Timber harvest alters mercury bioaccumulation and food web structure in headwater streams

Abstract

Timber harvest has many effects on aquatic ecosystems, including changes in hydrological, biogeochemical, and ecological processes that can influence mercury (Hg) cycling. Although timber harvest’s influence on aqueous Hg transformation and transport are well studied, the effects on Hg bio-accumulation are not. We evaluated Hg bioaccumulation, biomagnification, and food web structure in 10 paired catchments that were either clear-cut in their entirety, clear-cut except for an 8-m wide riparian buffer, or left unharvested. Average mercury concentrations in aquatic biota from clear-cut catchments were 50% higher than in reference catchments and 165% higher than in catchments with a riparian buffer. Mercury concentrations in aquatic invertebrates and salamanders were not correlated with aqueous THg or MeHg concentrations, but rather treatment effects appeared to correspond with differences in the utilization of terrestrial and aquatic basal resources in the stream food webs. Carbon and nitrogen isotope data suggest that a diminished shredder niche in the clear-cut catchments contributed to lower basal resource diversity compared with the reference of buffered treatments, and that elevated Hg concentrations in the clear-cut catchments reflect an increased reliance on aquatic resources in clear-cut catchments. In contrast, catchments with riparian buffers had higher basal resource diversity than the reference catchments, indicative of more balanced utilization of terrestrial and aquatic resources. Further, following timber harvest THg concentrations in riparian songbirds were elevated, suggesting an influence of timber harvest on Hg export to riparian food webs. These data, coupled with comparisons of individual feeding guilds, indicate that changes in organic matter sources and associated effects on stream food web structure are important mechanisms by which timber harvest modifies Hg bioaccumulation in headwater streams and riparian consumers.

1. Introduction

Atmospheric mercury (Hg) emissions from natural and anthropogenic sources have resulted in widespread deposition of Hg to the earth’s surface (Driscoll et al., 2013). Landscape perturbations play an important role in mobilizing Hg, as well as its transformation to methylmercury (MeHg; Hsu-Kim et al., 2018). Similarly, land-use change is one of the primary modulators of MeHg bioaccumulation in the face of global change (Eagles-Smith et al., 2018), often affecting the complex interplay among Hg inputs, production and transport of MeHg, and MeHg biomagnification through food webs (Benoit et al., 2003; Kidd et al., 2012). Timber harvest is one such widespread perturbation that has been linked to changes in Hg cycling (Bishop et al., 2009; Hsu-Kim et al., 2018).

Globally, approximately 50% (2.2 million hectares) of all forested land is utilized for timber production (Köhl et al., 2015). Timber harvest affects aquatic ecosystems through a variety of physicochemical alterations (Danehy and Johnson, 2013; Richardson and Béraud, 2014), as well as changes in food webs and community structure (Bilby and Bisson, 1992; Danehy et al., 2007; Göthe et al., 2009; Kiffney and Richardson, 2010). These changes can alter Hg cycling in affected watersheds through three primary mechanisms: 1) influencing aqueous Hg pathways, availability, or MeHg production (Eckley et al., 2018; Skyllberg et al., 2009); 2) altering dietary exposure to MeHg via changes in food webs that affect basal resources or trophic interactions (Bilby and Bisson, 1992; Erdozain et al., 2018, 2019; Göthe et al., 2009; Kiffney and Richardson, 2010); and 3) influencing growth and life history dynamics in some taxa via changes in stream habitat, chemical characteristics, or productivity (Gregory et al., 1987; Herendeen and Hill, 2004). Thus, timber harvest can be an important anthropogenic driver influencing Hg cycling and bioaccumulation.

Mercury bioaccumulation in aquatic food webs are often correlated with aqueous Hg concentrations (Chasar et al., 2009; Riva-Murray et al., 2011), which timber harvest can modify via changes in Hg deposition (Eagles-Smith et al., 2016b; Eckley et al., 2016), transport (Eckley et al., 2016; Obrist et al., 2016), or MeHg production (Eklöf et al., 2018; Kronberg et al., 2016; Skyllberg et al., 2009). Forest harvest also influences the relative availability of basal resources within stream food webs. In particular, changes in availability of organic matter derived from terrestrial litter versus instream production (Bilby and Bisson, 1992; Kiffney and Richardson, 2010), can alter food web structure (Göthe et al., 2009; Kreutzweiser et al., 2008a), and potentially the pathways of Hg bioaccumulation (Jardine et al., 2012; Riva-Murray et al., 2013) in both aquatic and associated terrestrial food webs. Such changes in relative availability of aquatic and terrestrial basal resources may be particularly important for Hg bioaccumulation because MeHg concentrations are generally lower in terrestrial organic matter compared to aquatic resources (Jardine et al., 2012; Ward et al., 2012). The overall productivity of aquatic systems can also be altered by forest harvest, either through increased instream photosynthesis or reduced inputs of terrestrial resources (Gregory et al., 1987; Kiffney and Richardson, 2010). Productivity-influenced changes to individual growth rates of organisms may thus also alter Hg bioaccumulation via growth dilution (Herendeen and Hill, 2004; Walters et al., 2015; Ward et al., 2012).

Whereas the impacts of timber harvesting on abiotic Hg cycling have been assessed in several studies (Bishop et al., 2009; Eklöf et al., 2016), little is known about how Hg concentrations in aquatic and riparian food webs change in response to forest harvest. Yet this is important for determining the cumulative effects of timber harvest on Hg risk in forested ecosystems. To address this information gap, we evaluated Hg bioaccumulation in the food webs of 10 headwater catchments in the Pacific Coastal Range, USA. Three of these catchments were left unharvested to serve as controls whereas the remaining catchments were subject to one of two harvest treatments – buffered clear-cut and clear-cut. The region is characterized by steep mountainous topography, high precipitation and wet Hg deposition rates (Eckley et al., 2016), and large foliar and soil Hg pools (Obrist et al., 2016). The region’s forests are also intensively managed for timber production, accounting for over 30% of the nation’s timber harvest despite occupying only 10% of the nation’s forest area (Smith et al., 2004). Our goals were to assess the effects of forest harvest on 1) Hg concentrations in food webs and individual feeding guilds, 2) abiotic and biotic drivers of Hg bioaccumulation in stream biota, particularly the role of forest harvest induced changes in food webs, and 3) Hg concentrations in songbirds utilizing the riparian areas of harvested streams.

2. Methods

2.1. Study area and experimental treatments

We conducted this study in the Trask Watershed Study Area located along the western slope of the Coast Range in northwestern Oregon, USA (Fig. S1). The Trask Watershed Study Area consists of 26 km² of publicly and commercially owned coniferous forest divided into small (0.2–0.5 km²) experimental catchments (Fig. S1) for assessment of timber management practices. Prior to harvest, the study area was dominated by 40–70 year old regenerating Douglas-fir (Pseudotsuga menziesii) stands, with riparian corridors composed of mixed conifers, red alder (Alnus rubra), and big leaf (Acer macrophyllum) or vine maple (A. ciccinatum; Hagar et al., 2012; Raggon, 2010). Individual catchments are representative of headwaters throughout much of the Coast Range, each with small (0.7–2.2 m wide and 2.2–6.5 cm deep; Chelgren and Adams, 2017), high gradient (12–22%) 2^nd order perennial streams draining areas from 0.2 to 0.7 km².

In the spring and summer of 2012 seven experimental catchments were subject to one of two harvest treatments (buffered clear-cut [subsequently referred to as buffered] and clear-cut; Chelgren and Adams, 2017). Three catchments received the buffered treatment, in which an 8-m wide riparian buffer was retained and the remaining catchment was clear-cut, resulting in an average reduction in total basal area of 82%. Four catchments were clear-cut in their entirety, leaving only mandatory wildlife habitat trees and reducing the total basal area of trees in the catchments by an average of 87%. Logging debris/slash was not removed from the harvested catchments. Three additional catchments without harvest activity were included as references for comparing to the harvested catchments.

2.2. Stream biota sampling

Between 2013 and 2015 we sampled biota representing five trophic guilds from each experimental catchment 1–3 times annually between mid-May and late August. These guilds encompass the breadth of consumers in each stream’s food web (Chelgren and Adams, 2017; Herlihy et al., 2005) including: a shredder (stonefly, Plecoptera: Pteronarcys spp.), a grazer (mayfly, Ephemeroptera: Heptageniidae), a collector-gatherer (mayfly, Ephemeroptera: Ephemerellidae), an invertebrate predator (stonefly, Plecoptera: Perlidae), and a vertebrate predator, the aquatic (i.e., larval or paedomorphic) Coastal Giant Salamander (Dicamptodon tenebrosus; hereafter referred to as salamanders), the top predator in these fishless systems (Parker, 1994). Salamanders were captured using a backpack electroshocker, and invertebrates were primarily collected using kick-nets. Except for mayflies, individual organisms were processed and analyzed. Individual mayflies had insufficient mass for chemical analyses, therefore we composited 5–20 similarly-sized individuals into 1–3 replicate samples of each taxon from a sampling date and catchment. We analyzed up to 10 samples of each taxon per catchment, although not all taxa were collected on each sampling date. Upon capture of salamanders, we measured snout-vent length (length) to the nearest mm and removed a 2-cm segment from the tip of each salamander’s tail. Total Hg concentrations in this tail segment are strongly correlated with whole-body THg concentrations and provide a non-lethal indicator of Hg exposure in coastal giant salamanders (Pfleeger et al., 2016). All samples were stored on dry ice in polyethylene vials while in the field then frozen at −20 °C until processing.

2.3. Water sampling

During a subset of sampling events we collected grab samples of flowing stream water for analysis of filtered THg (THg-F), particulate THg (THg-P), filtered MeHg (MeHg-F), dissolved organic carbon (DOC), sulfate (SO₄), and ultraviolet absorbance (UVA). Detailed sampling methods can be found in Eckley et al. (2018) and SI.

2.4. Bird sampling

Songbirds were sampled during the breeding season both pre-(July 2011) and post-harvest (June 2014). We captured songbirds in the riparian area of each catchment (typically within 50-m of the stream channel) using 30-mm mesh mist nets and playback recordings. Blood was collected from the brachial vein and stored on ice for approximately 3–6 h until they could be flash frozen on dry ice, then stored in the laboratory at −20 °C until analysis.

2.5. Mercury and stable isotope analyses

We analyzed THg-F, THg-P, and MeHg-F in water samples following EPA method 1631E, USGS method 5A-8, and EPA method 1630, respectively. All MeHg-F concentrations were below the 0.05 ng/L method reporting limit; however, we included uncensored values in our analyses as they provide the best available estimate of the true value. We analyzed DOC following EPA Method 415.3, SO₄ using EPA Method 300.0, and UVA with a 200–800 nm absorbance scan. We used UVA to calculate specific UVA at 254 nm (SUVA₂₅₄), normalized to DOC concentrations (Weishaar et al., 2003). Additional details on analytical methods and quality assurance results for water analyses are found in Eckley et al. (2018) and SI.

All invertebrate and salamander samples were oven dried at 50 °C until they reached constant mass (~48hrs), then homogenized to a fine powder using a ceramic mortar and pestle (salamander tails) or glass rod (invertebrates) and stored in a desiccator prior to analyses.

We measured MeHg in all invertebrate samples and THg in salamander tail clips and bird blood because nearly all Hg in these tissues is MeHg (Bank et al., 2005; Rimmer et al., 2005; Townsend and Driscoll, 2013). Total Hg was measured following EPA method 7473 (U.S. Environmental Protection Agency, 2000). Samples for MeHg analysis were digested in 4M nitric acid (Hammerschmidt and Fitzgerald, 2005) and analyzed following EPA method 1630 (U.S. Environmental Protection Agency, 2001). Invertebrate and salamander Hg concentrations are presented on a dry weight (dw) basis, whereas concentrations in bird blood are presented on a wet weight (ww) basis. Carbon (δ¹³C) and nitrogen (δ¹⁵N) stable isotope ratios were determined using isotope ratio mass spectrometry and are presented as ratios in standard δ-notation. See SI for additional details on chemical analyses including quality assurance measures.

2.6. Statistical analyses

Statistical analyses were conducted using natural-log transformed MeHg and THg concentrations in either JMP (SAS Institute Inc, 2016) or R (R Core Development Team, 2017) statistical software. Unless otherwise noted, Hg results are presented as back-transformed least-squares means (ng/g dw) with standard errors estimated using the delta method (Williams et al., 2002).

Mean salamander length varied among catchments and THg concentrations were positively correlated with length in some catchments (Fig. S2). Therefore, we size-adjusted individual THg concentrations to the median length of all salamanders sampled (39 mm, n = 418). See SI for additional details on size adjustment methods. The resulting size-adjusted THg concentrations were used for subsequent analyses.

We employed a tiered analytical approach to assess the effects of harvest treatments on Hg bioaccumulation and food web structure. In our first-tier analysis, we tested whether biotic Hg concentrations differed among the reference, buffered, and clear-cut catchments in the years following harvest activities. We used a linear mixed-effects model that included harvest treatment and feeding guild as fixed effects, a treatment × guild interaction, and catchment and collection year as random effects.

Our second-tier analysis examined the relationships between Hg concentrations in either predatory invertebrates or salamanders and THg-F, THg-P, MeHg-F, DOC, SO₄, or SUVA₂₅₄ measured in water samples collected within two weeks of biota sampling. Due to limited sample sizes, we used nonparametric Spearman’s rank correlations (ρ) to assess relationships between biotic and aqueous THg-F, THg-P, or MeHg-F. The relationships between biotic Hg and DOC, SO₄, or SUVA₂₅₄ were examined using general linear models in which we included a single water quality parameter as a continuous covariate, harvest treatment as a fixed categorical variable, and catchment as a random effect. We included the interaction between harvest treatment and each water quality parameter to assess whether the relationships between biotic Hg concentrations and aqueous covariates differed among harvest treatments.

Third, we compared Hg biomagnification rates among treatments using trophic magnification slopes (TMSs) calculated as the slope of the relationship between log₁₀-transformed Hg concentrations [MeHg for all invertebrates, un-adjusted THg for salamanders] and δ¹⁵N (Lavoie et al., 2013). Differences in the TMSs among treatments were assessed by testing the significance of the interaction between treatment and δ¹⁵N in a general linear model that also included catchment and collection year as random effects to account for possible influence of spatial-temporal variation in baseline δ¹⁵N values on the TMSs (Ménard et al., 2007).

In our fourth-tier analysis, we tested for differences in indices of food web structure among treatments. Specifically, we compared basal resource diversity (i.e., potential organic matter sources; range of δ¹³C) and food web length (range of δ¹⁵N; Jackson et al., 2011; Layman et al., 2007) among treatments. These parameters are sensitive to catchment disturbance and can influence Hg bioaccumulation. We also calculated the isotopic niche size (as Standard Ellipse Areas: SEA_B) of each feeding guild within the three experimental watersheds to test whether harvest treatments altered the utilization of specific resources. Bayesian estimates of these three parameters (basal resource diversity, food web length, and isotopic niche size) were calculated in the SIBER package for R with two chains, 2 × 10⁵ iterations, a burn-in of 10⁴ values, and posteriors thinned by 10. The resulting values are robust to variations in sample size and can be compared using Bayesian credibility intervals. We considered Bayesian estimates to be significantly different when the 95% credible intervals did not overlap between the reference and treatment catchments of a watershed (Jackson et al., 2011).

In the final tier of our analysis, we examined differences in the THg concentrations of riparian songbird blood sampled in harvested and unharvested areas of the Trask Watershed Study Area. For this analysis we were able to utilize samples collected for another study prior to harvest activities, allowing us to test whether blood Hg concentrations changed in the harvested areas while using the unharvested downstream areas as a reference (i.e., a before-after-control-impact design) using a general linear model. The model included species, catchments (aggregated into sub-watersheds due to spatial extents of bird home ranges), sampling area (harvested headwater areas versus unharvested areas) and harvest period (pre-harvest and post-harvest) as fixed-effects, as-well-as an interaction between harvest period and sampling area. We did not test for differences in blood THg concentrations among specific harvest treatments because pre-harvest sampling areas often included portions of multiple treatments and birds rarely captured in the clear-cut catchments.

3. Results

3.1. Hg in stream biota

Across all sites, dates, and feeding guilds, Hg concentrations (ng/g dw; MeHg in invertebrates, THg in salamanders) in individual stream biota samples ranged more than 500-fold (0.6–301.1). This range was driven largely by differences among foraging guilds, with geometric mean (±standard error) Hg concentrations across all watersheds, treatments, and dates generally increasing with trophic position from 1.7 ± 0.1 in shredders to 37.0 ± 0.8 in salamanders, the top predators at these sites. Least-squares mean mercury concentrations (accounting for differences due to catchment, treatment, and year) were lowest (1.8 ± 0.3) in shredders, followed by collectors (6.0 ± 0.5), then grazers (Heptageniidae; 7.1 ± 0.6), invertebrate predators (Perlidae; 22.4 ± 1.7), and salamanders (34.7 ± 2.6). Except for grazers and collectors, least-squares mean Hg concentrations differed among all feeding guilds (F_4,839 = 494.7, p < 0.001; Tukey HSDs < 0.05).

Mercury concentrations in aquatic biota differed among unharvested, buffer, and clear-cut harvest treatments after accounting for variation due to feeding guild, catchment and year (F_2,788 = 59.9, p < 0.001). The least-squares mean Hg concentration in the clear-cut treatment (14.1 ± 1.5) was 48% and 165% higher than those in reference (9.6 ± 0.7) and buffered (5.3 ± 0.5) catchments, respectively (Fig. 1A). However, the effect of harvest treatment differed among feeding guilds (treatment × guild interaction: F _8,839 = 6.2, p < 0.001). Specifically, differences among treatments were not significant in the shredder guild, though the mean mercury concentration in the clear-cut treatment was 50% higher than that of the reference treatment, and 110% higher than the mean for the buffered treatment catchments. In the other guilds sampled, the least-squares mean Hg concentrations were highest in the clear-cut treatment, followed by the reference, and then the buffer treatment (Fig. 1B). Guild mean Hg concentrations from the clear-cut treatment were 34%–84% higher than those from the reference treatment. In contrast, guild means from the buffered treatment were 29–62% lower than in reference catchments (Fig. 1B).

Fig. 1.

Least-squares mean mercury (methylmercury in invertebrates; total mercury in salamanders) concentrations (ng/g dry weight ± standard error) in stream biota A) and individual feeding guilds B) sampled from reference (i.e., no harvest operations; green bars; k = 3), buffered (i.e., clear-cut with an 8-m riparian buffer strip retained; orange bars; k = 3), and clear-cut (i.e., no riparian buffer; red bars; k = 4) catchments in the Trask Watershed Study Area, OR, USA. Least-squares means in A) account for the effects of taxa, experimental catchment, and collection year. Least-squares means in B) account for the effects of experimental catchment, and collection year. Letters indicate significant differences (p < 0.05, Tukey HSD) among treatments for each guild.

To determine whether differences in Hg bioaccumulation might be associated with changes in Hg availability among harvest treatments, we assessed the relationships between biotic Hg concentrations and concurrently collected aqueous THg-F, THg-P, and MeHg-F concentrations (ng/L). Mercury concentrations in predatory invertebrates and salamanders were not significantly correlated with aqueous THg or MeHg concentrations, or any other water chemistry parameter (Figs. S3 and S4, Table S1).

3.2. Hg biomagnification rates

To distinguish the roles of biomagnification rate and food web structure on Hg bioaccumulation following harvest, we first compared Hg trophic magnification slopes (TMSs) among treatments while accounting for catchment and year (Fig. 2). Mercury concentrations were positively correlated with δ¹⁵N in all treatments (F_1,306 = 513.2, p < 0.001), although TMSs differed among treatments (δ¹⁵N × treatment interaction: F_2,305 = 16.4, p < 0.001). Specifically, the clear-cut treatment had a lower TMS (0.17, F_1,305 = 5.1, p < 0.001) than either the reference (0.30) or buffered (0.33) treatments, which did not differ (F_1,304 = 1.1, p = 0.268; Fig. 2).

Fig. 2.

Trophic magnification slopes (TMS) for the relationship between mercury (methylmercury in invertebrates; size-adjusted total mercury in salamanders) concentrations (ng/g dry weight) and nitrogen stable isotopes (δ¹⁵N) in stream biota sampled from reference (i.e., no harvest operations; green points and line; n = 160, k = 3), buffered (i.e., clear-cut with an 8-m riparian buffer strip retained; orange points and line; n = 59, k = 3), and clear-cut (i.e., no riparian buffer; red points and line; n = 95, k = 4) catchments in the Trask Watershed Study Area, OR, USA. Models also include catchment and collection year as random effects.

3.3. Food web structure

We used Bayesian estimates of food web length (vertical trophic structure; indicated by δ¹⁵N) and basal resource diversity (horizontal food web structure; indicated by δ¹³C) in stream communities, as well as the isotopic niche size of individual feeding guilds, to compare the food web structure of reference, buffered, and clear-cut catchments. Bayesian estimates of food web length did not differ between treatments (Fig. 3A), suggesting that harvest operations did not change the length of the food webs. Conversely, basal resource diversity was higher in catchments from the buffer treatment than in those from the reference or clear-cut treatments (Fig. 3A). The isotopic niches (i.e., Bayesian standard ellipse area; SEA_B) of shredders, grazers, and salamanders did not differ in size among treatments (Fig. 3B). Collectors occupied a larger isotopic niche in the reference catchments compared to either the buffered or clear-cut treatment catchments, whereas the isotopic niche of predatory invertebrates was largest in the buffered catchments and did not differ between the reference or clear-cut treatments (Fig. 3B).

Fig. 3.

Bayesian estimates of A) food web length (δ¹⁵N range) and basal resource diversity (δ¹³C range), and B) Isotopic niche size (Bayesian standard ellipse area; SEA_B) for individual feeding guilds, from reference (unharvested; green plots), clear-cut with riparian buffer (orange plots), and clear-cut without a riparian buffer (red plots) streams in ten experimental watersheds of the Trask Watershed Study Area, OR, USA. Black dots represent the modes and the shaded boxes represent (from darker to lighter) the 50%, 75%, and 95% credible intervals of the Bayesian estimates. Plots with different letters have non-overlapping 95% credible intervals.

3.4. Riparian songbirds

We collected songbird blood both before and after harvest operations in both harvested and unharvested areas of the Trask Watershed Study Area. We collected 128 blood samples from nine species of songbirds (Table S2), three of which (Pacific-slope Flycatcher [Empidonax difficilis], Swainson’s Thrush [Catharus ustulatus], and Wilson’s Warbler [Cardellina pusilla]) comprised 83% of the total catch. Mercury concentrations in blood spanned a 30-fold range (17.3–523.9 ng/g ww) across all species, sites and dates, with a geometric mean (±standard error) of 132.1 ± 8.5 ng/g ww. Geometric mean THg concentrations of individual species ranged from 17.3 ± 0.01 to 243.5 ± 18.3 ng/g ww (Table S2).

Blood THg concentrations differed among songbird species (Tables S2 and S3) and sub-watersheds after accounting for the effects of collection area (headwater versus downstream areas) and sampling period (pre-versus post-harvest). Mercury concentrations in birds from experimental and downstream areas did not differ statistically; however, when accounting for sampling period, concentrations in birds from harvested areas were approximately 20% higher than in birds from unharvested areas (Table S3). Total Hg concentrations were nearly 50% higher in birds sampled after harvest operations than in birds sampled prior to harvest when accounting for the effects of harvest area (Fig. 4). The interaction between sampling area and period was not significant (Table S3) indicating that blood THg concentrations were similarly elevated following harvest in both headwaters (where harvest operations occurred) and in downstream areas, which were not disturbed during harvest operations.

Fig. 4.

Least-squares mean total mercury (THg) concentrations (ng/g wet weight ± standard error) in riparian songbird blood sampled from the Trask Watershed Study Area, OR, USA prior to, and following experimental timber harvests. Least-squares means account for the effects of species and capture area (i.e., within the experimentally harvested area versus in unharvested downstream areas).

4. Discussion

Employing a replicated control-impact study design, we examined the effects of timber harvest on Hg bioaccumulation in multiple feeding guilds of 10 headwater streams. In our analyses both across feeding guilds, as well as within most individual guilds, biotic Hg concentrations were highest in catchments that were fully clear-cut, intermediate in unharvested reference catchments, and lowest in catchments that were clear-cut except for an 8-m wide riparian buffer. These differences did not correspond to differences in either aqueous MeHg or THg in the experimental catchments. Rather, guild-level analyses and quantitative measures of food web structure suggest differences in the relative importance of terrestrial litter inputs compared to instream production play a role in determining the effects of timber harvest on Hg bio-accumulation in headwater streams.

We observed a 20-fold range in Hg concentrations among aquatic feeding guilds, consistent with the bioaccumulative properties of MeHg in stream environments (Jardine et al., 2012; Riva-Murray et al., 2013). Importantly, MeHg concentrations differed between the two primary consumer guilds, with least-squares mean concentrations in grazers more than 4-fold higher than those in shredders when accounting for catchment, treatment, and year. Similar differences in Hg concentrations between shredders and grazers have been reported across a variety of stream types and orders (Jardine et al., 2012; Mason et al., 2000; Riva-Murray et al., 2013; Tsui et al., 2009) and reflect lower MeHg concentrations in terrestrial plant litter, the basal resource upon which shredders are largely reliant, compared to instream biofilms and periphyton consumed by grazers (Jardine et al., 2012). Although we do not have direct measures of Hg concentrations in these basal resources from our study streams, the difference between periphyton and litter MeHg concentrations are well established in the literature, with a 31-fold higher median concentration reported for periphyton than for terrestrial litter (Table S4). Further, our data are in agreement with an extensive body of literature showing 1.5- to 3-fold lower MeHg concentrations in invertebrates relying on terrestrial litter compared to those utilizing organic matter produced within streams (Jardine et al., 2012; Obrist et al., 2016; Riva-Murray et al., 2013; Tsui et al., 2009). The incorporation of MeHg at the base of the food web is a primary determinant of Hg bioaccumulation in aquatic food webs (Chételat et al., 2011; Ward et al., 2010), and even small shifts in reliance on instream-produced organic matter (as opposed to terrestrially derived organic matter), similar to those observed in the current study, have been associated with elevated Hg concentrations in upper trophic level consumers of streams (Jardine et al., 2012; Riva-Murray et al., 2013).

Timber harvest affects both the type and amount of basal resources available to stream food webs, with clear-cut operations typically reducing terrestrial litter inputs and promoting instream primary production (Bilby and Bisson, 1992; Kiffney and Richardson, 2010). Consistent with this change, we encountered the shredder guild (specifically Pteronarcys spp.) much less frequently in the clear-cut catchments than in either the reference or buffered catchments. Indeed, across three years of extensive sampling with similar effort across catchments, we collected only six individual Pteronarcys spp. from clear-cut catchments compared to 55 individuals from the reference catchments. Although we lack quantitative data on guild abundance, these data suggest that the shredder niche (consumption of terrestrial organic matter) may be diminished in clear-cuts, as is a commonly-reported consequence of this such activities (Bilby and Bisson, 1992; Kiffney and Richardson, 2010; Kreutzweiser et al., 2008a). Such a reduction in the shredder niche is also consistent with the lower (though not statistically significant) basal resource diversity we observed in clear-cut compared to reference treatment (Fig. 3A).

The scarcity of our target shredder taxon in the clear-cut catchments also underlies our finding that the Hg biomagnification rate (i.e., trophic magnification slope) was significantly lower in the clear-cut treatment. Shredders had the lowest MeHg concentrations and δ¹⁵N of any guild, and thus exerted substantial leverage on the TMS. Poor representation of the shredder taxon (n = 3) in the TMS calculation for the clear-cut treatment substantially reduced that leverage, contributing to the lower overall slope. This is apparent in the similarity of TMSs among the three treatments when calculated using guild mean Hg concentrations and δ¹⁵N (reference TMS = 0.27, buffered TMS = 0.32, clear-cut TMS = 0.26; data not shown) or when excluding shredder data from all three treatments (reference TMS = 0.18, buffered TMS = 0.20, clear-cut TMS = 0.16; data not shown), suggesting the diminished importance of terrestrial litter in the stream food webs was a more likely driver of differences in TMSs than changes in Hg biomagnification rates. This may also help reconcile our results with a previous study in which biomagnification rates did not differ between harvested and unharvested catchments in Norway (de Wit et al., 2014). In that study, terrestrial leaf litter was rare in the streams because the catchments had very few deciduous trees, and the primary consumers sampled were both generalist herbivores foraging predominantly on instream biofilms (de Wit et al., 2014). The TMSs observed in the current study were all within the range reported for MeHg in rivers and streams, and similar to the global mean TMS of 0.27 ± 0.08 reported for rivers and streams by Lavoie et al. (2013).

Our data suggest that reduced terrestrial litter inputs likely contributed to the elevated Hg concentrations we observed in the food webs of clear-cut catchments, as consumers would be subsequently more reliant on instream primary production associated with higher MeHg concentrations. However, Hg bioaccumulation in the buffered treatment catchments appeared to be influenced by different processes. In contrast to the clear-cut treatment, we found that both across and within guilds, Hg concentrations were lower in the buffered treatment than in the reference treatment (Fig. 1). Coincident with these differences in Hg concentrations, basal resource diversity was substantially higher in the buffered treatment, indicative of differences in the food web structure of streams with riparian buffers (Fig. 3A). We did not observe any differences in the niche size of either shredding or grazing primary consumers (Fig. 3B) in buffered catchments. Therefore, the increased basal resource diversity in catchments with riparian buffers likely reflects utilization of both terrestrially derived and instream produced resources rather than a shift in the isotopic characteristics of either resource. In recent studies by Erdozain et al. (Erdozain et al., 2018, 2019) upland harvest intensity was positively correlated with both biofilm production and reliance on terrestrial resources in 12 streams with 30 m wide riparian buffers, despite limited effects on riparian canopy cover and direct litter inputs to the streams. These results support those of previous studies that have shown riparian buffers effectively mitigate clear-cut associated reductions in terrestrial litter inputs (Kiffney and Richardson, 2010; Kreutzweiser et al., 2010) and associated changes to macroinvertebrate assemblages (Kiffney et al., 2003; Newbold et al., 1980; Richardson and Danehy, 2007), while also stimulating instream productivity by increasing light and nutrient availability (Gregory et al., 1987; Kiffney et al., 2004; Kreutzweiser et al., 2008b).

Riparian buffers can also result in increased quantity and quality of litter inputs when the removal of overstory conifers facilitates increased abundance of deciduous vegetation (Hoover et al., 2011; Zhang et al., 2009). In our study catchments, riparian vegetation was dominated by red alder (Alnus rubra), a deciduous tree that produces large quantities of high quality litter (Hart et al., 2013; Richardson et al., 2005; Richardson et al., 2004), and is stimulated by the removal of overstory conifers (Richardson et al., 2005). Invertebrate richness, density, and biomass are higher in streams with riparian alder stands, particularly following harvest of upland conifers (Cole et al., 2003; Hernandez et al., 2005), and these effects can persist for decades following harvest (Frady et al., 2007). Importantly, similar increases in abundance and richness have been observed in both shredder and grazing guilds from alder influenced streams, suggesting that both basal resource pathways are stimulated (Cole et al., 2003; Hernandez et al., 2005). Alder abundance in our study catchments differed dramatically among treatments immediately following harvest operations. Catchments with riparian buffers retained a larger proportion of their pre-harvest alder cover compared to clear-cut catchments (mean reduction in alder basal area = 63% and 91%, in buffered and clear-cut catchments, respectively), despite similar reductions in the basal area of non-alder species between the two treatments (84% and 87%, respectively). These differences indicate that alders in the buffered catchments were concentrated in the riparian zone and suggest terrestrial inputs to streams in buffered catchments were likely to be maintained to a much higher degree than in clear-cut catchments (Kiffney et al., 2003, 2004; Newbold et al., 1980).

These differences in basal resource availability may provide a common mechanism explaining observed differences in Hg bio-accumulation across harvest treatments. We evaluated the relationship between Hg concentrations and basal resource diversity by comparing mean Hg concentrations in biota from individual catchments with estimates of basal resources for those catchments (this analysis was limited to those catchments with enough data to calculate robust estimates of each metric). We found that Hg concentrations in biota increased as basal resource diversity decreased across all treatments and catchments (Fig. 5). It is widely recognized that Hg bioaccumulation can vary among consumers relying on different basal resources (Jardine et al., 2012; Karimi et al., 2016; Riva-Murray et al., 2013) and our results suggest that the diversity of basal resources incorporated into food webs may independently influence Hg bioaccumulation. Food webs with higher basal resource diversity often have diffuse energy flows resulting from dilution of productivity across many relatively weak trophic interactions compared to channelizing energy flow through a few strong trophic interactions as is associated with simpler food webs (McCann et al., 1998; Thompson et al., 2012). Since MeHg largely follows energetic pathways through food webs, it is reasonable to expect an analogous process for MeHg bioaccumulation, though the exact mechanisms are not yet known.

Fig. 5.

Relationship between least-squares mean mercury (methylmercury in invertebrates; total mercury in salamanders) concentrations (ng/g dry weight ± standard error) and basal resource diversity (δ¹³C range) in stream biota sampled from six catchments in the Trask Watershed Study Area, OR, USA. Harvest treatments within catchments include: reference (i.e., no harvest operations; green points), buffered (i.e., clear-cut with an 8-m riparian buffer strip retained orange point), and clear-cut (i.e., no riparian buffer; red points). Least-squares means account for the effects of taxa and collection year. Dashed lines represent the 95% confidence interval around the regression line.

Biotic Hg concentrations were not significantly correlated with aqueous THg-F, THg-P or MeHg-F, despite higher THg concentrations and loads in the clear-cut catchments compared to reference catchments (Fig. S3; Eckley et al., 2018). Mercury bioaccumulation is often decoupled from aqueous THg concentrations (Eagles-Smith et al., 2016a; Tsui et al., 2009) because of the biogeochemical processes that regulate the production and availability of MeHg (Suchanek et al., 2008; Wiener et al., 1990). For these reasons, aqueous MeHg is more often correlated with concentrations in biota (Chasar et al., 2009; Riva-Murray et al., 2011). Although not statistically significant, MeHg concentrations in predatory invertebrates tended to increase with increasing aqueous MeHg concentrations (Fig. S3), suggesting a possible role of aqueous MeHg concentrations on MeHg bioaccumulation in these streams. We did not observe a similar relationship in salamanders, suggesting a weaker connection between aquatic MeHg availability and salamander MeHg concentrations, as would be expected considering the potential for salamander diets to be subsidized with high energy, low MeHg terrestrial prey (Parker, 1994; Ward et al., 2012). Aqueous MeHg could be more closely coupled with concentrations in primary consumers, however, we lacked enough overlap between water samples and either shredder or grazer samples to test this. Further, in other streams the relationship between aqueous and biotic MeHg concentrations is not limited to primary consumers (Chasar et al., 2009; Riva-Murray et al., 2011). Rather, the lack of relationships between aqueous and biotic Hg concentrations likely reflects the uniformly low aqueous MeHg concentrations in these streams and our small sample sizes for co-collected aqueous and biotic Hg data. Although we used uncensored values in our analyses, 97% of aqueous MeHg measurements were below our reporting limit (0.05 ng/L) and concentrations spanned a narrow range (<3-fold) compared with studies demonstrating correlations between aqueous and tissue MeHg concentrations (5–14 fold; Eckley et al., 2018).

The influence of forest harvest on Hg bioaccumulation may not be limited to the streams flowing through harvested catchments. Emergent aquatic insects provide important subsidies to riparian food webs (Baxter et al., 2005), and serve as a conduit for the entry of aquatic contaminants into riparian food webs (Becker et al., 2017; Cristol et al., 2008; Kraus et al., 2014). Forest harvest often increases net aquatic insect emergence (Banks et al., 2007; Moldenke and Ver Linden, 2007), and we found MeHg concentrations were higher in larval insects from harvested catchments. Thus, MeHg flux to riparian areas may have also increased following harvest, potentially explaining the higher blood THg concentrations in riparian song birds post-harvest (Fig. 4). Indeed, we observed the greatest differences in pre-versus post-harvest blood THg concentrations in species that are typically more reliant on aquatic resources (e.g., Pacific-slope Flycatchers: 70% increase; Wilson’s Warblers: 40% increase), whereas, there were only minor differences in species that utilize fewer aquatic resources (e.g., Swainson’s Thrush: 5% decrease; Rodewald, 2015). However, THg concentrations in bird blood were higher post-harvest in both harvested headwaters and unharvested downstream areas, possibly indicating that differences in songbird Hg exposure were not due to harvest practices, but reflective of an unrelated driver. Although we accounted for differences among species in our analyses, it is possible that differences in habitat use among species contributed to this result. There is strong species-level and individual-level variation in the sources of mercury exposure in songbirds, much of which are likely related to habitat and diet preferences as well as variation in individual home range (Tsui et al., 2018). Pacific-slope Flycatchers, which had one of the highest mean THg concentrations and greatest pre-to post-harvest changes in THg, were rarely captured in harvested areas during the post-harvest period but were common in adjacent unharvested areas. Given the substantial increase in the THg concentrations of these birds, it is likely they were influenced by harvest operations despite being captured outside of the harvested catchments. This result suggests that songbirds captured in unharvested areas were utilizing harvested habitats, which is likely given the close proximity (50–500 m) of harvested and unharvested sampling areas and the size of home ranges typical of songbirds (King and DeGraaf, 2000; Rodewald, 2015). Alternatively, these results may suggest aquatic subsidies of Hg into terrestrial habitats can extend through harvested areas into adjacent unharvested areas.

It is important to note that our analysis was largely constrained to a control-impact design (as opposed to a before-after control-impact [BACI] design). Thus, it should be stressed that our results represent differences among catchments subject to differing harvest treatments rather than effects of the harvest treatments on each catchment, per se. As such, it is possible that our results reflect natural variation among catchments rather than treatment effects, however, we contend that this is relatively unlikely given our replication of each treatment across several experimental catchments. Indeed, given our relatively robust replication, the primary limitation of our study design is likely to be a reduction in our ability to detect true differences among treatments (Osenberg et al., 2006; Smokorowski and Randall, 2017).

Our data indicate that Hg concentrations in stream and riparian biota are influenced by forest harvest practices. Although forestry operations have been shown to influence THg and MeHg export (Bishop et al., 2009; Eckley et al., 2018; Eklöf et al., 2018; Eklöf et al., 2016; Skyllberg et al., 2009), from harvested catchments, our data suggest that Hg bioaccumulation in these food webs is not simply a function of aqueous THg or MeHg concentrations. Rather, multiple lines of evidence suggest that differences in basal resource availability and food web structure, potentially modulated by differences in the productivity of streams following forest harvest, may account for the observed differences in biotic Hg concentrations among treatments. Regardless, our data highlight the importance of understanding food web responses to landscape disturbances when assessing Hg bioaccumulation. Importantly, MeHg concentrations in our study catchments were low in comparison to similar streams in many other areas and Hg exposure overall was unlikely to pose direct health risks to biota. However, given the global scope of timber harvest, a 50–80% increase in Hg bioaccumulation of top predators in stream food webs due to clear cutting, as observed in our study, could substantially increase ecological risk to Hg exposure in areas with higher pre-harvest Hg concentrations, areas that are more prone to Hg methylation, or are more sensitive to changes in food web structure. Thus, our findings that different harvest operation techniques (i.e. riparian buffers versus clear-cuts) may mitigate potential increases in Hg bioaccumulation associated with harvest offer potential avenues for management consideration and modifying best management practices.