Landscape characteristics govern the impacts of beaver ponds on surface water methylmercury concentrations in boreal watersheds

Studies in boreal regions concerning the bioaccumulative neurotoxin methylmercury (MeHg) in natural wetlands and experimental reservoirs have shown that these waterbodies contribute to high MeHg levels in underlying sediments, inundated vegetation, and aquatic organisms. Beaver ponds are natural reservoirs that are ubiquitous in the Canadian boreal region and have been reported to increase downstream MeHg concentrations. However, the reported impacts of beaver ponds on stream MeHg vary widely across a limited number of studies, and factors influencing this variation are not well understood. To quantify the effect of beaver ponds on stream mercury concentrations, water samples were taken upstream and downstream of 10 in-channel beaver impoundments in northwestern Ontario, Canada. The downstream:upstream MeHg concentration ratios were related to pond and landscape characteristics to examine potential factors that play a role in determining the effect of beaver ponds on stream MeHg concentrations. Overall, MeHg concentrations were 1.6 times greater downstream of the beaver ponds, though this increase was not consistent; downstream concentrations up to 12 times greater and up to 5 times less were also observed. Landscape characteristics that can be readily obtained from existing spatial datasets or quantified using remote sensing techniques emerged as better predictors of downstream:upstream MeHg concentrations than site-specific stream chemistry parameters or pond characteristics that are more difficult to ascertain, with drier landscapes indicative of lower background MeHg export being more likely to exhibit greater increases in MeHg downstream of a beaver pond. These results suggest that the effects of beaver ponds on surface water MeHg concentrations are generally small but highly variable, and that the magnitude of the pond’s influence on stream MeHg are lessened in landscapes already conducive to higher MeHg concentrations.


Introduction
The North American beaver (Castor canadensis) and its European cousin (Castor fiber) build dams that dissipate stream energy and create slower-moving ponded areas (Gurnell 1998), expanding lentic habitat along streams in forested watersheds (Martell et al 2006).Beaver dams and the resulting ponds are ubiquitous throughout headwater streams in North America; dam density may exceed 10 ponds per km in streams up to the fourth order (Westbrook et al 2006, Smith and Mather 2013, Burchsted and Daniels 2014).Dams are likely to grow in number and density as beaver populations increase both in Canada (Naughton 2012) and abroad (Halley et al 2021).Beaver ponds have been reported to improve forest resilience to disturbances such as droughts and wildfires and thus there are increasing efforts to re-introduce beaver populations and mimic effects of beaver engineering (Wohl 2021).
These impermanent reservoirs, however, also present ideal opportunities for the production of methylmercury (MeHg), a neurotoxin that bioaccumulates through aquatic food webs.Beaver ponds dissipate stream energy (Gurnell 1998), which raises water temperature (Błe ¸dzki et al 2011) and allows for the accumulation of sediment and organic matter (Karran et al 2018).The subsequent decay of submerged organic matter results in anoxic conditions (Collen and Gibson 2000, Rosell et al 2005) ideal for mercury (Hg) methylation (Herrero Ortega et al 2018), an anaerobic microbial process that occurs in oxygen-deficient environments (Bravo and Cosio 2020).In addition, beaver ponds have been reported to retain sulphate and raise concentrations of dissolved organic carbon (DOC) (e.g.Margolis et al 2001); both may affect the availability of Hg to methylating microbes and co-transport of Hg and DOC (Stoken et al 2016, Graham et al 2017).Beaver ponds also affect stream carbon and nutrient sources and cycling, altering downstream food webs (Anderson andRosemond 2010, Fuller andPeckarsky 2011) that may influence MeHg biocaccumulation (Jardine et al 2012).
There exists a wide range of literature on MeHg in wetlands, but very little on the effects of beaver ponds on surface water MeHg concentrations.Though similar, beaver ponds differ from natural wetlands in terms of carbon dynamics (Johnston 2014, Vehkaoja et al 2015), hydrologic environment (Puttock et al 2021), biogeochemical processing (Wegener et al 2017), and habitat diversity (Wohl 2021).Roy et al (2009aRoy et al ( , 2009b) ) examined MeHg concentrations upstream and downstream of ponds in southwestern Quebec and found that downstream MeHg concentrations could be up to 27 times greater than upstream concentrations, though the magnitude of this increase was affected by season, in-pond vegetation, and pond age.Painter et al (2015) also noted a significant downstream increase in MeHg in water, periphyton, and invertebrates for ponds in the Rocky Mountains, although the reported increase was much lower, 1.5-2 times that of upstream concentrations.Levanoni et al (2015) reported no significant downstream differences for recolonized ponds, but significant downstream increases (up to 3.5 fold higher) in MeHg for ponds that had recently been inundated for the first time.
This wide range of reported pond impacts on stream MeHg suggests a need to identify additional possible explanatory factors beyond pond characteristics.Recent landscape mercury studies have indicated that landscape factors can be used to predict surface water THg and MeHg concentrations in Canadian boreal watersheds.For example, Harrow-Lyle et al (2023) employed a structural equation modelling approach to demonstrate that watershed characteristics have a significant influence on stream MeHg concentrations.Fink-Mercier et al (2022) found that wetland coverage was a more accurate predictor of riverine total mercury (THg) and MeHg concentrations than physico-chemical variables.In the absence of reliable wetland coverage data, Lam et al (2022) suggested that % forest dominated by tree species favouring wet soil conditions could be an effective proxy for estimating streamwater THg and MeHg concentrations and loads.These studies suggest that landscape factors that can be relatively easily obtained (e.g. via remote sensing methods) could be used in addition to or in place of certain pond characteristics (e.g.age, colonization status), which are more difficult to ascertain, to predict the effects of beaver dams on stream MeHg concentrations.This study examined 10 beaver ponds in south-central Canada to test this hypothesis, by (1) quantifying the changes in surface water MeHg as it passes through beaver ponds and examining if any changes are accompanied by explanatory changes in stream chemistry and (2) identifying pond and landscape characteristic predictors of among-pond variation in downstream to upstream MeHg concentration ratios.We hypothesize that MeHg concentrations will be higher downstream of the ponds, especially in watersheds with landscape conditions less conducive to MeHg export in the absence of flooding by beaver ponds.

Study sites
This study was conducted in the Lake Nipigon ecoregion, which lies north of Lake Superior in northwestern Ontario, Canada and is part of the Boreal Shield ecozone.The Lake Nipigon ecoregion is underlain by Precambrian Shield bedrock with isolated morainal deposits (Crins et al 2009).Coniferous and mixed forests dominate the landscape, characterized by stands of black spruce (Picea mariana), white spruce (Picea glauca), balsam fir (Abies balsamea), trembling aspen (Populus tremuloides), white birch (Betula papyrifera), and jack pine (Pinus banksiana) (Crins et al 2009).Based on Environment and Climate Change Canada 1981-2010 climate normals, the mean annual, summer, and winter temperatures are 3.2 • C, 16.6 • C, and −11.2 • C, respectively.The mean annual precipitation is 796 mm, of which 30% falls as snow.This study occurred during a slightly warmer and much drier year than usual; the mean summer temperature was 17.2 • C and the annual precipitation was 544 mm.
Ten active beaver ponds on second or third order streams in the Mackenzie River watershed were selected for this study.Watersheds were delineated from the downstream point of the pond and ranged from 110 to 354 ha in size (figure 1).Watersheds were predominantly forest covered, with harvest having occurred in up to 30% of the watershed area over the past 10 years.This study included 2 sets of nested catchments, i.e., some study ponds were downstream of other study ponds.The Global Moran's I statistic was used to test for spatial autocorrelation (via the 'ape' package in R (Paradis et al 2019)) for all water quality parameters described below; in all cases, values were not found to be spatially correlated.Thus, all ponds were considered to be independent of the others.
Dams at all ponds were actively maintained, with clear signs of recent beaver activity in and/or around the ponds.This was assessed via beaver sightings, as well as observations of freshly cut saplings, freshly stripped branches in the water, and growing lodges across sampling periods (Naughton 2012).Beavers have been active in the area for decades and it is unlikely that any of the study ponds are pioneer systems.

Field sampling
To meet objective #1, water samples were collected from running water of the inlets (upstream) and outlets (downstream) of each pond 5 times from early July to late October 2021 (see breakdown of sampling periods in supplemental information, table S1).In-pond samples were also taken near the edge of the pond on each occasion.Samples for THg and MeHg were collected using 'clean-hands-dirty-hands' methods (Gill and Fitzgerald 1985) in PET or PETG bottles rinsed three times with streamwater.Half of each sample was filtered through ashed 0.7 µm glass fibre filters.All THg and MeHg samples were preserved with 0.5% trace-grade hydrochloric acid and stored refrigerated in the dark prior to analysis.
Samples for all other analytes were collected in 1 l HDPE bottles thrice-rinsed with stream water and filtered through weighed 0.7 µm glass fibre filters for determination of total suspended solids (TSSs).The filtered water was stored in polystyrene tubes for determination of anion (chloride, nitrate, sulphate, and phosphate) concentrations and in amber glass bottles for determination of DOC concentrations.

Laboratory analyses
Filtered and unfiltered samples were analysed for THg by cold vapour atomic fluorescence spectrometry using a Tekran 2600 automated total Hg analyser following EPA Method 1631 (United States Environmental Protection Agency (USEPA) 2002).MeHg analysis for filtered and unfiltered samples was carried out by aqueous phase ethylation and gas chromatography separation with speciated isotope dilution mass spectrometry (Hintelmann andEvans 1997, Hintelmann andOgrinc 2003) using an Agilent 7700x ICP-MS.Deionized water blanks and sample duplicates were included every 10-15 samples for both THg and MeHg analyses.Mercury analytical quality control measures are included in supplemental information (table S2).
Filter-passing samples are presented here as THg f or MeHg f ; unfiltered samples are presented as THg TOT or MeHg TOT .Particulate Hg (THg p or MeHg p ) is presented as the difference between Hg TOT and Hg f and may therefore be below the detection limits reported in table S2.
Samples were analysed for DOC concentrations using a Shimadzu TOC analyser following EPA Method 415.1 (USEPA 1974) and for nitrate, sulphate, phosphate, and chloride concentrations using a Metrohm 930 Compact ion chromatograph following EPA Method 300.1 (United States Environmental Protection Agency (USEPA) 1999a).Calibration curves from 0.5 to 50 mg l −1 were used in these analyses.Standard method 2540D was used for analysis of TSS, with a detection limit of 0.002 g l −1 (USEPA 1999b).

Pond and watershed characterization
For objective #2, potential predictors of pond effects on stream chemistry were split into 3 categories: pond, ecosite, and landscape characteristics (table 1).Pond characteristics were derived from drone imagery taken during the same summer as the water samples.Landscape characteristics were derived from the Ontario Watershed Information Tool (MNRF Provincial Mapping Unit 2023).Ecosite characteristics were derived from the Forest Resources Inventory (Ontario Ministry of Natural Resources and Forestry 2019).
Ecosites are a type of ecological unit that is part of Ontario's Ecological Land Classification system, which delineates areas of similar ecology at different scales, for example from Ecozones (100 000 s to 1000 000 s km 2 ) for broad applicability in international reporting, to Ecoelements (0.1-10 s ha) for field identification of rare species habitat.Ecosites are the second-finest level of the system (10 s to 100 s ha) and serve as 'tactical and operational planning units for resource conservation and management for all land types' (Crins et al 2009).They are characterized by similar topography, substrate, hydrology, and vegetation.This study extracted the moisture (dry, moist, or wet) and material (sand/silt, loam, sand/loam, organic, or organic mineral) elements of each ecosite for the analyses described below.For example, ecosite type B040 is described as 'Boreal-Dry, Sandy: Aspen-birch hardwood' and was here categorized as having dry moisture conditions and sandy material, whereas type B129 is described as 'Boreal-Rich Conifer Swamp' and was here categorized as having wet moisture conditions and organic material.If a watershed was 50% type B040 ecosites and 50% type B129 ecosites, it would then be classified as 50% dry and 50% wet for moisture conditions, and 50% sandy and 50% organic for soil composition.

Statistical methods
All data analyses were performed using R version 4.2.0 (R Core Team 2022).Data were assessed for normality using histograms and the Shapiro-Wilks test and transformed using logarithms if appropriate.In most cases, the assumption of normality could not be met through data transformation and non-parametric tests were used.Pearson regression and correlation were used for normally distributed data, and Spearman rank-order correlation was used for non-normally distributed data.Differences in stream chemistry between paired upstream and downstream observations were assessed using the Wilcoxon signed-rank test.
Ratios of downstream to upstream concentrations were used to assess pond effects on stream chemistry and are hereafter discussed as 'D:U ratios' in text and presented as 'D:U' (Downstream:Upstream) in figures.To meet objective #2, backward stepwise regression was used to predict pond MeHg TOT D:U ratios based on each category of parameters in table 1. Pairwise correlations were computed between all parameters within a category to test for collinearity, with one member of a pair (marked with an asterisk ' * ' in table 1) being dropped from the analysis if the absolute value of the associated correlation coefficient was greater than 0.7 (Dormann et al 2013).Which of the pair to drop was determined to maximize the remaining number of parameters.All parameters within a category were initially included in the model, and the parameter with the highest p-value dropped for the subsequent model.This was repeated until the value for Akaike's information criterion adjusted for small sample sizes (AICc, calculated using R package 'MuMIn' (Bartoń 2022)) did not decrease further when additional variables were removed, with the final model considered the most parsimonious.

Upstream, in-pond, and downstream water chemistry
There were no clear patterns across all ponds in THg TOT , DOC, TSS, or anion concentrations along the upstream to pond to downstream gradient.For example, median THg TOT concentrations decreased from upstream (2.66 ng l −1 ) to pond (1.92 ng l −1 ) to downstream at 5 K (1.74 ng l −1 ), but increased upstream-pond-downstream (1.35-1.44-1.69ng l −1 ) at UM, and increased upstream to pond (1.14-1.49ng l −1 ) but decreased pond to downstream (0.98 ng l −1 ) at W51.Similar inconsistencies emerged for most other measured water quality parameters as well, with the direction of change sometimes increasing (e.g.SO 4 2− at TASTAN), sometimes decreasing (e.g.SO 4 2− at W6), and sometimes fluctuating (e.g.SO 4 2− at SF) along the upstream-pond-downstream gradient (see supplemental information for comparisons, figures S1-S3).In most cases, downstream and in-pond water chemistry were found to be quite similar to upstream water chemistry.Though D:U ratios of anion concentrations were quite variable, median D:U ratios when all sites were considered jointly fell within the range of 1.00-1.02,suggesting that the observed dams did not result in changes to Cl − , NO 3 − , SO 4 2− , or PO 4 3− concentrations.The same applied to DOC and TSS concentrations; though the range in TSS D:U ratios (0-32.23 times greater downstream) was much greater than the range in DOC D:U ratios (0.56-2.09 times greater downstream), median D:U ratios for both were low: 1.01 and 1.19, respectively.These findings suggest that (a) pond effects on water chemistry may be highly variable, (b) the direction of change resulting from beaver ponds at any given time is inconsistent, and as such, (c) upstream and in-pond water quality conditions are not indicative of downstream conditions.
Though some studies (e.g.Roy et al 2009a, 2009b, Błe ¸dzki et al 2011) have reported significant downstream increases in N and DOC (from enhanced microbial activity), decreases in SO 4 2− (from reducing conditions), and decreases in TSS (from sediment accumulation), others have reported the opposite results ( Čiuldienė et al 2020) or no significant changes (Painter et al 2015) in water chemistry downstream.The inconsistent results across the literature, paired with the wide ranges of D:U ratios with median D:U ratios almost equal to 1 observed here, may be a reflection of how ponds exhibit variable source-sink dynamics as a function of flow (Wegener et al 2017).
The exceptions to these similarities were THg f , MeHg f , and MeHg TOT , which increased from upstream to downstream at almost all ponds (THg f : figure S1; MeHg f and MeHg TOT : figure 2).Specifically, when all sites and all measured water quality parameters were considered, significant downstream increases were only observed for THg f (p = 0.014), MeHg f (p < 0.001), MeHg TOT (p = 0.001), and % MeHg (p < 0.001).When each pond was considered individually, significant increases in any of these parameters were only observed at 2 or 3 sites, likely due to the low number of samples taken at each site.THg f D:U ratios ranged from 0.24 to 3.64, though the median D:U ratio across all ponds was much lower at 1.12.This downstream increase in THg f concentrations may be due to (a) the lower pH in the pond (Fracz and Chow-Fraser 2013) leading to decreased binding of Hg(II) to dissolved organic matter (Haitzer et al 2003), and/or (b) the association of THg with finer filter-passing colloidal material resulting from the breakdown of organic matter in the pond (Babiarz et al 2001).Both (a) and (b) would shift the partitioning of THg to favour the filter-passing fraction, resulting in the observed downstream increase in THg f but not THg TOT .
The ranges of MeHg f and MeHg TOT D:U ratios (0.10-18.00 and 0.21-12.19,respectively) were larger than the range of THg f D:U ratios, and the median D:U ratios were greater as well (1.66 and 1.65, respectively), indicating that filtered and total MeHg generally increased downstream of the observed ponds.Beaver ponds create temporary wetlands, which have long been thought of as ideal environments for Hg methylation (Chen et (Mangal et al 2022).However, as these processes occur on a small scale when compared to overall watershed factors, the overall downstream increase in MeHg is small.

Upstream, in-pond, and downstream changes in Hg-DOC relationships
THg f concentrations were significantly correlated with DOC concentrations upstream (ρ = 0.32, p = 0.026) and downstream (ρ = 0.39, p = 0.006).MeHg f concentrations were significantly correlated with DOC concentrations at all sampling positions (upstream: ρ = 0.45, p = 0.002; in-pond: ρ = 0.32, p = 0.027; downstream: ρ = 0.31, p = 0.033).A significant linear Hg-DOC relationship has been consistently reported in watershed studies (e.g.Dittman andDriscoll 2009, Riscassi andScanlon 2011), but not in pond studies (e.g.Roy et al 2009a, Sinclair et al 2012).Roy et al (2009a) suggested that this may be because in beaver ponds, Hg dynamics are controlled by reducing conditions rather than by co-transport with DOC.The significant relationships observed here suggest that co-transport with DOC may still be an important factor, though the decline in strength and significance of the relationship along the upstream-pond-downstream gradient suggests a slight decoupling as water passes through the pond.This may be due to within-pond processing of DOC leading to a reduction in DOC aromaticity (Kothawala et al 2006, Pugh et al 2021) and binding with MeHg (Mangal et al 2022).
When all ponds were considered jointly, there were no differences in Hg:DOC ratios among pond positions; median THg:DOC ratios were 0.1 ng mg −1 and median MeHg:DOC ratios were 0.02 ng mg −1 upstream, in-pond, and downstream.The median observed THg:DOC ratio is slightly lower and the MeHg:DOC ratio exactly on par with global averages reported by Lavoie et al (2019).When ponds were considered individually, there were also no clear patterns in median THg:DOC among pond positions.However, the median MeHg:DOC ratios increased along the upstream-pond-downstream gradient for 6 of the 10 ponds.The 4 ponds where MeHg:DOC ratios did not increase along the upstream-pond-downstream gradient were ROCKY, SF, W51, and W10A.These ponds' respective watersheds are characterized by higher % organic-mineral soil coverage (7%-12%) than the others (2%-6%).This suggests that the input of organic matter to the ponds and in-pond methylation play a role in in-pond MeHg-DOC complexation and MeHg bioaccumulation in biota (Chételat et al 2018).

Pond, ecosite, and landscape predictors of D:U MeHg
Backward stepwise multiple linear regressions indicated that pond characteristics did not account for much of the variance observed in MeHg D:U ratios.The most parsimonious model of pond predictors (ρ = 0.32, p = 0.36) included only pondshed area and did not produce a significant relationship (figure 3(a)).Regression models were improved when metrics derived from ecosite classifications were used instead of pond characteristics.The most parsimonious model of ecosite predictors (ρ = 0.67, p = 0.039) included, in order of relative importance, % organic mineral soil and % moist ecosite (figure 3(b)).Decreases in both predictors led to greater pond MeHg D:U ratios.Previous studies have shown that increases in these predictors generally correspond with higher MeHg concentrations, as methylation occurs in wet anoxic environments and MeHg is often transported with organic molecules.As such, in landscapes already conducive to higher MeHg concentrations in surface waters, the effect of beaver ponds on downstream MeHg may be more muted.This was supported by regression results from landscape predictors, in which average watershed slope and % brush/alder produced the most parsimonious model (ρ = 0.48, p = 0.17) for pond MeHg D:U ratios, though the relationship was not significant (figure 3(c)).
Similar results emerged when the most important predictors from all three groups were jointly considered, with increasing watershed slope (corresponding to less surface water storage), decreasing % organic-mineral soil, and decreasing % brush/alder (indicating drier soils) producing the most parsimonious model for median MeHg D:U ratios (ρ = 0.85, p = 0.0035; figure 3(d)).Thus, in landscapes less conducive to microbial methylation and therefore producing lower background MeHg concentrations in surface waters, the effect of beaver ponds on downstream MeHg may be heightened.

Beaver ponds as a window into Hg and DOC transport on the landscape scale
The MeHg D:U ratio models indicate that watersheds with lower background MeHg concentrations in surface waters are the most sensitive to the addition of beaver ponds, i.e.D:U ratios, an indicator of a pond's effect on MeHg concentrations, will be higher in these watersheds (figure 4).Conversely, streams in watersheds with higher background MeHg concentrations are less affected by the creation of beaver ponds, i.e., D:U ratios will be closer to 1. Put simply, as upstream MeHg concentrations are indicative of watershed conditions without the pond, they should be negatively correlated with the pond MeHg D:U ratio, which was the case in the ponds observed here (figure 5    regions (e.g.Levanoni et al 2015, Painter et al 2015), it would be reasonable to assume that these findings could be applied to beaver ponds in the Canadian boreal forest more generally.However, without further field investigations, this hypothesis remains untested.
Though organic-mineral soils and moist environments are known to favour methylation, it was unexpected that % organic-mineral soil and % moist ecosite emerged as more significant than % wet ecosite, which was strongly correlated with % organic soil (ρ = 0.92, p < 0.001).The latter predictors might be expected, based on existing literature on the prevalence of wetlands as methylation hotspots (Chen et al 2012, St. Louis et al 1994, Hall et al 2008, Mitchell et al 2008), to characterize upstream landscapes that export higher concentrations of THg and MeHg and thus would be less sensitive to the addition of a beaver pond and result in lower pond Hg D:U ratios.Ecosites classified as wet with organic material include, but are not limited to, bogs and conifer swamps (Crins et al 2009).These are wet environments and thus might be considered conducive to methylation but may be less hydrologically connected to surface waters; bogs are fed primarily by precipitation and swamps may be seasonally flooded and thus less hydrologically connected to the stream network in dry years such as the one in this study (Environment Canada 1987).Comparatively, ecosites classified as wet with organic-mineral material include but are not limited to meadow marshes and thicket swamps (Crins et al 2009); these are areas that are directly fed by, and output to, the stream network but where stream energy is low.Assuming methylation potential is similar between both types of ecosites, hydrologic connectivity to surface waters may be greater in the latter.
While dry ecosites tend to be characterized by sandy soils and hardwood stands that provide neither ideal conditions for methylation nor corridors of strong surface hydrologic connectivity, moist ecosites tend to be characterized by shrubs and mixed forests with coarser material that may be more conducive to surface and near-surface flows that facilitate connections between consistently wet areas and the stream network (Crins et al 2009).Thus, the total wet coverage of a watershed may be less important than (a) the type of wet coverage and (b) whether wet areas are connected to the stream network for determining stream MeHg concentrations.If stream MeHg concentrations are already high, the addition of a beaver pond that provides ideal methylating conditions will not have as sizeable an effect as it would in a stream where MeHg concentrations were lower.

Conclusion
This study aimed to (1) quantify the effects of beaver ponds on stream chemistry and downstream MeHg concentrations and (2) identify pond and landscape characteristic predictors of among-pond variation in pond MeHg D:U ratios.Pond D:U ratios for all measured water quality parameters were highly variable but relatively low when observations at all ponds were considered jointly.Landscape predictors explained much of the variance in pond MeHg D:U ratios, with landscapes already conducive to higher stream MeHg concentrations being less affected by the beaver ponds.Though this study took place across a limited number of ponds in a small study area in northern Ontario, the ponds and forest landscapes examined are very similar to conditions that may be encountered throughout the central Canadian boreal forest.As beaver densities increase, existing landscape metrics such as those derived from ecosite classifications may be of use in modelling the impacts of beaver dams on the transport of MeHg.

Figure 1 .
Figure 1.Watersheds delineated from the downstream sampling points of the 10 study dams.
al 2012, Tjerngren et al 2012) due to (a) the flooding of labile organic carbon (Hall et al 2005), (b) the subsequent anoxic conditions (St. Louis al 1994, Driscoll et al 1998) and (c) the higher and more stable water temperature (Collen and Gibson 2000, Sun et al 2023), all favouring microbial activity.Beavers also continually add fresh organic matter to the inundated area via direct delivery for food and dam construction (Johnston 2017) as well as by creating foraging canals around the pond that facilitate additional transport of upland material (Grudzinski et al 2020), further fuelling MeHg formation (Herrero Ortega et al 2018).While foraging, beavers will preferentially cut deciduous species (Gerwing et al 2013) and may shift forest stand structure over time, creating a conifer buffer up to 60 m wide around the pond (Gable et al 2023).Conifer cover has been linked to greater MeHg in water and fish (e.g.Drenner et al 2013, Eagles-Smith et al 2016) as (a) needles provide greater surface area for Hg adsorption than deciduous leaves (Tsui et al 2008, Tabatchnick et al 2012) and (b) conifer needle inputs may provide key MeHg transport molecules

Figure 3 .
Figure 3. Results of linear regression models for pond D:U MeHg ratios based on (a) pond characteristics, (b) ecosite characteristics, (c) landscape characteristics, and (d) best predictors from (a)-(c).Regression lines for significant relationships are in blue, with the 95% confidence interval in light grey.

Figure 4 .
Figure 4. Upstream watershed conditions influence the effect of a beaver dam on MeHg concentrations.

Figure 5 .
Figure 5. Correlation between upstream MeHg TOT and D:U MeHg TOT .Values are log-transformed for normality.The regression line is in blue, with the 95% confidence interval in light grey.

Table 1 .
Pond, ecosite, and landscape characteristics considered as predictors of pond effects on stream chemistry.Parameters marked with an asterisk (' * ') were dropped from statistical analyses to reduce multicollinearity; see footnotes for correlation results.
, R = −0.68,p < 0.001).As MeHg concentrations in this study were comparable to those reported in other Canadian boreal headwater streams (e.g.Roy et al 2009a, Lam et al 2022) and MeHg D:U ratios were comparable to those reported from other beaver ponds in northern