with methods such as projection and transformation, mosaicking, and clipping, using the model builder procedure in ArcGIS 10.

Proximity to the coastline may affect the presence of shield ferns because allochthonous nutrient enrichment derived from windblown sea spray alters plant communities by favoring dense stands of graminoids that outcompete other taxa (Maron et al. 2006). We therefore implemented a 500-m buffer around all coastlines and considered these areas unsuitable. Habitat altered by the anthropogenic footprint should also be excluded as suitable sites because these areas have been heavily impacted (e.g., introduction of concrete slabs, pollution, and erosion, as well as regular scheduled mowing in some cases) by humans. Therefore, we considered areas within a 200-m buffer from roadways as unsuitable. However, it is worth noting that most of these areas within the anthropogenic footprint are at low elevations with relatively flat terrain. Therefore, nearly all these areas were excluded based on topographical characteristics alone. We extracted raster cells in the National Elevation Dataset layer that were outside the coastal and anthropogenic footprint areas to derive the elevation, slope, and aspect layers that we used in the analysis.

A suitability map was created that prioritized suitable areas by weighting each topographical habitat factor equally. Pixels meeting the topographical criteria described above were reclassified as either suitable (“1”) or unsuitable (“0”) for each habitat factor. The layers were then summed and pixels that met all topographical habitat criteria proposed in this study (Value  =  “3”) were classified as “Highly Suitable.” Areas that met at least two of the topographical habitat criteria proposed in this study (Value  =  “2”) were classified as “Moderately Suitable.” All remaining areas were considered “Unsuitable.”

We conducted a sensitivity analysis by creating a series of maps to examine how the amount of suitable habitat changed by assigning more weight to one topographical habitat factor while holding the weight for other factors constant. We changed the weights for each topographical habitat factor systematically. For example, when making the map in which aspect was considered more vital than elevation and slope, suitable raster cells in the aspect layer were given a value of “4,” suitable raster cells in the slope and elevation layers were given a value of “3,” and unsuitable raster cells for all layers were given a value of “0.” The layers were then summed and we considered a raster cell “Highly Suitable” if it received a value of “10” because it met all criteria. A raster cell was considered “Moderately Suitable” if it received a value of “7” because this meant the raster cell met at least two of the factors, one of which was the factor with greater weight. All other raster cells were considered “Unsuitable.” We used this protocol to create maps in which elevation or slope was considered more vital than the other factors considered and then compared the area considered suitable for each weighting scenario.

When weighting each topographical habitat factor equally to delineate suitable shield fern habitat, the known extant locations of shield fern populations on Mount Reed were classified as “Highly Suitable” (Figure 3). Also, 18 islands contained habitat that was considered both “Highly Suitable” and “Moderately Suitable” (Table 1; Figure 4). Islands from the greatest amount of area considered “Highly Suitable” to the least were Atka, Amlia, Adak, Tanaga, Great Sitkin, Kagalaska, Seguam, Kanaga, Gareloi, Amatignak, Umak, Little Tanaga, Chugul, Bobrof, Tagalak, Igitkin, Kanu, and Kasatochi. Areas considered “Highly Suitable” and “Moderately Suitable” totaled 80.76 km² and 408.50 km², respectively. However, Kasatochi Island recently erupted, which caused massive die-offs of resident flora and fauna (Scott et al. 2010); therefore, Kasatochi likely does not currently have shield fern populations despite meeting the topographical criteria used for this study.

All areas identified as “Highly Suitable” by the equally weighted topographical habitat factor scenario did not change for the weighting scenarios used in the sensitivity analysis. This is because weighting factors differently does not change whether or not an area meets all three topographical criteria used to delineate suitable habitat. However, the total area considered “Moderately Suitable” decreased when going from the equally weighted map to the elevation map to the aspect map and then to the slope map (Table 1).

All known extant locations of shield fern populations occurred in “Highly Suitable” areas under all mapping scenarios (Figure 3) considered in this study, which supports our criteria for selecting suitable localities based on topographical characteristics. The equally weighted habitat-factor scenario delineated 489.26 km² (12.8%) of land as suitable (both highly and moderately) shield fern habitat across the entirety of the Andreaonof Islands. Although this may seem like a relatively small amount of suitable land, this finding seems to be consistent with our understanding about the rarity of the species.

Ferns are generally not dispersal-limited (small spores are well known to travel long distances and fern diversity does not decline on islands as other taxa do; Williamson 1981; Cody 2006); therefore, it is logical to assume that the distribution of shield ferns is limited by one or more dimensions of niche space. Although we used ecologically reasonable topographical criteria (i.e., elevation, aspect, slope) to quantify habitat suitability, some caveats still exist. First, our topographical habitat criteria were based on only four locations that are highly clustered relative to the species' potential range. Therefore, difficulty arises when determining important niche parameters. Nevertheless, we delineated habitat based on the best information currently available concerning topographical characteristics. Second, using a 60 m × 60 m National Elevation Dataset layer may be too coarse of a resolution to identify smaller microhabitats within predominately “Unsuitable” raster cells. This may lead to an underestimation of suitable localities. Conversely, smaller unsuitable areas within a predominately “Suitable” raster cells may not be recognized for the same reason. Therefore, it would be useful to identify suitable locations using higher resolution spatial data when it becomes available. Third, although the use of the terrain slope as a proxy for exposed mountainside-rock seems probable, field observations examining possible correlations between these habitat characteristics are warranted. Fourth, shield fern occupancy is coupled with particular plant communities (Talbot and Talbot 2002), and other biotic and abiotic characteristics (e.g., competitors, soil type, depth, parent material, and nutrient availability) should also play a key factor in suitability of habitat for shield ferns. However, high-resolution geospatial data for these habitat characteristics is lacking for the Aleutians.

Introduced caribou and reindeer (domesticated caribou) Rangifer tarandus populations on Adak and Atka, respectively, pose a threat to shield fern populations. Caribou and reindeer populations have impacted some plant communities in the Aleutians through increased herbivory and soil compaction associated with trampling (Williams and Tutiakoff 2005; A. Duarte and M.A. Ricca, personal observation). Notably, caribou habitually occupy areas of Mount Reed (Williams and Tutiakoff 2005) and have recently been observed foraging in close proximity to existing shield ferns (L. Spitler, USFWS, personal communication). Moreover, caribou have recently immigrated to Kagalaska Island (located directly to the east of Adak; Ricca et al. 2012), which harbors 18.09 km² of potential shield fern habitat (both “Highly Suitable” and “Moderately Suitable” area under equally weighted habitat factor scenario). It is suspicious that reindeer have existed on Atka for a century, and shield ferns have not been located again since their initial detection in 1932. Perhaps, the lack of successful shield fern relocation efforts on Atka is because reindeer have eaten and trampled the past documented shield fern population. Detailed caribou and reindeer spatial use patterns (e.g., GPS-satellite telemetry data) would allow refinement of model-based estimates of shield fern habitat suitability and associated risk from these large mammalian herbivores.

The sensitivity analysis conducted in this study identified considerably less suitable area when weighting slope more than elevation and aspect (Table 1). This means that the area within the range of terrain slope classified as suitable in this study is more limited than the other two topographical characteristics considered. If the terrain slope is indeed a more important factor determining the occupancy of shield ferns, then shield fern habitat is scarcer than what the results are describing herein.

Our study presents a habitat suitability map for the endangered Aleutian shield fern using topographical characteristics, which can assist resource agencies with current and future recovery efforts for the species by providing a quantitative means of identifying future survey areas. Ideally, we would have refined our model by excluding areas where previous unsuccessful shield fern surveys have been conducted (see Byrd and Williams 2007). However, these data are not publicly available. Further, because ground-truthing (i.e., actively surveying field sites and comparing them with the suitability map) did not occur during this study, we recommend that if species introduction efforts are deemed necessary and pursued in the future, then microhabitat characteristics in identified suitable locations should be measured and compared with documented shield fern habitat.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author.

Reference S1. Anderson BL. 1992. Aleutian shield fern (Polystichum aleuticum C. Chr. In Hulten) recovery plan. Anchorage, Alaska: U.S. Fish and Wildlife Service.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S1 (2.5 MB PDF).

Reference S2. Byrd GV, Williams JC. 2007. Management plan for Aleutian shield fern (Polystichum aleuticum): an endangered species. Homer, Alaska: U.S. Fish and Wildlife Service Report.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S2; also available at http://alaska.fws.gov/fisheries/endangered/pdf/management_plan_aleutian_shield_fern.pdf (863 KB PDF).

Reference S3. Williams JC, Tutiakoff V. 2005. Aerial survey of barren-ground caribou at Adak Island, Alaska in 2005. Homer, Alaska: U.S. Fish and Wildlife Service Report.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S3; also available at http://alaskamaritime.fws.gov/visitors-educators/caribhunting/2005%20Caribou%20census.pdf (2.6 MB PDF).

Data S1. Geographic Information System (GIS) raster layer of the Aleutian shield fern Polystichum aleuticum habitat suitability map for weighting elevation, slope, and aspect equally. Areas within 500 m of coastlines and 200 m of roadways were removed from raster file.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S4 (1522 KB LPK).

Data S2. Geographic Information System (GIS) raster layer of the Aleutian shield fern Polystichum aleuticum habitat suitability map when weighting slope greater than aspect and elevation. Areas within 500 m of coastlines and 200 m of roadways were removed from raster file.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S5 (1536 KB LPK).

Data S3. Geographic Information System (GIS) raster layer of the Aleutian shield fern Polystichum aleuticum habitat suitability map when weighting elevation greater than slope and aspect. Areas within 500 m of coastlines and 200 m of roadways were removed from raster file.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S6 (1531 KB LPK).

Data S4. Geographic Information System (GIS) raster layer of the Aleutian shield fern Polystichum aleuticum habitat suitability map when weighting aspect greater than slope and elevation. Areas within 500 m of coastlines and 200 m of roadways were removed from raster file.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S7 (1536 KB LPK).

Data S5. Geographic Information System (GIS) polygon layer of the Andreanof Islands in the Aleutian archipelago of Alaska. Use this layer as a background layer when viewing the Aleutian shield fern Polystichum aleuticum suitability maps (for all scenarios) in order to have coastlines and roadways visible.

Found at DOI: http://dx.doi.org/10.3996/032012-JFWM-023.S8 (173 KB LPK).

We are grateful to the Fall 2011 Texas State University Geographic Information Systems (GEO 7417) class for their helpful comments and suggestions during the development of this project. We thank Alex Hartman (USGS), two anonymous reviewers, and the Subject Editor for critiques on earlier manuscript drafts. We would also like to acknowledge the U.S. Geological Survey's Seamless Data Warehouse and the Alaska State Geo-Spatial Data Clearinghouse for the GIS layers used in this analysis. The USFWS–Invasives with Volunteers Program and USGS Western Ecological Research Center provided funding for field work that indirectly spawned the ideas behind this manuscript.

Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.