Do beaver ponds increase methane emissions along Arctic tundra streams?

Beaver engineering in the Arctic tundra induces hydrologic and geomorphic changes that are favorable to methane (CH4) production. Beaver-mediated methane emissions are driven by inundation of existing vegetation, conversion from lotic to lentic systems, accumulation of organic rich sediments, elevated water tables, anaerobic conditions, and thawing permafrost. Ground-based measurements of CH4 emissions from beaver ponds in permafrost landscapes are scarce, but hyperspectral remote sensing data (AVIRIS-NG) permit mapping of ‘hotspots’ thought to represent locations of high CH4 emission. We surveyed a 429.5 km2 area in Northwestern Alaska using hyperspectral airborne imaging spectroscopy at ∼5 m pixel resolution (14.7 million observations) to examine spatial relationships between CH4 hotspots and 118 beaver ponds. AVIRIS-NG CH4 hotspots covered 0.539% (2.3 km2) of the study area, and were concentrated within 30 m of waterbodies. Comparing beaver ponds to all non-beaver waterbodies (including waterbodies >450 m from beaver-affected water), we found significantly greater CH4 hotspot occurrences around beaver ponds, extending to a distance of 60 m. We found a 51% greater CH4 hotspot occurrence ratio around beaver ponds relative to nearby non-beaver waterbodies. Dammed lake outlets showed no significant differences in CH4 hotspot ratios compared to non-beaver lakes, likely due to little change in inundation extent. The enhancement in AVIRIS-NG CH4 hotspots adjacent to beaver ponds is an example of a new disturbance regime, wrought by an ecosystem engineer, accelerating the effects of climate change in the Arctic. As beavers continue to expand into the Arctic and reshape lowland ecosystems, we expect continued wetland creation, permafrost thaw and alteration of the Arctic carbon cycle, as well as myriad physical and biological changes.

Modeling studies indicate that greenhouse gas emissions following permafrost thaw will amplify current rates of atmospheric warming, a process known as the permafrost-carbon feedback [11]. North American beavers (Castor canadensis) have colonized the Arctic tundra during the last half-century [12]. In the Arctic, beaver dams often trigger local inundation and subsequent permafrost thaw [12][13][14], a mechanism known to mobilize permafrost carbon (C) to the atmosphere in the form of the greenhouse gasses carbon dioxide (CO 2 ) and methane (CH 4 ). Shortterm and long-term measurements linking beaver pond construction in the Arctic with CH 4 release are lacking. Here, we use remotely sensed data to examine the potential for beaver colonization in the tundra to increase Arctic CH 4

emissions.
Northern permafrost regions contain vast amounts of soil organic carbon. Permafrost is thawing across these regions, evident as increased thaw slumps [15], formation and disappearance of thermokarst lakes [16], thawing ice wedge polygonal ground [17], increased erosion rates, and warming temperatures in boreholes. Permafrost thaw surrounding streams and rivers is evident in the altered timing and partitioning of streamflow [18]. Geochemical signatures of increased subsurface compared to surface flow indicate that thawing permafrost enhances flowpaths and connectivity underneath stream channels [19]. These processes are expected to continue with future warming, but non-linearities associated with abrupt thaw processes hamper predictions of the strength of the permafrost-carbon feedback [20]. As beavers colonize streams in the Arctic, they construct ponds along streams, likely enhancing subsurface flows [21] and advective heat transfer to permafrost.
When permafrost thaw induces subsidence and ponding, anoxic conditions favor the microbial conversion of thawed permafrost organic C to greater quantities of CH 4 as opposed to CO 2 [22,23]. Though the fine-scale spatiotemporal variability of total CH 4 flux remains difficult to quantify in-situ at landscape scales, recent advances in airborne remote sensing have quantified CH 4 hotspot occurrences across regional scales [24,25]. Hotspot behavior is indicative of sites with intensive biogeochemical activity that are likely complemented by biotic factors (e.g. vegetation and microbial metabolism) and abiotic factors (e.g. physical thermokarst characteristics and labile C) that drive extreme emissions [25]. Beaver colonization and the associated flooding and hydrologic changes alter adjacent vegetation composition, conflating physical and biologic drivers of CH 4 emissions.
North American beavers are a keystone species whose engineering is known to heavily influence streams, rivers, riparian corridors, and lakes in North America, Eurasia, and South America [26]. Beavers induce hydrologic and geomorphic alteration of the landscape by flooding terrestrial surfaces and converting lotic to lentic environments, which alter biogeochemical cycling in non-permafrost temperate ecosystems [27]. Many beavers build dams to increase the size and depth of waterbodies to elude predators, access upland forage areas, facilitate transport of wood for forage and dam construction, increase winter unfrozen water habitat, and promote early successional forage species [28].
Through their engineered expansion of waterbodies, beaver dams often increase water depth, flood nearby vegetation, raise groundwater levels, and perennially thaw regions that would otherwise freeze in winter, all of which are known mechanisms for promoting CH 4 emissions. Furthermore, the presence of beaver dams drives accumulation of organicrich sediment in impounded stream reaches, which is often hundreds of times greater than that found in free-flowing channels [29]. Sediment accumulation in lentic environments promotes microbial decomposition of organic matter, leading to anoxic conditions suitable for methanogenesis. As a result of a likely combination of these factors, beaver ponds have been shown throughout the non-permafrost region to emit >30 times more CH 4 per unit area than other types of wetlands, lakes, and lotic systems [27,[30][31][32][33][34]. While information on northern beaver pond CH 4 emissions are scarce, theory suggests that emissions may be higher yet in beaver ponds beneath which organic C-rich permafrost is rapidly thawing. Recent aerial remote sensing observations over Big Trail Lake near Fairbanks, AK indicate persistent AVIRIS-NG CH 4 hotspots driven by thermokarst hydrologic dynamics [25], but no connections were made to the active beaver engineering in the area.
Beavers have constructed over 12 000 ponds in the Arctic tundra of Alaska, including a doubling between 2003 and 2017, as they colonized new regions and transformed lowland permafrost ecosystems [12,35]. People in remote subsistence-based communities of Alaska and Canada have witnessed the influx of beavers and associated hydrologic engineering with concern due to effects on fish, water quality, and boat access [36,37]. Beaver engineering, where prevalent, is the dominant control on increases in surface water, which is closely linked to permafrost dynamics and ecosystem processes [14]. Time-series images of beaver colonization in the Arctic tundra show that inundation and hydrologic route diversion associated with pond construction may trigger permafrost thaw and thermokarst within and surrounding the nascent wetland [12][13][14]. Beaver ponds alter the local thermal balance and increase heat transfer to the underlying permafrost [38,39]. This hydrologydriven heat transfer is more effective at warming permafrost than the pre-existing vegetation cover, driving increased permafrost thaw via both lateral pond expansion and subsidence, vertical talik formation, and formation of thermokarst ponds. A case study from an ice-rich permafrost stream in western Alaska demonstrates how rapidly beaver hydrologic engineering can trigger thermo-erosional processes to thaw permafrost and restructure lowland streams [13]. Once initiated, actively thawing permafrost margins on ponds and lakes can quickly become sites of high CH 4 emissions, which can remain active for decades [40].
We explore the relationship between beaver engineering and CH 4 release in a permafrost landscape by examining spatial correlations between remotely-sensed CH 4 hotspots and beaver ponds. We compare the occurrence of CH 4 hotspots, derived from AVIRIS-NG data, in beaver-affected waterbodies and non-beaver affected waterbodies over 429.5 km 2 of an Arctic tundra region in Northwestern AK. We predicted that AVIRIS-NG CH 4 hotspots are more abundant in the terrestrial areas surrounding beaver ponds compared to terrestrial areas surrounding non-beaver affected waterbodies. Results demonstrate how this rapidly expanding new disturbance regime is altering the carbon cycle along Arctic streams.

Study area
We selected an area (429.5 km 2 ) of the lower Noatak River basin in Northwestern Alaska where six overlapping AVIRIS-NG flightlines were acquired on 24 July 2018 (figure 1). The study area includes abundant lowland wetlands, riparian wetlands, lakes, and loworder streams that drain alpine and subalpine watersheds. The study site is located in the continuous permafrost zone [41]. Mean annual air temperature is −5.06 • C and mean annual precipitation is 279.4 mm at nearby Kotzebue, AK [42].

Beaver disturbance mapping
Beaver ponds (n = 118) were identified and delineated from very high resolution cloud-free satellite imagery from June, July, and August 2019 (<1 m, GeoEye-1) using expert photo interpretation at a scale 1:2000 following previous methods [35](figure 2). Complete coverage of cloud-free imagery was unavailable for 2018. Beaver pond density is moderate in the study area compared to elsewhere in the western Alaska tundra, and the number of ponds within and nearby this study area increased from 153 to 364 between 2003 and 2017 [35].

Airborne remote sensing of CH 4 hotspots
AVIRIS-NG imagery was retrieved from the study region on 24 July 2018 [45]. These data include six overlapping flight lines with a ∼5 m spatial resolution. Identification of AVIRIS-NG CH 4 hotspots followed previously published methodology [24,46,47]. AVIRIS-NG CH 4 hotspots are defined as pixels with ppm m (integrated concentration path length units) CH 4 enhancement >2500 ppm m [24,46,47]. All enhancements above the 2500 ppm m threshold are integrated as binary detections [47]. Surface reflectance was estimated as in [48]. CH 4 bias was removed using models described in [25].
Ground measurements were conducted outside the scope of this study, at Big Trail Lake in interior Alaska. Big Trail Lake sits within Goldstream Valley, a boreal forest marsh underlain by discontinuous permafrost. Beavers are often found in a lodge on the bank of the lake, and their engineering is common in nearby Goldstream Creek, including 30 m downstream of the lake outlet. Big Trail Lake and the Goldstream Valley is a warmer version of the study area here, probably including less permafrost but ample shrubs. We acknowledge some of these differences, but we also assume that the relationships that were established between AVIRIS-NG hotspots and CH 4 emissions at Big Trail Lake apply similarly in our study area [25].

Water layer development
To compare beaver and non-beaver (control) waterbodies, we created a water layer for the study landscape (figure 2) utilizing the National Hydrography Dataset (NHD) [49] and beaver pond polygons. Flowlines and waterbodies were rasterized at ∼5 m resolution to form the basis of the water layer. Beaver pond polygons were rasterized at ∼5 m resolution from very high resolution imagery (GeoEye-1) and added to the water layer. The main channel of the Noatak river was removed from the water layer through photo interpretation and manual editing of the water layer. The water layer was manually checked for errors by visual comparison with true color AVIRIS-NG images.
To minimize the potential effects of comparing beaver waterbodies to distant non-beaver waterbodies with dissimilar environmental features and to exclude possible inadvertent inclusion of adjacent beaver waterbodies, the all water layer was refined to produce a second water layer (nearby water). For the nearby water layer, non-beaver waterbodies were defined as water pixels >100 m from beaver water pixels, <450 m from beaver water pixels, and with no apparent beaver dams, lodges, or evidence of flooding. The refinement was to provide a more direct comparison to beaver water pixels in terms of spatial proximity, geographic and ecological similarity, and total surface water area (beaver affected water area: 1623 687 m 2 , non-beaver affected water area: 1600 038 m 2 ).

Streams and lakes
Water pixels were classified as streams or lakes through expert assessment. Streams were derived from the flowlines of the NHD and considered as flowing waterbodies including beaver ponds on flowing waterbodies. Lakes were derived from the waterbodies of the NHD and considered as non-flowing waterbodies (figure 2). Of the beaver ponds, 44 were classified as lakes and 74 were classified as streams.  [43]. Flightlines include ang20180724t183517, ang20180724t184947, ang20180724t190505, ang20180724t191928, ang20180724t193427, and ang20180724t194858. Esieh Lake is described in Sullivan et al [44].

Distance to and elevation above water
Distance to water was calculated for each pixel to use in hotspot analysis. Pixels <10 m from water were excluded from analysis to account for any mixed water or land pixel effects. Elevation above water was calculated for each non-water pixel. Elevation was obtained from the 5 m IFSAR digital elevation model [50].

Analysis
We calculated CH 4 hotspot occurrence ratios (ratio of hotspot pixels to total pixels, CH 4 HOR) for beaver pond margins and non-beaver waterbody margins [24,47]. Water margins were defined as the area around water features out to a distance of 60 m. CH 4 HOR were compared between beaver waterbodies and non-beaver waterbodies using an uppertailed modified z-test [51]. We predicted that beaver CH 4 HOR were greater than non-beaver CH 4 HOR. Statistical tests were performed for 5 m distance bins from each water pixel.

Results
AVIRIS-NG CH 4 hotspots covered 0.496% of the area adjacent to waterbodies in the study. Comparing the beaver waterbodies to nearby non-beaver waterbodies, in this western Alaska study area, we find, on average, 51% greater CH 4 HOR around beaver waterbodies to a distance of 25 m (p < 0.01, figure 3). Comparing beaver waterbodies to all nonbeaver waterbodies (including waterbodies >450 m from beaver-affected water), we find significantly greater CH 4 HOR around beaver waterbodies to a distance of 60 m (p < 0.001, figure 3). Nearby nonbeaver waterbodies have a slightly elevated CH 4 HOR relative to all non-beaver waterbodies to a distance of 60 m (p < 0.01; figure S1).

Streams and lakes
We find a strong enhancement of CH 4 HOR adjacent to beaver stream segments relative to nonbeaver streams to a distance of 30 m (p < 0.05, figure 3). However, lakes show no significant differences in CH 4 HOR between beaver and non-beaver . CH4HOR for beaver and non-beaver water pixels with significance levels calculated at each 5 m distance. Beaver waterbodies have elevated CH4HOR from 10 m-25 m relative to nearby waterbodies and from 10 m-30 m+ for all waterbodies and non-beaver streams. For the number of pixels analyzed for each group see table S1. lakes at most distances (figure 3). Beaver streams show greater CH 4 HOR (10-30 m) compared to beaver lakes (p < 0.05, figure S2). Non-beaver streams and lakes show no consistent difference in CH 4 HOR.

Elevation above water
The elevation of hotspots relative to the elevation above the nearest waterbody (∆Z) was calculated to understand the potential influence of flooding and raised water tables on AVIRIS-NG CH4 hotspots. A majority of AVIRIS-NG CH 4 hotspots occur at <3 m above nearby waterbodies to 30 m distance (figure 4).

Hotspot spatial decay ratio
We calculated the power law fits for the beaver and non-beaver CH 4 HOR (figure 5). The power law fit for 10-35 m shows strong exponential decay with distance for both beaver and non-beaver hotspots (R 2 = 0.87, 0.85). The decay rate of the power law fit (10-35 m) for the ABoVE domain (Northwestern Canada and Alaska) is very close to that of the decay rate of the beaver fit (−0.649 vs −0.6254). Beyond 40 m, CH 4 HOR declined gradually ( figure 5). Overall, we see an enhancement of CH 4 HOR in this study landscape compared to the ABoVE domainwide average [24]. The non-beaver CH 4 HOR are ∼50% greater while the beaver CH 4 HOR are ∼400% greater than those for the entire ABoVE domain.

Discussion
Detecting CH 4 efflux at landscape and regional scales using hyperspectral imaging and spectroscopy is a relatively new technique that has not been applied to test the effects of beaver ponds. This study provides evidence relating AVIRIS-NG hotspots, shown to represent high CH 4 efflux in a similar boreal forest setting, to beaver ponds on a landscape scale. Specifically, we used airborne imaging spectroscopy surrounding 118 beaver pond locations over an area of 429.5 km 2 and found that beaver ponds in the Arctic enhance AVIRIS-NG hotspots from adjacent terrestrial surfaces. We interpret the increased hotspots surrounding beaver ponds as being indicative of increased CH 4 emissions, likely due to a combination of wetland formation and permafrost thaw. New beaver damming and construction of ponds along Arctic streams, including a doubling of beaver ponds in the Alaskan Arctic from 2003-2017 [35], is thus increasing CH 4 emissions from adjacent terrestrial surfaces and amplifying the warming already underway. Future work should include ground measurements of methane flux and a better understanding of the processes driving methane production and release surrounding tundra beaver ponds. Figure 5. Power law fits for CH4HOR for beaver and all non-beaver water pixels. Power law functions and R 2 are presented in the figure for beaver (blue), non-beaver (green) water pixels for 10-35 m and 40-250 m distance from water. Power law functions from [24] in black and from [47] in gray.

Beaver engineering and CH 4 in non-permafrost ecosystems
Beaver engineering shifts streams from free-flowing to standing water systems. If beavers persist in an area, they create dynamic landscape mosaics that transition over time between free-flowing streams, impounded water in ponds, and wetlands [52]. Beaver engineering increases the width of the riparian zone and the length of riparian and upland interface, creating ponds and wetlands [53], which promote anoxic conditions and CH 4 production through methanogenesis [30,31]. The evolving and complex nature of beaver engineered ecosystems, however, makes it difficult to generalize spatially and temporally about the effects on the carbon cycle (of which CH 4 is only one component). Nummi et al [34] estimate that beaver ponds globally range from a C sink (−0.47 Tg yr −1 ) to a C source (0.82 Tg yr −1 ). Whitfield et al [26] estimated that the resurgence and introduction of beavers since 1900 accounted for emission of 0.18-0.80 Tg CH 4 yr −1 in 2000, a 200-fold increase from unengineered streams in 1900. A study from the boreal forest of Quebec found that beaver ponds were responsible for emitting 18% of the total CH 4 flux [54]. The high uncertainty on existing estimates of beaver pond CH 4 fluxes is partly due to the difficulty of monitoring CH 4 fluxes in-situ or remotely and the challenges of extrapolating spatially and temporally limited samples. Our results indicate that in the Arctic tundra environment, where permafrost predominates, beaver ponds enhance CH 4 release.

Comparison to other Arctic CH 4 hotspot studies
In this study, CH 4 HOR declines exponentially with distance up to 35 m from standing water whereas at distances >35 m the decay ratio is more than an order of magnitude slower (−0.02, figure 5). The rapid decay for distances <35 m from standing water is consistent with previous CH 4 hotspot work [25,47], and is likely due in part to topography near waterbodies, which reduces the potential for wetlands and CH 4 production beyond 35 m. The majority of beaver ponds in the study landscape occur along streams in valleys with greater slope than the few beaver ponds surrounded by relatively flat topography. The relatively higher CH 4 HOR found in the Noatak landscape could be explained by thawing of permafrost with higher ground ice content, greater permafrost C stocks, younger beaver ponds, or greater general methanogenic activity compared to the ABoVE domain in aggregate [45]. Our findings are similar to those from the uplands of the Mackenzie Delta, Canada, where the non-beaver CH 4 HOR are slightly lower (∼0.005-0.008), and the beaver CH 4 HOR are slightly higher (∼0.009-0.02) for 10-35 m (∼0.007-0.015) [47].

CH 4 hotspots and beaver ponds in the Arctic
High CH 4 flux rates (1000-24 000 mg CH 4 m −2 d −1 ) have been linked to hotspots identified by AVIRIS-NG using in-situ measurements to validate hotspots [25]. Thermokarst lakes are known to have large CH 4 fluxes to the atmosphere [55,56]. At a thermokarst lake in central Alaska, elevated CH 4 flux measurements have been associated with thawing permafrost [57], though beaver engineering at the lake outlet and elsewhere on the adjacent stream system was not mentioned. Beaver engineering and the creation of interconnected wetlands is known to induce a cascade of environmental changes that are likely to promote CH 4 emissions. These effects are likely enhanced in the permafrost domain, where thermal effects of inundation are especially consequential for thawing permafrost C [58]. By constructing ponds along streams in permafrost, beavers are essentially emplacing thermokarst ponds along streams. Though we have associated increased AVIRIS-NG CH 4 hotspots with beaver ponds and wetlands, we cannot quantify the relative contributions of CH 4 production around nascent beaver ponds compared to CH 4 released from thawed permafrost adjacent to ponds.
The 51% increase of AVIRIS-NG CH 4 hotspots surrounding beaver ponds presented here is substantial. In this study area, beaver engineering was minimal prior to 2000, and by 2020 the density of ponds was about half of the more favorable areas to the south and east [35]. Strong increases in beaver engineering, which have occurred recently and are expected to continue, would extend the 51% local increase around beaver ponds to entire stream lengths, and possibly even across landscapes where lowland streams and beaver habitat are prevalent. The size of CH 4 emissions in the tundra due to beaver engineering depends heavily on how quickly they disperse to new areas and how favorable the habitat is in these regions. The damming of streams and the conversion of relatively small flowing waterbodies to wetland complexes could be the dominant mode of abrupt permafrost thaw (in ice-rich permafrost) and C release in young beaver-dominated stream systems undergoing colonization. The landscape response to beaver ponding varies according to the setting on a stream, slough, or lake outlet.
Beaver impacts to permafrost are most evident at the surface where active layer thickness is relatively shallow (e.g [13]), reflecting cold soil thermal conditions. In these locations, beaver engineering and expansion of surface water over near-surface permafrost leads to enhanced heat transfer into the underlying permafrost, leading to thermokarst in ice-rich terrain, and high potential for thaw and C release [13]. The presence of groundwater flow paths through gravel bars and floodplain soils likely reduces permafrost presence in braided river corridors prior to beaver arrival. Upon colonization by beavers, we expect that dams and ponds in these riverine settings (side-channels or sloughs) will have a lesspronounced effect on permafrost. However, lateral expansion of the pond over time may still function to thaw permafrost in adjacent riparian or upland soils.
In pre-existing lakes, beaver dams at the outlets typically have a reduced influence on water level, surface water extent, and lakeshore geometry. Because new terrestrial surfaces are usually minimally inundated in these situations, permafrost thaw is less common. This is consistent with our results showing that beaver dams at lake outlets are not correlated with increased AVIRIS-NG CH 4 hotspots. Where beavers dam outlets of drained thaw lake basins (e.g [14]), or where dammed outlets lead to catastrophic drainage, we expect to find permafrost thaw and CH 4 hotspots, but the study landscape contained no obvious examples of this. Beaver preference for damming drained thermokarst lakes and related landforms would contribute to enhanced CH 4 hotspot occurrence since these landscapes are rich in soil organic carbon and their inundation would be expected to promote significant methanogenesis.

Detecting CH 4 hotspots
Since the CH 4 survey excluded open water areas due to limitations of the hyperspectral imaging, potentially large ebullitive fluxes from pond sediments, thawing permafrost underneath ponds, and wetlands went undetected. However, some evidence suggests that if ebullition is prevalent within a few meters of terrestrial shorelines, and if wind conditions are favorable, this water-borne emission could supplement CH 4 enhancements over the nearby land surface where SWIR reflectances would allow CH 4 detection [25]. Even in this scenario, we are confident that the AVIRIS-NG adequately identified sites of intense CH 4 emission across our 429.5 km 2 study area and did not bias detections between beaver-affected and nonaffected locations. If anything, AVIRIS-NG's capability to only observe areas of relatively extreme CH 4 provides a robustness and conservatism to our results.
Quantifying the source of CH 4 (i.e. contemporary carbon, ancient permafrost carbon, and/or geologic carbon) emitted at our observed hotspots was outside the scope of this study. However, it is worth noting that geologic emissions may influence a small number of our observations, especially near Esieh Lake where a large geologic CH 4 seep has recently been discovered [44]. To our knowledge, the primary Esieh Lake seep is located within the lake body itself, making its detection by AVIRIS-NG challenging. This is because without glint, water surfaces do not reflect enough short-wave infrared photons to perform confident CH 4 retrievals. A small number (1-3) of hotspots were observed near the shorelines of Esieh Lake and may be related to the geologic seep either directly or due to dispersal of the seep's plume over the reflective land surface. While differentiating CH 4 sources of detected hotspots was outside the scope of our study, we conclude that beavers have a significant effect on their occurrence.
Further study is warranted to both investigate the direct mechanisms that link beaver activity to increases in AVIRIS-NG CH 4 hotspots in permafrost environments and estimate actual CH 4 emission ratios as a result of beaver activity. Our approach of only considering nearby water pixels for analysis is more conservative than including all water pixels, as demonstrated by the rapid decline and lower significance of hotspots nearby water compared to all water (figure 3). Moreover the use of only nearby water pixels and the exclusion zone (0-10 m) creates a conservative approach to detecting AVIRIS-NG CH 4 hotspots. There are likely many hotspots near water (distance <10 m) that are being excluded that could have been included and likely have elevated CH 4 fluxes due to their proximity to water. A more accurate and higher resolution water mask could help to refine this process and lead to better detection of AVIRIS-NG CH 4 hotspots at the margins of waterbodies, but the resolution of AVIRIS-NG may still be limiting.
Finally, it is conceivable, though unlikely, that beavers are targeting pre-existing CH 4 hotspots to construct their ponds, in which case beaver pond construction would not be responsible for the correlated hotspots. If we assume that beavers are not targeting CH 4 hotspots to construct their ponds, then the increases in AVIRIS-NG CH 4 hotspots are attributable to beaver pond construction and the associated changes to permafrost and lowland ecosystems. We believe the evidence presented here makes a strong case for this latter scenario where beaver engineering is causing the associated changes in permafrost and CH 4 emissions. Future work will exploit time series imagery, imagery along a gradient of permafrost coverage, and ground based gas flux measurements to resolve this issue definitively.

Conclusion
As the Arctic warms, local-scale changes have the potential to amplify large-scale climate forcings causing cascading effects to landscapes and ecosystems. As beavers proliferate and spread into new tundra areas, they actively flood and inundate permafrost terrain, driving increases in CH 4 release through permafrost thaw and thermokarst processes. This study is the first attempt to link large numbers of new beaver ponds to CH 4 emissions at the landscape scale, and these results suggest that beaver engineering in the Arctic will at least initially increase CH 4 release.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https:// doi.org/10.3334/ORNLDAAC/1569 [43].