Differential reliance on aquatic prey subsidies influences mercury exposure in riparian arachnids and songbirds

Abstract Cross‐ecosystem subsidies move substantial amounts of nutrients between ecosystems. Emergent aquatic insects are a particularly important prey source for riparian songbirds but may also move aquatic contaminants, such as mercury (Hg), to riparian food webs. While many studies focus on species that eat primarily emergent aquatic insects, we instead study riparian songbirds with flexible foraging strategies, exploiting both aquatic and terrestrial prey sources. The goal in this study is to trace reliance on aquatic prey sources and correlate it to Hg concentrations in common riparian arachnids (Families Tetragnathidae, Opiliones, and Salticidae) and songbirds (Common Yellowthroat Geothlypis trichas, Spotted Towhee Pipilo maculatus, Swainson's Thrush Catharus ustulatus, Song Sparrow Melospiza melodia, and Yellow Warbler Setophaga petechia). We used stable isotopes of δ13C and δ15N and Bayesian mixing models in MixSIAR to determine the reliance of riparian predators on aquatic prey sources. Using mixed effects models, we found that arachnid families varied in their reliance on aquatic prey sources. While songbird species varied in their reliance on aquatic prey sources, songbirds sampled earlier in the season consistently relied more on aquatic prey sources than those sampled later in the season. For both arachnids and songbirds, we found a positive correlation between the amount of the aquatic prey source in their diet and their Hg concentrations. While the seasonal pulse of aquatic prey to terrestrial ecosystems is an important source of nutrients to riparian species, our results show that aquatic prey sources are linked with higher Hg exposure. For songbirds, reliance on aquatic prey sources early in the breeding season (and subsequent higher Hg exposure) coincides with timing of egg laying and development, both of which may be impacted by Hg exposure.

. In many temperate climates, terrestrial ecosystems have seasonal shifts in insect prey availability with emergent insect biomass peaking in spring, followed later by terrestrial prey after leaf-out (Nakano & Murakami, 2001).
The flux of aquatic prey is an important nutrient source, as emergent insects have higher polyunsaturated fatty acids (PUFA) than terrestrial prey (Martin-Creuzburg et al., 2017;Moyo, 2020). Both terrestrial predatory invertebrates (e.g., arachnids) and songbirds often concentrate near aquatic habitats ostensibly to exploit the emergent insect subsidy during peak emergence times (Hagar et al., 2012;Uesugi & Murakami, 2007).
Despite the benefits of aquatic insect emergence as energetic subsidies to riparian communities, they can also degrade riparian habitats through the export of aquatically derived environmental contaminants. Many aquatic ecosystems are accumulation zones for environmental contaminants, such as mercury (Hg), other heavy metals, and polychlorinated biphenyls (PCBs). When assimilated by aquatic invertebrates, these contaminants can accompany their movement into surrounding terrestrial food webs (Jones et al., 2013;Kraus et al., 2014;Latta et al., 2015;Walters et al., 2008). As a result, riparian taxa ranging from terrestrial invertebrates to invertebrateeating birds and bats have been shown to accumulate aquaticsourced contaminants (Becker et al., 2018;Moy et al., 2016;Yates et al., 2014).
A number of studies have assessed the influence of aquatic factors on the magnitude of emergent aquatic insect and aquatic-derived contaminant flux into the terrestrial ecosystem (Kelly et al., 2019;Walters et al., 2008). Contaminant concentrations in aquatic insects are influenced by aquatic habitat (Jackson et al., 2019), and the biomass of insects that survive to emerge from the aquatic system can be mediated by habitat, water quality, and fish abundance (Jones et al., 2013;Paetzold et al., 2011). For some contaminants, high contaminant loading can reduce invertebrate fecundity and survival, ultimately constraining insect biomass flux and contaminant transfer to riparian food webs (Kraus, 2019;Kraus et al., 2014;Paetzold et al., 2011). Other contaminants, such as Hg, are not known to affect aquatic insect survival in Hg contaminated areas; therefore, high insect emergence rates can move large amounts of Hg into riparian zones (Tweedy et al., 2013). Mercury is of particular concern for songbirds as it continues to be found in the environment at high concentrations across North America (Cristol & Evers, 2020), and Hg exposure negatively impacts many critical aspects of the songbird life cycle, including reproduction Jackson, Evers, Etterson, et al., 2011) and migration (Seewagen, 2018).
While we understand many of the aquatic factors that influence emergence of aquatic insects, less is known about how terrestrialbased factors may play a role in riparian predator Hg exposure, especially across bird taxa. Many Hg studies focus on species, such as aerial insectivores, that feed directly and almost exclusively on emergent aquatic prey (Alberts et al., 2013;Custer et al., 2007). Other riparian songbirds have been shown to accumulate high levels of aquatic-based contaminants but have much more varied exposure patterns (Jackson et al., 2015). While some variation can be explained by broad foraging guild classifications (e.g., granivore, omnivore, insectivore), there is still a large amount of unexplained variation both among species and individuals (Jackson et al., 2015;. Within species classified as insectivorous, it is assumed that individuals eat both emergent aquatic insects but also terrestrial-based prey with little connection to the aquatic ecosystem and contaminants. Additionally, many riparian songbirds eat a variety of terrestrial invertebrate predators, such as spiders, and those invertebrate predators also can eat a varied diet consisting of both aquatic and terrestrial prey Kelly et al., 2019;Speir et al., 2014).
We hypothesize that some of the variation in Hg exposure among insectivorous predators (arachnids and songbirds) can be explained by species and individual-level differences in prey selection, with individuals or species that rely more on aquatic prey having higher Hg exposure. Our goal in this study is to trace reliance on aquatic-based prey and correlate it to Hg concentrations in common riparian arachnids (Families Tetragnathidae, Opiliones, and Salticidae) and songbirds (Common Yellowthroat Geothlypis trichas, Spotted Towhee

Pipilo maculatus, Swainson's Thrush Catharus ustulatus, Song Sparrow
Melospiza melodia., and Yellow Warbler Setophaga petechia). We use δ 13 C and δ 15 N stable isotopes to differentiate between aquatic and terrestrial signatures, which has been used in other studies with success (Walters et al., 2008). To answer this question using stable isotopes, we must 1) determine differences in stable isotope signatures of aquatic and terrestrial invertebrate food webs, 2) quantify differences in MeHg between aquatic and terrestrial invertebrate prey, 3) use δ 13 C and δ 15 N stable isotopes in a Bayesian mixing model determine terrestrial predator (arachnids and songbirds) reliance on the aquatic prey source, 4) evaluate seasonal changes and speciesspecific differences in the proportion of aquatic prey in the diet of terrestrial predators, and 5) correlate seasonal changes in aquatic prey reliance to their Hg exposure for both arachnids and songbirds.

| Fieldwork
Field sites were chosen in riparian forest sites along the Willamette River in western Oregon (Figure 1). The Willamette River, Oregon, USA, is a major tributary to the Columbia River and drains the eastern Coast Range and western Cascades. It also has a history of anthropogenic influences, including Hg contamination (Henny et al., 2005;Hope, 2006;Hope & Rubin, 2005). The Willamette River supports a diversity of subhabitats including backwater alcoves and open channel flowing waters that differ in Hg concentrations (Jackson et al., 2019). To best differentiate aquatic and terrestrial isotope signals, we focused here on the main channel environments.
From 1 May 2013 to 23 July 2013, we sampled 7 main stem sites along the Willamette River ( Figure 1). Since we are focused on birds that are not necessarily riparian obligates, we sampled songbirds living within the riparian forest using targeted mist-netting to focus on individuals with territories within 150m of the river, the general riparian area defined in other studies (Nakano & Murakami, 2001;Walters et al., 2008). Singing individuals were identified, and mist nets (6m or 12m length, 30mm mesh) were placed opportunistically (where habitat allowed) within their territories. Playback recordings of conspecific songs were used to attract and capture riparian songbirds in a mist net. Although numerous bird species were captured and sampled, the majority (94%) of samples were of five species present at all sites: Common Yellowthroat (Geothlypis trichas), Spotted Towhee (Pipilo maculatus), Swainson's Thrush (Catharus ustulatus), Song Sparrow (Melospiza melodia), and Yellow Warbler (Setophaga petechia). All birds were banded with an aluminum USGS band, and any recaptures were excluded from the analyses to preserve independence of samples. Blood samples of each bird were taken from the brachial ulnar vein, using 27-gauge needles (BD PrecisionGlide, Fisher Scientific) and heparinized microhematocrit capillary tubes (Fisherbrand, Fisher Scientific). Samples were capped with Critocaps™ (Leica Microsystems) and stored on ice in the field until they could be transferred to a freezer (within 6 hr of sampling).
No more than 1% of bird's body weight of blood was collected from each individual, usually between 20 µl and 100µl.
We recorded presence of brood patch (for females) or cloacal protuberance (for males) and limited samples to individuals in breeding condition to avoid late or early migrants. We also aged each bird (Pyle, 1997)  We also collected aquatic and terrestrial invertebrates from sites that were collocated in space and time with the songbird sampling (Table 1). We sampled all invertebrates encountered but targeted subsequent laboratory analysis on the numerically dominant invertebrate families at each site. We did not estimate invertebrate biomass each site because our goal was focused on tracking energetic signals and not taxonomic abundance on the landscape. Aquatic invertebrates were collected via kick net and dip net in the aquatic habitat near where mist nets had been set. Terrestrial invertebrates were collected by beat sheet and sweep net in forest or shrub habitat near mist nets. All invertebrates were composited by site, transferred to glass scintillation vials, and kept on ice in the field until they could be transferred to a freezer (within 6 hr of sampling).

| Invertebrate laboratory analyses
Aquatic invertebrates were identified to family (Merritt et al., 2008), and terrestrial invertebrates were identified to either order (i.e., Hemiptera, Coleoptera, etc.) or family for arachnids (i.e., Tetragnathidae, Opiliones, Salticidae), etc. All invertebrates were composited based on the lowest taxon (order or family) identified per site and sampling date (mean, SD, F I G U R E 1 Study area along the Willamette River in western Oregon. Sites (shown as circles) had similar gallery forest terrestrial habitat paired with main channel aquatic habitat. Major cities along the Willamette are represented with stars TA B L E 1 Number of samples collected at each site over different sampling date ranges. minimum, and maximum number of individuals in the composite samples can be found in Table 2). Once composited, invertebrates were rinsed with deionized water, placed in glass vials, and dried in an oven at 50°C for a minimum of 48 hr hrs. Once dried, they were homogenized in their drying vials into a fine powder with a clean glass rod.

| Songbird blood analysis
Songbird whole blood samples were not composited, but instead run on an individual basis as dry weight. Because 95%-99% of Hg in bird blood is MeHg (Rimmer et al., 2005), we analyzed bird blood for total

| Statistical analysis
All statistical modeling was conducted with Program R (Version 4.0.3 "Bunny-Wunnies Freak Out," R Foundation for Statistical Computing).
We grouped invertebrates into broad categories based on diet and life history. Aquatic-collected samples were grouped into aquatic (fully aquatic life stage, no emergent life stage; e.g., snails), emergent (larval emergent aquatic insects with herbivorous or omnivorous feeding habits; e.g., Trichoptera), and emergent predators (larval emergent aquatic insects with entirely predatory feeding habits; e.g., Odonata).
Terrestrially collected samples were grouped into terrestrial (primarily herbivorous feeding habit with no aquatic life stage; e.g., Hemiptera), terrestrial-emergent (adult stage of emergent aquatic insects, those that do not feed as adults; e.g., Ephemeroptera), terrestrial-emergent predators (adult life stage of emergent aquatic insects, those that are predators as adults; e.g., Zygoptera), and terrestrial-mixed (terrestrial insect orders that have both aquatic and terrestrial larval stages; e.g., Diptera). The terrestrial-mixed category is necessary because we only identified terrestrial insects to order, and some of the orders have mixed life history strategies (Table 3). Arachnids and songbirds were not included in these groups, as they were the consumers of interest for the stable isotope models to follow.

| Prey sources
To determine suitable stable isotope endmembers for our analysis, we first explored differences in δ 13 C and δ 15 N between our invertebrate groups. We used a one-way analysis of variance (ANOVA) followed by Tukey's HSD test to quantify differences in both δ 13 C and δ 15 N among the invertebrate groups. We also compared MeHg concentrations among the invertebrate groups (one-way ANOVA on log-transformed MeHg concentrations).
Based on these analyses, we designated the "emergent" insect group as being representative of the aquatic prey source and the TA B L E 3 Summary statistics stable isotopes and MeHg for invertebrate taxa and groups. Different letters moving down a column indicate statistically significant differences between groups in δ 13 C or δ 15 N "terrestrial" insect group as representative of the terrestrial prey source (groups explained above and in Table 3). When we refer to either aquatic or terrestrial prey sources, we are strictly indicating only the stable isotope signal came from those different endmembers, but we do not necessarily know the exact prey the consumers ate. Thus, the stable isotopes results would only be distinguishing if the arachnids or songbirds were receiving energy that originated in those respective habitats and would not inform that they were directly consuming emergent aquatic insects. No TDF exists for the species sampled in this study, but we also ran our MixSIAR results using TDF for songbirds used in other studies (Michelson et al., 2018) and found no difference in results.

| MixSIAR Bayesian models
The arachnid model included individual sample composite as a random effect, uninformative priors, a residual*process error structure and used the "long" run length to achieve model convergence (chain length = 300,000, burn-in = 200,000, thin = 100, # chains = 3). The songbird model included individual bird as a random effect, uninformative priors, a residual*process error structure and used the "very long" run length to achieve model convergence (chain length =1,000,000, burn-in = 500,000, thin = 500, # chains = 3). We assessed model convergence using Gelman-Rubin and Geweke diagnostics.

| Mixed effects models
We used the mixSIAR-calculated proportion aquatic prey source for both arachnids and songbirds in all subsequent analysis. We developed mixed effects models using packages lme4 (Bates et al., 2019)

| Analysis of prey sources
We analyzed a total of 214 invertebrate samples for δ 13 C, δ 15 N, and MeHg and 137 songbird individual samples for δ 13 C, δ 15 N, and THg (Table 1). We first compared stable isotope signatures of the invertebrate groups in both aquatic and terrestrial food webs (Figure 2a).
Invertebrates sampled from the terrestrial environment were generally depleted in both δ 13 C and δ 15 N compared to invertebrates sampled from the aquatic environment. One notable exception was emergent insects caught as terrestrial nonfeeding adults (mayflies and stoneflies); their isotope signatures were similar to the aquatic environment. Overall and taxa-specific means and SD for δ 13 C, δ 15 N, and geometric means and back-transformed SE for MeHg can be found in Table 3. We found a significant difference in both δ 13 C (one-way ANOVA, F = 51.3, p <.001) and δ 15 N (one-way ANOVA, F = 56.0, p < .001) between invertebrate groups. Tukey pairwise analysis revealed consistent differences (p <.01) between terrestrial insects and aquatic, emergent aquatic, and emergent aquatic predators in both δ 13 C and δ 15 N (results from all pairwise comparisons are included in Table 3).
Pairwise comparisons (Tukey HSD, p <.001) indicated that terrestrial insects were significantly lower in MeHg than all other groups (including both aquatic invertebrates and emergent insects caught in both their aquatic and terrestrial life stages). Although there was no difference in δ 13 C or δ 15 N between aquatic invertebrates and emergent aquatic invertebrates, there was a significant difference in MeHg between these groups.

| MixSIAR Bayesian isotope mixing models
We used emergent aquatic insects as our aquatic prey source (δ 13 C =

| Arachnids
Using the results for proportion aquatic prey as the dependent variable in a mixed effects model with site as a random effect, we found that proportion aquatic prey differed among arachnid groups (F = 22.5, p <.001, Figure 4a) but not over time (F = 1.15, p =.29). We F I G U R E 2 (a) Stable isotopes of carbon-13 and nitrogen-15 in invertebrate groups, based on sampling location (aquatic=circles or terrestrial =triangles). Biplots indicate mean and SD. Samples that make up each mean and SD can be found in Table 1. (b) MeHg differences between invertebrate groups. Different letters indicate statistically significant differences (p <.05). Aquatic-collected samples were grouped into aquatic (aq; fully aquatic life stage, no emergent life stage, N = 48), emergent (em; larval emergent aquatic insects with herbivorous or omnivorous feeding habits, N = 36), and emergent predators (em-pred; larval emergent aquatic insects with entirely predatory feeding habits, N = 14). Terrestrially collected samples were grouped into terrestrial (terr; primarily herbivorous feeding habit with no aquatic life stage, N = 45), terrestrialemergent (terr-em; adult stage of emergent aquatic insects, those that do not feed as adults, N = 11), terrestrial-emergent predators (terr-em-pred; adult life stage of emergent aquatic insects, those that are predators as adults, N = 5-all Coenagrionidae), and terrestrial-mixed (terr-mix; terrestrial insect orders that have both aquatic and terrestrial larval stages, N = 25) next used MeHg concentration as the dependent variable and found that MeHg concentrations also differed among arachnid groups (Opiliones, Salticidae, Tetragnathidae, arachnid composite; F = 6.85, p =.002) and was positively related to proportion aquatic prey (F = 25.71, p <.0001). However, the interaction between proportion of aquatic prey and arachnid family (F = 3.51, p =.035) indicated that the slopes for the relationship between proportion of aquatic prey and MeHg concentrations varied among taxa (Figure 4b).

| Terrestrial songbird predators
We ran similar models for riparian songbirds (Common Yellowthroat,

| D ISCUSS I ON
Aquatic productivity provides important energetic subsidies to surrounding terrestrial communities (Baxter et al., 2005;Nakano & Murakami, 2001), and our findings are consistent with this body of Hg are generally higher in aquatic than terrestrial environments; thus, these subsidies may represent substantial vectors of aquatic contaminants into terrestrial communities (Kraus et al., 2014;Moyo, 2020;Walters et al., 2008).
Seasonal pulses of emergent aquatic invertebrates into terrestrial ecosystems have been identified in many different habitats (Ballinger & Lake, 2006;Bartrons et al., 2015;Baxter et al., 2005). Seasonal weather shifts in the Pacific Northwest, from wet spring through dry summer months, which span the songbird breeding season, can influence terrestrial invertebrate abundance as well as emergence pulses of aquatic invertebrates (Nakano & Murakami, 2001). Using isotopic signatures, we found that songbirds shifted from being largely reliant on aquatic-sourced prey to a greater reliance on terrestrially sourced prey. Songbird reliance on aquatic-sourced prey declined throughout the season, presumably due to the later emergence of terrestrial invertebrates following the leafing out of deciduous vegetation. Our findings are supported by previous studies that have used field observations to determine reliance on terrestrial versus aquatic prey in riparian songbirds (Nakano & Murakami, 2001;Uesugi & Murakami, 2007).
For arachnids, MeHg concentrations were also correlated with aquatic prey reliance but the relationship between reliance on aquatic prey and MeHg exposure varied among arachnid families.  Walters et al., 2008Walters et al., , 2010  . Contrary to these benefits, our study also shows that reliance on aquatic prey increases Hg exposure, which can interfere with migration (Seewagen, 2018), reproduction (Jackson, Evers, Etterson, et al., 2011;Varian-Ramos et al., 2014), and survival (Ma et al., 2018) at environmentally relevant exposures. The varying reliance on aquatic prey among species and individuals complicates the calculation of Hg risk to riparian communities near contaminated water bodies because a more detailed understanding of foraging ecology (beyond broad classification of granivore, omnivore, or insectivore) is needed to assess risk. Despite that all of the songbirds in this study are reported to be insectivorous during the breeding season, they still differed in the relative contributions of aquatic and terrestrial sourced prey, which directly influenced their Hg exposure.
Birds and arachnids that rely more on aquatic prey have higher Hg exposure, which follows findings of others who have shown that emergent aquatic insects are an important source of aquatic contaminants to terrestrial ecosystems (Chumchal & Drenner, 2015;Speir et al., 2014;Walters et al., 2008Walters et al., , 2010. Very few studies have quantified individual or species-specific reliance on aquatic prey outside of aerial insectivores, which are known to focus almost entirely on emergent insects (Alberts et al., 2013;Custer et al., 2008). Our expanded effort on other forest riparian songbirds indicated that species with more flexible foraging strategies demonstrated more plastic reliance on prey source over time.
This study provides an example of how individual-level factors in species foraging ecology influence variation in mercury exposure.
Moreover, we did not find evidence that proximity of nesting territory (inferred from capture location) to water was the sole determinant of aquatic energy to riparian songbird diet.
The Hg levels we measured in songbirds were below general thresholds thought to cause reproductive harm (Jackson, Evers, Etterson, et al., 2011;Varian-Ramos et al., 2014, p.). It is important to understand, however, that these thresholds are developed for a limited number of species, none of which were sampled in this project. It is likely that species and individuals vary in their sensitivity to Hg, and so taxa-wide threshold levels should be used with caution (Varian-Ramos et al., 2014, p.). While the Hg levels are relatively low, our findings related to habitat, species, and season may apply to other study areas with higher Hg loading and so reinforce the importance of studying interactions between behavior, season, and habitat.

| CON CLUS IONS
We used stable isotopes of carbon and nitrogen to determine how the riparian forest songbird communities relied on aquatic energy subsidies. These species (Common Yellowthroat, Spotted Towhee, Swainson's Thrush, Song Sparrow, and Yellow Warbler) or closely related conspecifics are widespread throughout North American riparian areas and represent an understudied community in aquatic contaminant studies. The species we studied varied in their reliance on aquatic prey and subsequent Hg exposure at both a species and individual level. We were not only able to correlate Hg concentrations in songbirds to their reliance on aquatic-based prey, but also showed that birds forage on more aquatic-sourced prey early in the season than later. These findings suggest that pulsed emergence of aquatic invertebrates may be an important vector of Hg to avian insectivores. These findings are particularly relevant in the face of climate change, which can alter the timing and magnitude of emergent aquatic subsidies (Larsen et al., 2016).