Journal list menu

Volume 7, Issue 8 e01422
Open Access

Seasonal variation exceeds effects of salmon carcass additions on benthic food webs in the Elwha River

S. A. Morley

Corresponding Author

S. A. Morley

Fish Ecology Division, Northwest Fisheries Science Center, National Marine Fisheries Service, NOAA, Seattle, Washington, 98112 USA

E-mail:[email protected]Search for more papers by this author
H. J. Coe

H. J. Coe

Ocean Associates, Arlington, Virginia, 22207 USA

Search for more papers by this author
J. J. Duda

J. J. Duda

Western Fisheries Research Center, U.S. Geological Survey, Seattle, Washington, 98115 USA

Search for more papers by this author
L. S. Dunphy

L. S. Dunphy

School of Aquatic and Fishery Sciences, University of Washington, Seattle, Washington, 98105 USA

Search for more papers by this author
M. L. McHenry

M. L. McHenry

Natural Resources Department, Lower Elwha Klallam Tribe, Port Angeles, Washington, 98363 USA

Search for more papers by this author
B. R. Beckman

B. R. Beckman

Environmental and Fisheries Sciences Division, Northwest Fisheries Science Center, National Marine Fisheries Service, NOAA, Seattle, Washington, 98112 USA

Search for more papers by this author
M. Elofson

M. Elofson

Natural Resources Department, Lower Elwha Klallam Tribe, Port Angeles, Washington, 98363 USA

Search for more papers by this author
E. M. Sampson

E. M. Sampson

Natural Resources Department, Lower Elwha Klallam Tribe, Port Angeles, Washington, 98363 USA

Search for more papers by this author
L. Ward

L. Ward

Natural Resources Department, Lower Elwha Klallam Tribe, Port Angeles, Washington, 98363 USA

Search for more papers by this author
First published: 18 August 2016
Citations: 12
Corresponding Editor: D. P. C. Peters.


Dam removal and other fish barrier removal projects in western North America are assumed to boost freshwater productivity via the transport of marine-derived nutrients from recolonizing Pacific salmon (Oncorhynchus spp.). In anticipation of the removal of two hydroelectric dams on the Elwha River in Washington State, we tested this hypothesis with a salmon carcass addition experiment. Our study was designed to examine how background nutrient dynamics and benthic food webs vary seasonally, and how these features respond to salmon subsidies. We conducted our experiment in six side channels of the Elwha River, each with a spatially paired reference and treatment reach. Each reach was sampled on multiple occasions from October 2007 to August 2008, before and after carcass placement. We evaluated nutrient limitation status; measured water chemistry, periphyton, benthic invertebrates, and juvenile rainbow trout (O. mykiss) response; and traced salmon-derived nutrient uptake using stable isotopes. Outside of winter, algal accrual was limited by both nitrogen and phosphorous and remained so even in the presence of salmon carcasses. One month after salmon addition, dissolved inorganic nitrogen levels doubled in treatment reaches. Two months after addition, benthic algal accrual was significantly elevated. We detected no changes in invertebrate or fish metrics, with the exception of 15N enrichment. Natural seasonal variability was greater than salmon effects for the majority of our response metrics. Yet seasonality and synchronicity of nutrient supply and demand are often overlooked in nutrient enhancement studies. Timing and magnitude of salmon-derived nitrogen utilization suggest that uptake of dissolved nutrients was favored over direct consumption of carcasses. The highest proportion of salmon-derived nitrogen was incorporated by herbivores (18–30%) and peaked 1–2 months after carcass addition. Peak nitrogen enrichment in predators (11–16%) occurred 2–3 months after addition. All taxa returned to background δ15N levels by 7 months. Since this study was conducted, both dams on the Elwha River were removed over 2011–2014 to open over 90% of the basin to anadromous fishes. We anticipate that as the full portfolio of salmon species expands through the basin, nutrient supply and demand will come into better balance.


Dams have long been a major feature on the global landscape, particularly in North America and Europe, where 80% of all large rivers are impounded (Lehner et al. 2011). There is no shortage of studies investigating the ways in which dams alter riverine ecosystems by disrupting hydrologic, sediment and thermal regimes; disconnecting rivers from their floodplains; and blocking passage of migratory species (Petts 1984, Nilsson et al. 2005).

However, a growing body of literature now documents the opposite trend, describing what happens to rivers when aging and obsolete dams are removed (O'Connor et al. 2015). Until recently, most of these cases involved relatively small structures less than 10 m in height (Poff and Hart 2002). In the last 5 yr, a number of dams taller than 30 m have been or are in the process of being removed in the U.S. Pacific Northwest (Service 2011, Blumm and Erickson 2012).

The largest such undertaking to date has been a 3-yr project (2011–2014) to simultaneously remove two dams from the Elwha River in Washington State (Warrick et al. 2015). Neither dam was built with fish passage, which restricted anadromous species to the lowermost 8 km of river for the last century. Although greatly reduced from historical levels, all five species of Pacific salmon as well as native trout (Oncorhynchus spp.), char (Salvelinus spp.), and lamprey (Entosphenus tridentatus and Lampetra spp.) have persisted downstream from the previously impassable Elwha Dam (Brenkman et al. 2012).

The Elwha Dam removal project provides a unique opportunity to study ecosystem restoration on a watershed scale. Due to both surficial geology and forest composition, many Pacific Northwest rivers have naturally low levels of nutrients available to primary producers (Vitousek and Howarth 1991, Gresh et al. 2000). These watersheds may become nutrient-limited when dams and other barriers prevent salmon migration between marine and freshwater habitats (Naiman et al. 2002). Recolonization by Pacific salmon upstream of the Elwha Dam is hypothesized to increase ecosystem productivity through incorporation of marine-derived nutrients into riparian and riverine food webs (Munn et al. 1999, Morley et al. 2008, Duda et al. 2011).

Long-term studies examining food web effects of salmon recolonization are few—but we can draw inferences from numerous nutrient enhancement studies that have become increasingly common in western North America (for review, see Compton et al. 2006 and Janetski et al. 2009). Although placement of salmon carcasses or nutrient analogs differs from naturally spawning salmon in a number of important ways (Holtgrieve and Schindler 2011, Tiegs et al. 2011), such experiments that examine energy transfer through freshwater food webs are relevant in the context of understanding the ecosystem outcomes of salmon recolonization.

To better interpret ecosystem response following dam removal, we conducted a salmon carcass placement experiment in the Elwha River over 2007–2008. Our study was designed to examine how nutrient dynamics and benthic food webs vary seasonally and in response to carcass addition. Although much has been written on heterogeneous carcass placement response across different watersheds, relatively little attention has been paid to seasonal context (but see Wilzbach et al. 2005, Lessard and Merritt 2006). By this, we refer to how the timing of carcass availability influences the relative importance of different pathways by which nutrients are incorporated into freshwater food webs, and how long these effects persist in time.

Our study had two main objectives. First, we wished to evaluate natural variability of background conditions (i.e., independent of salmon carcass addition) by asking the following questions: (1) Is the Elwha River nutrient-limited? and (2) How do nutrient limitation and benthic food web measures vary seasonally? Our second objective was to examine how salmon carcass addition changed these background conditions. Here, we asked the following questions: (1) What is the timing and magnitude of salmon-derived nutrient uptake through direct and indirect pathways? and (2) Can we detect a response in nutrient limitation status and benthic food web metrics from salmon carcass addition?


Study area

The sixth-order Elwha River originates in the Olympic Mountains and flows northward 72 km before emptying into the Strait of Juan de Fuca at the Lower Elwha Klallam Tribal Reservation (Fig. 1). Eighty-three percent of the river basin is protected within Olympic National Park, with 14 km of the Elwha downstream of the park boundary flowing through second-growth forests on a patchwork of private, state, and tribal lands. Western hemlock (Tsuga heterophylla) and Douglas fir (Pseudotsuga menziesii) are the dominant vegetation zones at lower elevations.

Details are in the caption following the image
Study region showing location of three reference and treatment reaches downstream (D1–D3) and upstream (U1–U3) of the Elwha Dam on the Olympic Peninsula (inset map) of Washington State, USA.

The Elwha (32 m high) and Glines Canyon (64 m high) dams were located at 8 and 22 km, respectively, from the river's mouth. Population numbers for all anadromous fish species remaining in the lower 8 km of river are critically low, with four listed as threatened under the U.S. Endangered Species Act. Hatchery intervention has further shifted historical species composition from an assemblage numerically dominated by pink (O. gorbuscha) and chum (O. keta) salmon to one composed largely of hatchery-reared coho (O. kisutch) and Chinook (O. tshawytscha) salmon (Pess et al. 2008). For further background on the natural setting of the Elwha River and the history of the dams, see Duda et al. (2008).

Our study was conducted in floodplain channels upstream and downstream of the Elwha Dam. In anastomosing river systems of the Pacific Northwest, side channels and other floodplain habitats often represent a disproportionately high percentage of available spawning and rearing habitat for fish and other fluvial species (Morley et al. 2005). Practically speaking, the smaller channel area and lower discharge of side channels allowed us to place carcasses at target loading densities, collect data year round, and retain more carcasses during periods of high flow. The Elwha River exhibits a bimodal hydrograph, with seasonal peaks in discharge during spring snow melt and again in the fall and early winter when the majority of precipitation falls.

Experimental design

Our experimental design was a before/after control/impact (BACI) model with spatially paired control and impact sites sampled at multiple times before and after treatment. All sampling occurred prior to dam removal from October 2007 to August 2008. We conducted our experiment in three side channels downstream of the Elwha Dam (elevations 5–9 m) and three between the Elwha and Glines Canyon dams (elevations 65–70 m). Within each side channel, we delineated a 50 m long treatment reach at the downstream end, separated by at least 200 m from a 50 m long reference reach at the upstream end. All channels were low gradient, with bankfull widths between 10 and 25 m, and a largely closed canopy at leaf-out (Table 1).

Table 1. Natural features of study side channels downstream (D) and upstream (U) of the Elwha Dam and percentage of carcass biomass remaining 1 and 2 months post-placement
Channel feature D1 D2 D3 U1 U2 U3
Water source GW Combo SW GW Combo SW
Elevation (m) 5 5 9 67 67 67
Bankfull width (m) 10 20 25 11 12 11
Gradient (%) 1.4 1.2 0.7 1.0 1.5 2.0
Canopy (%) 93 92 68 84 89 92
Mean flow (cm/s) 0.03 0.54 1.25 0.04 0.11 0.23
D50 (mm) 27 77 55 14 19 55
Carcass remaining (%)
1 month after 48 0 75 60 49 98
2 months after 0 0 8 27 39 39


  • Water source refers to dominant input by groundwater (GW), surface water (SW), or a combination of the two. Canopy coverage was measured at full leaf-out, mean flow calculated across sample events, and D50 is the median surface particle size diameter in sample riffles.

We began the nutrient limitation component of the study in October 2007. All other sampling began in December 2007 and was repeated again in January 2008. A week after January sampling, we evenly distributed coho salmon carcasses across all treatment reaches at a loading density of 0.75 kg/m2. After carcass addition, we resampled all reaches in February, March, April, and August 2008.

We selected coho because its late fall–winter spawn timing allowed us to sample in a season typically overlooked in salmon subsidy studies (but see Bilby et al. 1998, Wilzbach et al. 2005, Harvey and Wilzbach 2010). Within-basin coho carcasses were supplied by the Elwha tribal hatchery (mean mass 3.76 kg, SD 0.98 kg). Carcass loading density was driven by the total number of carcasses available and our intention of approaching N15 saturation—reported elsewhere as between 0.15 and 1.5 kg/m2 of salmon carcass (Wipfli et al. 1999, Bilby et al. 2001). Coho carcasses had a mean δ15N value of 13.39‰ (SD = 0.4‰, n = 10) and were composed of 14.57% N by mass (SD = 1.09%).

In the Elwha, coho enter the river from August through December and spawn October through January. Estimated annual escapement of coho salmon during the period of this study is 1000–3000 adults; it is unknown how many of these fish spawn in the river rather than return directly to the tribal hatchery (Pess et al. 2008). With the exception of the occasional winter steelhead (O. mykiss), we did not observe naturally spawning salmon in our study reaches over the course of the study.

In river systems with healthy salmon runs, carcasses that naturally wash out or are scavenged by predators are replaced by the arrival of new spawners or the delivery of carcasses from upstream reaches. In the absence of new sources and lacking the resources to continually place new carcasses throughout the experiment, we constructed loose-fitting plastic mesh bags that held two carcasses each and anchored these to the channel bottom with rebar. Although mesh size of the bags was large enough (5 × 5 cm) to allow access to the carcasses by invertebrates, juvenile fish, and birds, anchoring could have impeded scavenging by other riparian carnivores. To determine carcass decomposition rates, we reweighed five randomly selected carcass bags at each treatment reach every other week for 2 months following January placement.

Sampling parameters

Nutrient limitation

We evaluated nutrient limitation of algal accrual using nutrient-diffusing substrates (NDS) following protocols of Sanderson et al. (2009). Each experiment included four NDS types: control (no nutrients), nitrogen (N: 0.5 mol/L NH4NO3), phosphorus (P: 0.2 mol/L KH2PO4), and nitrogen + phosphorus (N + P: 0.5 mol/L NH4NO3 + 0.2 mol/L KH2PO4). NDS were constructed by filling polystyrene vials with bacteriological-grade agar at the nutrient concentrations described above. Each vial was then fused to a 27-mm silica crucible cover, which served as a periphyton colonization substrate. Five replicates of each NDS type were affixed to a wooden rack anchored within a riffle of each study reach. Vials were switched out every 29–40 d (mean 33 d) for each rack.

Following each monthly deployment, colonized silica covers were detached from the vials and held on ice until frozen at −20°C. In the laboratory, chlorophyll a was extracted in acetone and the absorbance of the resulting supernatant measured using a TD-700 fluorometer. Chlorophyll a concentration was converted to algal accrual rate (μg · cm−2 · d−1) based on crucible cover area and total colonization time (Sanderson et al. 2009). We deployed Hobo Water Temperature Pro (± 0.2°C) and Hobo Pendant Light data loggers (Onset Computer Corporation, Bourne, Massachusetts, USA) alongside each rack to measure average daily temperature and light availability.

Water chemistry

On each sampling occasion, we collected one set of dissolved and total nutrient samples from riffles at the downstream end of each treatment and reference reach. Dissolved nutrients samples were filtered in the field using a 0.45-µm (pore size) syringe filter. Samples were held on ice in the field and then frozen at −20°C until analyzed by the University of Washington Marine Chemistry Laboratory. Total nutrient samples were evaluated for total P (TP) and total N (TN). Dissolved nutrient samples were evaluated for phosphate (PO4-P), nitrate (NO3-N), nitrite (NO2-N), ammonium (NH4-N), and silicate (SiO4-Si). Concentrations of NO3, NO2, and NH4 were summed to yield dissolved inorganic N (DIN).


We measured periphyton standing crop from natural substrates by sampling three rocks (surface areas 50–340 cm2) from riffles distributed evenly across each sample reach. Periphyton was dislodged from each rock by scrubbing and rinsing with a stiff brush, pooled into one homogenized sample, and filtered onto three separate 47-mm glass–fiber filters (1 μm pore size) for analysis of chlorophyll a, ash-free dry mass (AFDM), and stable isotopes. Chlorophyll a concentration was measured as described above for nutrient-diffusing substrates and AFDM calculated following the gravimetric method (Steinman and Lamberti 1996). Both total periphyton (mg/cm2) and algal (μg/cm2) standing crop were converted to densities based on total rock surface area sampled at each site (Dall 1979).

Benthic invertebrates

We sampled benthic invertebrates by placing a Slack sampler (500-μm mesh, 0.25-m2 frame) at three riffle locations throughout each sample reach, pooling these into one sample, and preserving all invertebrates in 70% EtOH. From one-quarter of each sample, we identified all invertebrates to the family level and calculated total numeric density. From the remaining 75% of each sample, we selectively removed four taxa for stable isotope analysis: the predatory Plecoptera genus Sweltsa, the filter-feeding Trichoptera genus Hydropsyche, the algal-grazing Ephemeroptera genus Cinygmula, and the leaf-shredding Plecoptera family Capniidae. We selected these taxa because they represent different taxonomic orders and functional feeding groups and because they were sufficiently abundant across all of our study sites over most months of data collection.

Juvenile fish

For juvenile fish response, we focused on 0 + age O. mykiss (50–100 mm fork length) because this salmonid was present both upstream and downstream of the Elwha Dam, and this size cohort was most common across our study sites. On each sampling occasion, we collected fish using a backpack electrofisher. Each fish was weighed to the nearest 0.01 g and measured to the nearest 1 mm. We calculated condition factor as: = 105 × W/L3, where W is mass (g) and L is fork length (mm) (Fulton 1902). Other fish species present during collection trips included sculpin (Cottus spp.), lamprey ammocoetes, and juvenile coho and Chinook salmon at sites downstream of the dam. Upstream of the dam we observed sculpin and two char species: bull trout (S. confluentus) and the non-native brook trout (S. fontinalis).

We euthanized up to 10 O. mykiss per reach on each sample occasion to collect blood and measure energy density. Blood was centrifuged at 3000 g for 3 min to separate plasma from red blood cells. Red blood cells were frozen at −20°C for subsequent stable isotope analysis. Plasma samples were stored at −80°C until they were subjected to a radioimmunoassay for concentrations of the hormone insulin-like growth factor (IGF-I) (Beckman 2011).

Remains of each fish were frozen at −20°C for subsequent analysis of energy density. After thawing in the laboratory, wet mass was measured to the nearest 0.01 g, fish were dried for 24 h at 105°C to constant mass and then reweighed on the same scale to determine percent dry mass (PDRY = 100 × dry mass/wet mass). Energy density (kJ/g wet mass) was estimated proximately using the relationship previously established for juvenile coho with the equation: energy density = 0.34 × PDRY − 2.99 (Trudel et al. 2005).

Stable isotopes

Freeze-dried and homogenized tissues for invertebrates and fish were analyzed by mass spectroscopy at NOAA's Northwest Fisheries Science Center for δ15N, δ13C, [N], and [P]. Periphyton isotope filter samples were processed by the University of Alaska Fairbanks isotope laboratory for δ15N and δ13C. Previous comparisons of duplicate samples analyzed by both laboratories showed high correlations for both δ15N and δ13C (Duda et al. 2011). Nitrogen and carbon isotopic values are reported as per mil (‰).

We were unable to collect all target taxa on each sampling occasion. Missing data are due either to limited site access during periods of high flow or unavailability of specific taxa at time of sampling. December NDS data, along with temperature and light loggers, were lost due to a flood (the second highest event on record) on 3 December 2007 that buried equipment. Due to these missing data points and low sample size, we did not have adequate statistical power to include location (upstream vs. downstream of Elwha Dam) as a factor in our statistical analyses. Instead, we considered the overall response of our six paired reaches to nutrient-diffusing substrate experiments, seasonal variability, and carcass placement.

Statistical analyses

Nutrient limitation

We evaluated nutrient limitation status across reaches and months using a one-way analysis of variance (ANOVA, α < 0.05, n = 5) to compare chlorophyll a concentrations across NDS types. Data were log-transformed to meet assumptions of homogeneity of variance, and where treatment effect was significant, we applied Tukey's post hoc test to determine the type of nutrient limitation (N limitation, P limitation, primarily N limited and secondarily P limited, primarily P limited and secondarily N limited, or N and P co-limited) (Tank and Dodds 2003).

In addition to evaluating nutrient limitation status for each reach and month combination, we included two NDS metrics in subsequent analyses of overall seasonal variability and carcass response. Based on our prevalent finding of N and P co-limitation, we calculated the ratio of chlorophyll a concentrations on N + P vs. control substrates as a measure of the magnitude of nutrient limitation. We also used chlorophyll a data from control substrates as a measure of background algal accrual rate. Temperature and light data associated with NDS racks were collected for descriptive purposes only. We present these data graphically to illustrate the range of conditions over the course of the study.

Seasonal variability

To examine natural seasonal variability in all of our sample parameters independent of carcass placement, we analyzed data from reference reaches only. We tested for differences in algal accrual rate, magnitude of nutrient limitation, stable isotopes, water chemistry, periphyton, invertebrate, and fish response variables with a randomized block ANOVA, with month as a fixed factor and reach as a random effect. Model residuals were examined for approximate normality and equality of variance. Where appropriate, we log-transformed data and applied Tukey's post hoc test for comparisons of sample months.

To examine seasonal changes in benthic invertebrate assemblage structure, we used a suite of multivariate techniques available in the statistical software packages PRIMER (version 6.1.13; Clarke and Gorley 2006) and PERMANOVA (version 1.0.3; Anderson et al. 2008). We fourth-root-transformed invertebrate densities to balance contributions by dominant and rare species and then created triangular resemblance matrices of pairwise similarities between all reference sites using the Bray–Curtis distance. We tested for differences by month using a one-way randomized block design with month as a fixed factor and site set as a random effect. Where significant differences were detected, we used the SIMPER routine of PRIMER to determine which taxa contributed most to dissimilarities between groups.

Salmon-derived nutrient uptake

Based on our findings of significant enrichment in 15N but not 13C following carcass placement, we used a two-source single-element mass balance mixing model to estimate the proportion of salmon-derived N assimilated by sample taxa (Phillips and Koch 2002, Martínez del Rio and Wolf 2005). This model included concentration dependence for all taxa except periphyton, for which we did not have [N] data. In both the concentration dependent and simple linear mixing model, coho salmon carcasses were one food source (S), while reference taxa (R) represented the pool of all non-salmon base food sources. This approach assumes that with the exception of salmon carcasses, consumers in reference reaches were eating the same prey items in the same proportions as consumers in the paired treatment reach (Honea and Gara 2009). This approach does not identify or quantify these base food sources, but determines how much salmon contribute to the diet of consumers in treatment reaches.

Thus, to solve for the proportion of salmon-derived N assimilated by consumers, we used the formula:
where f represents the fractional contribution of N from each food source; the subscripts R, S, and T denote reference food sources, salmon food source, and the mixture of the two present in treatment reaches; [N] is the nitrogen concentration from each food source; δ15N is the N isotopic signature for each source; and Δ15Ntissue-S is the trophic fractionation of N between salmon carcass and consumer. We assumed that isotopic fractionation of N during assimilation by consumers was the same in reference and treatment reaches, but applied a correction of +1.3‰ per trophic level to the coho carcass food source (Post 2002). This value was based on 15N trophic enrichment observed in our reference reaches and in an earlier study (Duda et al. 2011).

Food web response

Taking advantage of our BACI design, we used a ratio approach to test the effects of carcass placement on the same sample parameters included in the seasonal variability analyses. For each of these variables, we calculated the paired ratio of treatment to reference reach (T:R) and log-transformed data where appropriate. As our data set did not meet the criteria of sphericity required for repeated-measures ANOVA, we instead used paired t tests to examine differences in T:R values before and after carcass placement. To reduce the total number of paired comparisons, we calculated a mean pre-treatment T:R value from sample months prior to carcass placement and compared this to T:R values at 1, 2, 3, and 7 months after carcass addition. For isotope data, we applied one-tailed paired t tests as we anticipated enrichment of 15N and 13C following carcass placement; for all other comparisons, we applied two-tailed t tests.

To test for differences in invertebrate assemblage structure between treatment and reference reaches before and after carcass placement, we used a two-way randomized block design in PERMANOVA. Our two fixed factors were treatment (carcass addition or no carcass addition) and time (before placement [December and January] and after [February, March, April, and August]). Site was set as a random effect, allowing us to block by paired reference and treatment reaches. Because we had uneven replication across months due to high flow events, we used the type III partial sum of squares to partition sources of variation in our model and test the interaction between treatment and time.


Salmon carcass tissue remained in the six treatment reaches from less than 1 week to over 2 months (Table 1). At the furthest downstream site, all carcasses disappeared within 1 week due to scavenging by gulls (Larus spp.). At the other two reaches downstream of Elwha Dam, over 40% of total carcass biomass remained 1 month after placement, but was completely depleted by 2 months. Carcasses remained for a longer period at all three reaches upstream of Elwha Dam, with over one-third of original biomass still present after 2 months. All carcasses were fully decomposed after 3 months.

Nutrient limitation

Across the six reference reaches, algal growth was co-limited by N and P in every season except winter (Fig. 2). This took either the form of N and P co-limitation (N + P > C) or primary N limitation and secondary P limitation (N + P > N > P > C). In both cases, chlorophyll a concentrations were significantly higher on N + P substrates than other NDS types for all study reaches at the onset of the study in October (one-way ANOVA, < 0.05, n = 5).

Details are in the caption following the image
Type and magnitude of nutrient limitation across reference reaches upstream and downstream of the Elwha Dam. Data are missing from D3 in June due to high flows limiting site access.

The consistent nutrient limitation observed in fall did not persist into winter (Fig. 2). In January, we detected no differences in algal growth between substrate types. In February, three of the six reference reaches continued to show no nutrient limitation, while the other half reverted to N and P co-limitation. By March, chlorophyll a concentrations were again consistently highest on N + P substrates at all reference reaches and generally persisted in this pattern through summer. Magnitude of nutrient limitation ((N + P)/control) also varied seasonally and was greatest in spring and summer relative to January (randomized block ANOVA, < 0.05, n = 7) (Fig. 3a).

Details are in the caption following the image
Seasonal variation across references reaches for (a) nutrient-diffusing substrates, (b) water chemistry, (c) periphyton, (d) benthic invertebrates, and (e) juvenile Oncorhynchus mykiss. Values are monthly means ±1 SD (n = 6). C = algal accrual rate on control substrates, NP/C = the ratio of accrual on N + P-amended substrates to control, DIN = dissolved inorganic nitrogen, AFDM = ash-free dry mass, K = Fulton's condition factor, ED = energy density, and IGF = insulin-like growth factor. Letters indicate significant differences between months (randomized block ANOVA,< 0.05, n = 6).

Seasonal variability

Independent of carcass addition (reference reaches only) we observed strong seasonal patterns in many of our study variables. Mirroring seasonal nutrient limitation, ambient nutrient concentrations at reference reaches were highest in winter (Fig. 3b) and light and temperature at their lowest levels (Fig. 4). Phosphate, TP, and DIN were all significantly higher in winter than in spring (randomized block ANOVA, < 0.05, n = 6).

Details are in the caption following the image
Environmental characteristics of Elwha River study channels showing (a) mean weekly water temperatures, (b) mean light intensity (±1 SE) and photoperiod, and (c) discharge in the Elwha River measured at USGS stream gage 12045500, upstream of Elwha Dam. Data gaps in temperature are due to logger malfunction and equipment loss during high flow events.

Across the six study channels, mean weekly water temperatures ranged 3–7°C during winter months, with groundwater-fed study channels typically 2–3°C warmer than surface water-fed channels (Fig. 4a). All six channels converged by early spring to 5–6°C and then steadily climbed to 11–16°C by late summer. Light levels followed a bimodal distribution with low values observed across winter and early summer and peaks in spring and late summer (Fig. 4b). This pattern reflects a combination of a photoperiod, spring leaf-out, and turbidity peaks associated with high flow events (Fig. 4c).

Background levels of periphyton standing crop observed on river rocks were opposite to those observed for water chemistry. Both total periphyton and algal densities were lowest in the winter and highest in the spring (Fig. 3c) (randomized block ANOVA, < 0.05, n = 7). However, we did not detect seasonal differences in algal accrual rates, as measured by periphyton growth on control NDS (Fig. 3a). This may be because benthic invertebrate mean density increased along with the periphyton density and was significantly greater in March than in January (randomized block ANOVA, < 0.05, n = 6) (Fig. 3d).

We did not observe significant differences in invertebrate taxa richness by month (Fig. 3d), but did detect changes in overall assemblage structure (randomized block PERMANOVA, < 0.01). Based on pairwise comparisons and SIMPER, December and January samples were similar to each other and distinguished from other months by a high proportion of Ephemeroptera from the families Ephemerellidae, Heptageniidae, and Baetidae (< 0.05). February, March, and April samples grouped together and were characterized by increased abundance of the Dipteran family Chironomidae. August samples significantly differed from all other months (< 0.01) and were dominated by Chironomidae, Oligochaetes, and the Plecopteran family Nemouridae.

Juvenile O. mykiss growth metrics displayed strong seasonal patterns. Fish energy density decreased significantly from December through April, while IGF rapidly increased in the opposite direction (Fig. 3e). Condition factor did not vary significantly by season. We were unable to collect enough fish in August to include summer in our seasonal analyses.

We observed limited seasonal variability in δ13C and δ15N across reference reaches (Fig. 5). Sweltsa δ15N was significantly higher in February than August and Hydropsyche δ15N was higher in January and March relative to April (randomized block ANOVA, < 0.05, n = 6). We detected no significant difference by month in δ13C for any sample taxa.

Details are in the caption following the image
Plots of mean δ15N (y-axis) and δ13C (x-axis) values (n = 3) for juvenile Oncorhynchus mykiss, four benthic invertebrate taxa, and periphyton collected from study channels downstream and upstream the Elwha Dam pre- and post-carcass placement. Symbols indicate values for different taxa, with blue and gray fills indicating treatment and reference reaches, respectively. Boxes highlight extent of overlap in isotopic range between reaches (solid boxes include all treatment taxa and dashed boxes all reference taxa). Missing data points are due to insufficient biomass for a given taxon on that sample occasion. Mean isotope values of coho carcasses placed during this experiment were 13.39‰ for δ15N (SD = 0.41‰) and –18.44‰ for δ13C (SD = 0.86‰).

Salmon-derived nutrient uptake

Prior to carcass placement, δ15N and δ13C values were nearly identical between paired reference and treatment reaches (Fig. 5). We detected significant enrichment of 13C in Capniidae one-month post-carcass placement (one-tailed paired t test, < 0.05, n = 6), but no other differences in δ13C for any taxa. In contrast, 15N was significantly enriched in treatment reaches for all sample taxa post-carcass placement (one-tailed paired t test, < 0.05, n = 6). Based on our mixing model, sample taxa from treatment reaches derived a monthly average of 5–30% of their N from coho salmon in the 3 months following carcass placement (Fig. 6).

Details are in the caption following the image
Boxplots of the percentage of nitrogen (N) derived from salmon by four benthic invertebrate taxa, periphyton, and juvenile Oncorhynchus mykiss following mid-January coho carcass placement. Blue circles display the data points summarized within each boxplot. Significant monthly differences in δ15N between reference and treatment reaches are indicated with asterisks (one-tailed paired t test, < 0.05). Missing data points are due to insufficient biomass for a given taxon on that sample occasion.

The largest observed changes in δ15N were in the shredder Capniidae, for which δN15 values were 3–4 times greater at 1 and 2 months post-treatment than at pre-treatment (one-tailed paired t test, < 0.05, Fig. 6a). This translated to a mean of 30.0% (SD = 13.5%) of total N derived from salmon 1 month following carcass placement. As we were unable to collect sufficient biomass of Capniidae at all study reaches before carcass addition, we directly compared reference to treatment isotope values post-treatment rather than pre- and post-T:R ratios. This approach assumes no preexisting difference between paired reaches prior to the experiment—which held true for all taxa sampled (Fig. 5).

Starting at the base of the food web, periphyton derived a monthly average of 9–22% of its N from salmon, with 15N significantly enriched at 1 and 2 months post-placement (Fig. 6b). The algae grazer Cinygmula displayed a pattern similar to that of periphyton, with treatment reaches significantly enriched in 15N relative to references reaches at 1, 2, and 3 months post-carcass placement (Fig. 6c). The predator Sweltsa was significantly elevated at 2 and 3 months post-placement, with 14–16% salmon-derived N (SD = 8.6%) (Fig. 6d). Juvenile O. mykiss reached significantly enriched 15N levels by 3 months post-carcass placement (salmon-derived N mean = 10.5%, SD = 8.1%) (Fig. 6e). The smallest level of detectable change we observed was for the filter feeder Hydropsyche, for which 15N was significantly elevated only at 1 month post-carcass placement (salmon-derived N mean = 4.7%, SD = 3.6%) (Fig. 6f). Seven months post-carcass placement, δ15N had returned to background levels for all sample taxa for which we were able to collect sufficient biomass in August (Fig. 6).

Food web response

We detected few changes in nutrient limitation and benthic food web measures in relation to the placement of salmon carcasses. Background concentrations of DIN doubled in treatment reaches at 1 month post-placement (two-tailed paired t test, n = 6, < 0.05), but other water chemistry parameters did not differ significantly before vs. after placement. At 3 months post-carcass placement, algal accrual rates on control substrates in treatment reaches increased significantly (two-tailed paired t test, < 0.05, n = 6). However, we detected no changes in rock periphyton or algal densities as a result of carcass placement (two-tailed paired t test, > 0.05, n = 6).

In the second month after carcass placement (March), N limited algal growth in half of treatment reaches vs. all of the reference reaches. For the remainder of the study period, nutrient limitation patterns in treatment reaches remained similar to those in reference reaches. There was also no detectable change in magnitude of nutrient limitation pre- and post-carcass placement (two-tailed paired t test, > 0.05, n = 6). Nor did we detect change in any of our response variables for benthic invertebrates (density, taxa richness, taxonomic structure) (two-way randomized block PERMANOVA, treatment × time > 0.05) or fish (energy density, IGF, condition factor) (two-tailed paired t test, > 0.05, n = 6).


Nutrient limitation

Our study confirmed the previously untested assumption that primary productivity on the Elwha River is limited by nutrient availability (Munn et al. 1999, Duda et al. 2011). Outside of winter months, benthic algae accrual rates were consistently higher when growth substrates were amended with N and P (Fig. 2). The most probable explanation for N and P co-limitation is that cycling was closely linked, and enrichment of one nutrient induced limitation of another (Elser et al. 2007). These results are consistent with numerous studies across habitat types and ecoregions that have demonstrated increased rates of algal growth and biomass following addition of both N and P (Tank and Dodds 2003, Chaloner et al. 2007, Harding et al. 2014, Marcarelli et al. 2014).

Seasonal variability

Seasonal patterns exceeded the treatment effects of carcass addition in the majority of our monitoring parameters. Natural seasonal variability in temperature, light, river discharge, and leaf litter input all influence aquatic primary and secondary production, and the patterns we observed in our sample variables largely reflected these natural seasonal gradients. Both light levels and water temperature were at their lowest annual levels for a large part of our study period from late December through February (Fig. 4a, b).

Absence of nutrient limitation during January and February suggests that low temperatures and light availability superseded nutrients as limiting factors during winter (Hill and Knight 1988, Morin et al. 1999, Sanderson et al. 2009). Dissolved nutrient concentrations were highest in the winter following deciduous leaf litter input and during a period of low biological demand (Fig. 3b). As day length and temperature increased in the spring (Fig. 4), dissolved nutrient concentrations decreased (Fig. 3b), and nutrient limitation increased (Figs. 2 and 3a).

Periphyton and algal standing crop peaked in the spring, before declining in the late summer (Fig. 3c). This decrease in standing crop is potentially due to higher grazing pressure from benthic invertebrates, whose densities also peaked in early spring and remained high through summer (Fig. 3d). Although we did not detect seasonal change in invertebrate taxa richness, we only identified taxa to the family level. Higher taxonomic resolution and direct measurement of growth would likely have improved our ability to detect benthic invertebrate response to both seasonal variability and salmon subsidies.

Among our fish response variables, energy density of juvenile O. mykiss slowly decreased from December through April (Fig. 3e). This reflects a depletion of stored lipids over a period of limited feeding opportunities (Biro et al. 2004, Trudel et al. 2007). Conversely, IGF-I concentrations increased 10-fold from December until April (Fig. 3e). These patterns indicate that O. mykiss put increasing spring prey availability toward growth rather than lipid storage, as is typically observed when energy intake is limited (Gardiner and Geddes 1980). The strong seasonal patterns we observed in IGF-I concentrations support more wide-ranging use of this physiological assay as a functional metric of freshwater fish growth (Beckman 2011).

Salmon-derived nutrient uptake

The timing and magnitude of changes in δ15N and δ13C across our study taxa suggest that the prevalent pathway by which salmon nutrients entered the benthic food web was uptake of dissolved nutrients by primary producers. Had consumers been feeding directly on carcasses, we would have expected to observe enrichment in both 15N and 13C, rather than in 15N only (Post 2002). The conclusion that bottom-up energy transfer was more prevalent in our experiment is supported by significant increases in DIN and algal accrual rate following salmon placement.

Salmon-derived N increased most rapidly and to the greatest extent in periphyton and herbivorous invertebrates (Fig. 6). However, the largest increase in δ15N was within a detrital pathway, as the leaf-shredding Capniidae received over 30% of its N from salmon 1 month after carcass placement. High 15N enrichment in Capniidae could reflect consumption of enriched microbes growing on leaf litter (Claeson et al. 2006, Marcarelli et al. 2014). Another possibility is that Capniidae fed directly on the carcasses themselves. This could explain why this is the only taxa for which we also detected a significant enrichment in 13C. Capniidae have been observed on salmon carcasses in Alaskan streams (Chaloner et al. 2002), and other studies report detritivorous invertebrates opportunistically switching to salmon carcasses (Minakawa et al. 2002).

For higher trophic-level taxa, statistically significant enrichment levels were reached after 2 and 3 months in the predatory invertebrate Sweltsa and after 3 months in O. mykiss. However, the proportion of salmon-derived N in Sweltsa and O. mykiss was smaller than for periphyton or herbivores. We hypothesize that this lag was due to predators feeding primarily on invertebrate prey sources enriched with salmon nutrients. The relative importance of bottom-up vs. direct consumption of carcasses is driven in part by the timing of nutrient availability, and underscores the importance of seasonal context in interpreting ecosystem response to salmon subsidies.

One interesting result of this work was the observation that 15N enrichment following carcass placement was more pronounced upstream of the Elwha Dam where anadromous salmon had been absent for 100 yr (Fig. 5). Our previous research showed enriched isotope levels across trophic levels downstream of the Elwha Dam, where salmon still had access prior to dam removal (Duda et al. 2011); this pattern was also observed in the present study (Fig. 5). However, after carcasses were introduced, δ15N values for periphyton and invertebrate taxa from treatment reaches upstream of the Elwha Dam were much higher than those downstream. It is unclear what was driving this pattern. Carcasses were slower to decompose at all upstream reaches (Table 1) and thus were available over a longer period of time. Existing differences in available nutrient pools may have also favored preferential uptake of the heavier marine isotopes upstream of the Elwha Dam (Phillips and Gregg 2003).

Food web response

Although our study suggests that Elwha River primary productivity may benefit from increased nutrient subsidies, autotrophic nutrient limitation did not appear to be alleviated in the presence of carcasses. Nutrient limitation (type and magnitude) differs between heterotrophs and autotrophs (Tank and Dodds 2003, Rüegg et al. 2011); therefore, it is possible that we would have seen a response in heterotrophic nutrient limitation had we directly measured microbial production. The high autotrophic index (AFDM/chlorophyll a) values we observed from periphyton rock samples during this and previous research (Morley et al. 2008, Coe et al. 2009) suggests that heterotrophic production is a large component of total primary production, particularly downstream of the Elwha Dam.

Although we observed short-term changes in DIN and algal accrual rate on control substrates following carcass placement, we did not detect changes in periphyton rock standing crop, invertebrate, or fish metrics. In a review of 37 salmon subsidy studies, Janetski et al. (2009) found that while all response variables except δ13C exhibited overall positive effects, they differed widely in magnitude and variability. Dissolved nutrient concentrations and δ15N displayed the largest positive effects, whereas dissolved organic carbon, periphyton, and invertebrate standing crop, the smallest. Fish response was most variable. Although many studies have demonstrated uptake of salmon-derived nutrients into aquatic food webs, this additional nutrient source does not always result in detectable shifts in benthic food web structure or productivity. This heterogeneity in response is often attributed to differences in environmental features, experimental approach, the timing of nutrient delivery relative to demand, and to retention capacity (Tiegs et al. 2011, Rinella et al. 2013, Marcarelli et al. 2014).

In our experiment, it is probable that pulsed nutrients available in the form of salmon carcasses were not available at a time when they could be fully utilized by other aquatic species (Wilzbach et al. 2005, Lessard and Merritt 2006, Harvey and Wilzbach 2010). Winter is typically a period of low productivity for many stream organisms. Juvenile fish that are relatively inactive may not respond to carcasses as they would during more active periods of growth. IGF levels were below detection limits for 41% of fish collected from December through February, compared to 5% collected during March and April. Energy density of these fish steadily decreased from December through April (Fig. 3e). The historically large flood that occurred 1 month prior to carcass placement also likely influenced our experimental outcomes (Fig. 4c). Following this event, we observed low densities of juvenile fish for the remainder of our study period.

Nutrient supply may have outstripped demand during our winter experiment, but we did observe nutrient retention into spring. Three of our six sample taxa were still significantly enriched in 15N in April, 3 months after carcass placement (Fig. 6). This was perhaps most notable for juvenile fish, where the rapid tissue turnover rate of blood (Church et al. 2009) made it more likely that salmon-derived nutrients had been recently consumed. April was also the only month in which periphyton growth rates were significantly elevated in treatment reaches, providing further evidence that salmon nutrients were retained through winter.

We hypothesize that under natural spawning conditions, total nutrient inputs and retention would be even higher. Unlike spawned-out carcasses, live spawners also contribute nutrients in the form of excreted NH4+ and energy-rich gametes (Janetski et al. 2009, Tiegs et al. 2011). During redd construction, nutrients can be moved into and stored in the benthos via the flocculation of salmon organic matter and sediment (Rex and Petticrew 2010). Unrestricted dispersal of carcasses to the floodplain by scavenging and flooding also increases nutrient storage potential in riparian soils, vegetation, and terrestrial organisms (Helfield and Naiman 2001, Fellman et al. 2008).

However, overall effects of salmon on primary producers can be hard to predict (Bellmore et al. 2014, Harding et al. 2014). Bioturbation of the bed by spawning salmon can reduce periphyton and benthic invertebrate standing crop (Moore and Schindler 2008, Honea and Gara 2009). But increased sediment mobilization, combined with elevated nutrient concentrations during spawning events, can also result in high rates of microbial heterotrophic production (Holtgrieve and Schindler 2011). The net effect of both increased nutrients and bed disturbance varies considerably by habitat and interannual variation in salmon density (Chaloner et al. 2007, Rüegg et al. 2012, Harding and Reynolds 2014).


We observed uptake of salmon-derived nutrients across multiple freshwater trophic levels, but this uptake did not translate into dramatic effects on algal nutrient limitation, benthic invertebrate assemblages, or fish growth. Rather, natural seasonal variability in our response metrics was greater than a salmon carcass effect. We hypothesize that this lack of response was due in large part to differences between carcass placement and natural salmon spawning runs and the timing of nutrient availability relative to demand. Neither of these factors is relevant for natural fish recolonization involving multiple species with diverse life-history strategies.

With both dams on the Elwha removed as of September 2014, Pacific salmon, steelhead, and lamprey are now poised to recolonize over 150 km of mainstem, side channel, and tributary habitat (Pess et al. 2008). Coho, steelhead, sockeye (O. nerka), lamprey, and Chinook have already recolonized the river section between the two former dams, and Chinook and bull trout moved upstream past the former Glines Canyon Dam in fall 2014 (M. Elofson, personal observations). If projections made prior to dam removal are accurate, full recovery of all anadromous fish could provide an annual increase of 8900–23,690 kg of N and 1050–2805 kg of P, with returning adult salmon present nearly year round in the Elwha watershed (Munn et al. 1999, Pess et al. 2008).

The pathways these nutrients take through the freshwater food web will depend on a number of factors including escapement levels, run timing, and species-specific spatial distribution. We expect that changes in freshwater productivity will occur both by direct consumption of salmon tissues (Tonra et al. 2015) and indirectly via the bottom-up stimulation of primary production observed in this study. A majority of projected salmon nutrients will be delivered by Chinook and pink salmon from late spring through early fall, seasons in which we observed consistent N and P limitation of algal growth. As recolonization by all anadromous fish species unfolds in the Elwha River, timing of marine-derived nutrient delivery will better match biological demand.

The strong seasonal variation observed in our sample parameters was not unexpected, but serves as a reminder of how sample timing can potentially bias data interpretation and the conclusions drawn from this and other monitoring projects (Larsen et al. 2001). For many practical reasons such as site accessibility and limited funding, most monitoring studies are conducted during a relatively narrow pre-determined index period. For instance, our 3-yr data set of intensive pre-dam removal monitoring was collected in late summer only (Morley et al. 2008, Duda et al. 2011). Results from this experimental study led us to fine tune our ongoing dam removal monitoring research to better incorporate natural seasonal variation in our response metrics.


This work could not have been undertaken without support from numerous colleagues, students, and volunteers from the Lower Elwha Klallam Tribe's Natural Resources staff, the NWFSC Watershed Program, and Peninsula College. In particular, we would like to thank T. Bennett, J. Ganzhorn, K. Kloehn, R. Moses, V. Pelekis, G. Pess, C. Tran, and C. Vizza for assistance in the field and laboratory. M. Liermann provided statistical consultation and O. Stefankiv assisted with figure preparation. This article benefited from comments by J. Bellmore, A. Collins, B. Sanderson, and two anonymous peer reviewers. Funding was provided by the NWFSC Internal Grant Program, the Lower Elwha Klallam Tribe, and the U.S. Geological Survey's Ecosystems mission area. Use of trade names is for descriptive purposes only and does not constitute endorsement by the U.S. Government.