Increase in beaver dams controls surface water and thermokarst dynamics in an Arctic tundra region, Baldwin Peninsula, northwestern Alaska

The Baldwin Peninsula was selected as the study area due to the optimum availability of sufficiently high spatial and temporal resolution satellite imagery and the presence of tundra vegetation and permafrost (figure 2). Beaver dams were mapped using high-resolution satellite imagery available from the DigitalGlobe Inc. imagery archive in two areas to account for spatial and temporal differences in cloud-free image availability between 2002 and 2019. The spatial resolution of the panchromatic imagery ranged from 40 cm to 70 cm and the frequent image acquisition dates provided an opportunity for cloud-free observations suitable for identifying beaver dams across the study period (figure 3). The imagery was orthorectified using the 2 arc-second resolution U.S. Geologic Survey National Elevation Dataset for Alaska and corrected to top-of-atmosphere reflectance to facilitate time series analysis. Beaver dam mapping positions were based on the orthorectification results from the 2019 images to account for subtle differences in image geolocation in prior years.

Zoom In Zoom Out Reset image size

Based on suitable and available imagery, the Kotzebue study area (100 km²) was mapped each year in 2002, 2007–14, and 2017–2019 (in total 12 individual years) and the entire northern Baldwin Peninsula (430 km²) was mapped in 2010, 2013, and 2019 (3 individual years). This hybrid approach was designed to optimize our spatial and temporal perspective and to assess whether the changes being observed near Kotzebue were reflective of changes occurring in the surrounding region. Beaver dams were manually mapped in each image in both study areas at a scale of 1:1000. All individual, apparently active dams were designated as a point feature in a GIS database in each image. If a dam failed or was abandoned in a subsequent year it was not counted in the latter time series unless it was re-established. Dams were also grouped as complexes to track the inhabitation of new dam building sites over time in the study region. The location of dams relative to permafrost-region landforms (thermokarst lakes, drained thermokarst lake basins, and tundra/beaded streams) was also determined manually. Surface water area change between 2002 and 2019 was then compared at beaver-influenced versus non-beaver-influenced waterbodies that were classified as either thermokarst lake, drained thermokarst lake basin, or tundra/beaded stream to assess whether there was a preference for dam building relative to thermokarst landforms and their potential connection to surface water changes.

We analyzed changes in surface water extent in the Kotzebue study area using the best available high-resolution panchromatic images available for a particular year. Image selection focused on cloud-free, ice-free, and calm surface water conditions with images being acquired between late-June and mid-August in a given year. All images were resampled to a spatial resolution of 70 cm to match the lowest resolution image in the time series prior to analysis. Within year image dates range from 25 June to 22 August with the average date of image acquisition being 17 July (table 1). Object-based image analysis was conducted in eCognition Essentials 1.3. Image pixels were grouped into objects using the multiresolution segmentation algorithm with a scale of 20, a color to shape parameter of 0.1, and a smoothness to compactness factor of 0.4. A supervised classification scheme was developed for each image by selecting water and non-water object samples. At least 500 samples were manually selected for each class in each image. Objects were classified using the object-based k-Nearest Neighbor (KNN) classifier with a k value of 1. The KNN is a simple machine learning algorithm where an object is classified by a majority vote of its neighbors, with the object being assigned to the class most common amongst its k nearest neighbors. Using a k value of 1 emphasizes stark contrasts between disparate object class types. An accuracy assessment was conducted using the eCognition workflow for each image object classification using the input training samples. The user and producer accuracy of surface water classification was >96% in each case. Classified objects were then merged and only those objects classified as water were extracted for further analysis in a GIS database. All objects smaller than 250 m² were removed from the surface water inventory, objects that fell outside of the overlapping extent of all images were removed from further analysis, and stable water feature centroids were used to reposition lake polygon files from prior years to match the orthorectification positioning of the 2019 images. Surface water area was summed in each time step and its changes were compared against the number of beaver dams mapped in the Kotzebue study area and their relation to thermokarst landforms between 2002 and 2019.

The number of beaver dams in the Kotzebue study area increased from 2 in 2002 to 98 in 2019, or ∼5000% (figure 4 and table 1). The rate of increase followed that of a 2nd order polynomial regression (R² = 0.99), with the years 2011 (+17.2), 2014 (+13.6), and 2018 (+15) exhibiting the largest annualized increase in dam numbers (figure 5). The area-weighted rate of dam increases for the Kotzebue study area based on the 17 year period was 5.6 dams per year per 100 km². The number of dams in the larger northern Baldwin Peninsula study area followed a similar trajectory between 2010 and 2019, increasing from 94 (2010), to 174 (2013), to 409 (2019) in the 430 km² study area (figures 4 and 5). The rate of increase also followed that of a 2nd order polynomial regression, and the area-weighted rate of dam increases was 8.1 dams per year per 100 km² over the shorter and more recent time interval. Over this same period, the area-weighted rate of dam increases was 9.2 dams per year per 100 km² for the Kotzebue study area.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As of 2019, there are now 42 distinct beaver dam/pond complexes in the Kotzebue study area. On average, 2.5 new beaver dam sites, or the initiation of new beaver dam complexes, were established per year over the 17 year period; however, there was substantial interannual variation (range of 13.0). In particular, 10.3 new sites were established between 2012 and 2013, and 13.6 new sites were established between 2013 and 2014, and less than one new site per year was established between 2002 and 2007. Analyzing the increase in the number of new dams relative to the number of new dam complexes shows periods of localized building versus expansion into new sites. The year 2011 and period 2014–2018 experienced a ratio of more than 2:1 new dams relative to new complexes in the study area, indicating development of existing dam complexes. In contrast, the years 2008 and 2013 experienced the opposite pattern, with more new sites being established relative to dam building at pre-existing complex sites. Between 2012 and 2013, the number of dams at existing sites and the establishment of new complexes were equally high at 13.6.

Total surface water area mapped in the Kotzebue study area showed a variable but increasing trend between 2002 and 2019 (figure 6 and table 1). Total surface water area increased from 594 ha in 2002 to 643 ha in 2019, or 8.3%. The mapped surface water area increase was significantly correlated with the increase in beaver dams over the 17 year period (R² = 0.61; p < .003). We further investigated the 8.3% increase in surface water area between 2002 and 2019 at beaver-influenced versus non-beaver-influenced waterbodies. Partitioning the 49 ha increase in surface water area between 2002 and 2019 into beaver influenced (+33 ha) versus non-beaver influenced (+16 ha) waterbodies establishes the role of beavers on controlling surface water area increases. Beaver-influenced waterbody areas increased by 14.3% over this period, whereas non-beaver influenced waterbody areas increased 4.6% (figure 7(a)). Thus, beaver-influenced waterbodies accounted for the majority (66%) of the increase in surface water area in the Kotzebue study area between 2002 and 2019.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

All but one beaver dam was associated with at least one of three thermokarst landforms in the study region: thermokarst lakes, drained thermokarst lake basins, or tundra/beaded streams (figure 7(b)). Beaver damming of drained lake basin outlets accounted for 68% of the increase in beaver-influenced waterbody area increases since 2002. In some cases, beaver dams led to more than a doubling in surface water area in specific drained lake basins over the study period. Damming of lake outlets accounted for 26% of the increase in surface water area. Damming at lake outlets tended to amplify natural expansion occurring along thermokarst lake margins due to permafrost degradation. Damming of beaded stream courses, though representing only a small fraction of total beaver influenced surface water area, increased in area by more than 150% (1.1–2.7 ha) between 2002 and 2019.

Our analysis adds the Baldwin Peninsula in northwestern Alaska to the scattered recent evidence that beavers are establishing themselves as a key disturbance agent in low arctic tundra regions (Jung et al 2017, Tape et al 2018). In the Kotzebue study area, we found an 8.3% increase in surface water extent between 2002 and 2019, with short term annual variation on the order of −0.2 to +3.7%. Over the twelve years of images analyzed, only three (2009, 2012, and 2017) experienced a reduction in total surface water based on the previous year, with one of the years being influenced by a partial lake drainage event. The 4.6% increase in non-beaver-influenced waterbodies likely results from natural thermokarst processes associated with expanding lakes in ice-rich permafrost terrain as well as annual variability in surface water area to some degree. The 14.3% increase in surface water area in beaver-influenced waterbodies likely results from the cumulative effects of beaver damming and inundation of low-lying areas, as well as the degradation of permafrost and expansion of waterbodies through thermokarst processes. Surface waterbodies in the Arctic are known to fluctuate seasonally and annually, but also to interact with permafrost through thermokarst processes that may cause both increases and decreases in surface water area (Bowling et al 2003, Smith et al 2005, Jones et al 2011, Cooley et al 2019, Nitze et al 2018). While it is difficult to separate the combined effects of seasonal and annual variability and geomorphic factors when analyzing regional surface water dynamics using remote sensing snapshots in time, the general trend over the 17 year study period for the Kotzebue study area was increasing surface water area resulting primarily from an increase in beaver dams (R² = 0.61, p < .003).

Few studies have directly addressed beaver-permafrost interactions (Lewkowicz and Coultish 2004, Jones et al 2018, Tape et al 2018). In this study, which represents a typical ice-rich permafrost landscape in the Arctic interspersed with thermokarst lakes, drained thermokarst lake basins, and beaded streams, we showed that beavers preferentially build dams and facilitate inundation of thermokarst landforms, which likely promotes permafrost degradation. Water impoundment on the landscape is known to drive permafrost degradation (Lachenbruch et al 1962, Jorgenson et al 2010, Arp et al 2016). Surface water inundation more readily conducts heat and therefore impacts the ground thermal regime almost instantaneously (Langer et al 2016). If the water depth is sufficient such that the mean annual water bottom temperature exceeds 0 °C (Burn 2002), permafrost thaw commences and in the case of ice-rich permafrost, thermokarst occurs (Grosse et al 2013). Several studies have highlighted the prominent role of permafrost degradation, and in particular abrupt permafrost thaw and thermokarst development on mobilizing carbon previously stored in permafrost (Schuur et al 2015, Walter Anthony et al 2018, Turetsky et al 2020). Future research should focus on the potential role of beavers on altering the carbon cycle in Arctic and Boreal regions (Gatti et al 2018, Nummi et al 2018).

Beaver colonization and dam building has substantial implications for the short-term dynamics and long-term evolution of permafrost lowlands (figure 8). On the Baldwin Peninsula, northwestern Alaska, fossil beaver dams and beaver-gnawed wood dating to the early Holocene (10 kya) have been found in association with pond sediments and in ice wedge casts below the pond sediments, which were inferred to represent ice wedge melting as a result of beaver dam building (McCulloch and Hopkins 1966). We likewise found that damming of drained thermokarst lake basin outlets accounted for the largest increase in surface water area. Presumably this extensive inundation due to beavers is accompanied by extensive permafrost degradation in the inundated basin floors and adjacent tundra slopes (figure 8(a)). Beaver damming of existing lake outlets was also observed to increase lake surface area, similarly promoting sub-aqueous and lateral degradation of permafrost (figure 8(b)). Lake drainage was also observed through dam failure, abandonment, or bank overtopping (figure 8(c)). While bank overtopping is a commonly attributed lake drainage mechanism (Hinkel et al 2007, Jones et al 2011, 2020), the ability of beavers to exacerbate this process has only been recognized once before (Lewkowicz and Coultish 2004) and at least for sites like the Baldwin Peninsula it could help explain some of the recent widespread lake drainage occurring in the region (Swanson 2019). Future work will need to take into account the potential role of beavers on driving these large regional differences in thermokarst lake dynamics and thus, permafrost landscape changes.

Zoom In Zoom Out Reset image size

Numerous regional remote-sensing based studies of changing lake surface water dynamics have been conducted in Arctic and Boreal regions (Smith et al 2005, Riordan et al 2006, Carroll et al 2011, Jones et al 2011, Roach et al 2011, Nitze et al 2017, 2018, Cooley et al 2019, Pastick et al 2019, Swanson 2019), yet beavers have seldom been considered among the primary drivers (Hood and Bayley 2008). Our observations showed that newly beaver influenced waterbodies accounted for 66% of the increase in surface water area between 2002 and 2019. This is comparable to observations from the Boreal forest in Canada when assessing the role of beavers relative to climatic variable controls on open water extent. Hood and Bayley (2008) found that beavers were responsible for 80% of the increase in open water extent in their study area in east-central Alberta, which dwarfed the effects of climate variables. Our results challenge the idea that current and future surface water changes in some Arctic and Boreal regions are, or will be, directly linked to changes in permafrost and thermokarst resulting from increased air temperature, through-going talik penetration, and precipitation and/or evaporation trends (Yoshikawa and Hinzman 2003, Smith et al 2005, Riordan et al 2006, Jones et al 2011, Roach et al 2011, Nitze et al 2018, Cooley et al 2019, Pastick et al 2019). Including the potential impact of beaver-driven engineering on Arctic and Boreal aquatic ecosystems may help explain the heterogeneity of lake area changes reported for a number of regions at the treeline and tundra ecotone or continuous to discontinuous permafrost zone transition.

While we have not addressed the causes of increases in beaver dam building in this study, it could be linked to three key components of changes occurring in the Arctic. First, the increased growing season length and increase in shrub cover and extent provide new forage and shelter necessary for beavers to survive in these landscapes (Tape et al 2006, 2018). Second, changes in winter climate are dominating the annual climate change signal in this region (Stafford et al 2000). The increase in winter air temperatures combined with more winter snowfall in Arctic Alaska is likely increasing the over-wintering habitat potential for beavers in a number of different types of thermokarst water bodies and as a result of thinner seasonal winter ice growth (Arp et al 2018). The shorter winters also reduce the duration that beavers must remain in their lodges relying on caches from the previous summer. Third, the change in native subsistence and hunting lifestyles that has occurred in this region and the reduction in the desire of beaver pelts and underfur for fashionable attire could be contributing to our observations of increased beaver activity (Bockstoce 2009, Tape et al 2018). Other factors such as changes in disease or predation seem less likely but cannot be eliminated as possibilities. We hope to tease apart the driving forces responsible for the increase in beaver dam activity in future studies to assist in the development of models to predict future Arctic landscape dynamics.

An interesting analogue to our observations of rampant beaver dam building in tundra regions in Alaska comes from Tierra del Fuego, Patagonia. The introduction of ∼20–50 beavers from northern Manitoba, Canada in southern Patagonia in the 1940s has swelled to an estimated 100 000 beavers in the region over a 60–70 year period (Choi 2008, Pietrek and Fasola 2014). Interestingly, one possible factor behind the explosive population growth has resulted from beavers moving out of the forests in which they were initially introduced, and establishing in adjacent steppe ecosystems which were previously thought to be subprime habitat (Pietrek et al 2017). The ability of beavers to utilize a range of habitat types in Patagonia, and to thrive in largely herbaceous vegetation-dominated ecosystems, has implications for what appears to be an ongoing beaver invasion of low Arctic tundra ecosystems.