Invertebrate prey contributions to juvenile Coho Salmon diet from riparian habitats along three Alaska streams: Implications for environmental change

Abstract Stream fish rely on a mix of terrestrial and aquatic prey sources. While the importance of terrestrial invertebrates as a food source for stream fish is well documented, the role of aquatic insects that emerge from the stream as winged adult insects (aquatic winged adults) and return to the stream as prey is less understood. In this study we determined the proportion of total diet for stream-rearing juvenile Coho Salmon (Oncorhynchus kisutch) that is derived from terrestrial and aquatic winged adult invertebrates which enter the stream from riparian habitats and consider how those cross-ecosystem prey contributions vary based on riparian habitat type. Study reaches were identified in three streams within the Kenai River watershed of Alaska that were representative of habitats found throughout the region and riparian vegetation was classified into grass/sedge, shrub and tree types using LiDAR. Juvenile Coho Salmon stomach contents were sampled seasonally in study reaches over a two-year period and ingested invertebrates were identified by taxa, life stage and origin. Our results showed that aquatic winged adult prey contributions to juvenile salmon diet were significantly lower in the grass/sedge study reach, and cross-ecosystem invertebrate prey represented a significantly higher proportion of juvenile salmon diet in the tree study reach. Invertebrate prey in the grass/sedge reach were composed primarily of the larval life stage of aquatic winged adults. These results suggest that change in riparian vegetation from tree/shrub to grass/sedge along Kenai streams as projected by regional climate change models, or that results from anthropogenic modification, will likely lead to lower availability of cross-ecosystem prey for stream fish. Management of riparian buffers along streams to preserve or increase occurrence of trees and shrubs is likely to help mitigate impacts of those possible changes.

Introduction lateral dispersal related to landscape openness. Movement by aquatic winged adults away from their natal stream reduces their chances of returning to the stream and becoming potential prey for stream fishes (Jackson and Fisher 1986;Briers and Gee 2004).
In this study, we measured contributions by terrestrial, aquatic and aquatic winged adult invertebrates to the diet of stream-rearing juvenile Coho Salmon Oncorhynchus kisutch within the Kenai River watershed of southcentral Alaska and calculated the total invertebrate contribution derived from invertebrates that enter the stream from riparian habitats (cross-ecosystem). Invertebrate contributions to juvenile salmon diet were compared among three study reaches representative of common riparian vegetation types found within the Kenai watershed. We hypothesized that terrestrial and aquatic winged adult invertebrates were an important component of juvenile Coho Salmon diet that is moderated by vegetation type. Our prediction was that the fractional contribution to juvenile salmon diet of aquatic winged adult and cross-ecosystem invertebrates would be lower in streams with more open riparian habitats (grass/sedge) than those in shrub or tree riparian habitats.

Study area
This study was conducted within the Kenai River watershed of southcentral Alaska which covers approximately 5,600 sq km and includes over 2,600 km of mapped tributary streams and rivers ( Extended, homogeneous study reaches within three salmon-rearing tributaries (Beaver Creek, Ptarmigan Creek, Russian River) of the Kenai River were chosen for study that were representative of the most common vegetation types (grass/sedge, shrub, tree) occuring along streams in the Kenai watershed ( Figure 3). These extended study reaches were selected to avoid mixing of invertebrate input from diverse vegetation types or the movement of juvenile salmon among riparian habitats (Wipfli 1997;Allan et al. 2003;Roon et al. 2016). Sampling replicates of similar extended, homogeneous reaches was not practical, consequently we could not statistically assess effects of riparian type. This study therefore provides a statistical comparison of study reaches and considers how observed differences in prey contributions could be attributed to variation in riparian type.
Beaver Creek (BC) is a lowland stream with an overall 2% average gradient and 0% direct glacial influence. Vegetation adjacent to the BC study reach was predominately grass and sedge (Calamagrostis spp., Deschmapsia spp., Arctogrostis spp., and Carex spp.). Ptarmigan Creek (PC) is a higher elevation glacially fed stream with an overall 14% average gradient and 7% glacial watershed coverage. Vegetation adjacent to the PC study reach was predominantly shrubs (Alnus spp. and Salix spp.) as well as deciduous (Betula spp. and Populus spp.) and coniferous trees (Picea spp.). Russian River (RR) is a montane tributary with an overall 9% average gradient and has minor (< 1%) glacial coverage within its watershed. Vegetation adjacent to the RR study reach is dominated by mixed  deciduous and coniferous trees (Betula spp., Populus spp. and Picea spp.) with deciduous shrub (Alnus spp. and Salix spp.) understory. While areas within the lower Kenai River watershed have experienced substantial human development, the study reaches were generally undeveloped, and as a result riparian habitats were relatively intact and undisturbed.

Characterizing vegetation type
Remote sensing data are useful for mapping ecological patterns (Dauwalter et al. 2017), and in this study we used LiDAR data to classify vegetation type adjacent to the three study reaches according to height. The LiDAR data were collected at a horizontal resolution of 1.2 m with an expected vertical accuracy of 0.15 m. Vegetation height categories were chosen based on vegetation classes defined for Alaska vegetation (Viereck et al. 1992): grass/sedge 0.0-0.6 m shrub >0.6-6.0 m and tree > 6.0 m. Vegetation was classified 150 m upstream/downstream and 20 m inland from the stream bank at fish sampling sites and along the study reaches.

Juvenile salmon sampling
Fish sampling sites of 150 m were identified at the upper and lower end of each of the three study reaches. Repeat sampling of juvenile Coho Salmon occurred at each sample site during Spring (June), Summer (July) and Fall (August-September) for 2015 and 2016 ( Figure 2). At each sampling event, six to 12 GeeV R minnow traps with 6.4 mm mesh were randomly placed. Traps were baited with sterilized salmon roe in perforated Whirl-Pak V R bags, and submerged 15-45 cm. A total of 1302 traps were set for an average period of 2.5 ± 1.4 hours (mean ± SD).
Captured juvenile salmon were anesthetized by submersion in a bath of AQUI-S 20E TM administered at 20 mg/L for 2 to 3 minutes or until fish exhibited total loss of equilibrium. Fork length was measured to the nearest millimeter, and weight was measured to the nearest tenth of a gram with an electronic balance. Gastric lavage sampling of a random subset of captured fish with fork length greater than 50 mm was conducted to determine diet. Stomach contents were preserved in 70% ethanol. All fish were released near the point of capture when sampling was complete at the end of each day. Juvenile Coho Salmon with no stomach contents or with salmon eggs as stomach contents were not included in the analysis (Table 1).
Identifiable invertebrates from stomach samples were classified to Family, or the next highest reliable taxa and life stage (Merritt and Cummins 1996). Length of partially digested prey was estimated based on intact individuals of the same taxon that appeared similar in size (Wipfli 1997). Estimation of prey dry mass was based on measured invertebrate body length with the allometric formula W ¼ aL b , where W is total dry body mass, L is total body length, and a and b are constants of the regression between W and L (Ricker 1973). Length-mass regression constants a and b were derived from a database of allometric juvenile salmon prey length-weight relationships (B.E. Meyer, unpublished data).
Invertebrate taxa were grouped into three primary prey categories based on origin and life stage: aquatic (larval and adult life stages of invertebrates that originate and reside instream), terrestrial (invertebrates that originate and reside in the terrestrial environment) and aquatic winged adult (aquatic-born invertebrates that emerge from the stream to the terrestrial environment as winged adults). The origin of some adult Diptera prey taxa were unknown, and were categorized as Diptera of unknown origin. The cross-ecosystem contribution to juvenile salmon diet that resulted from the movement of invertebrates from the terrestrial environment to the stream was calculated as the sum of terrestrial, aquatic winged adult and adult Diptera of unknown origin categories.
The percentage of total invertebrate prey consumed (dry mass) was calculated for each prey category in each study reach to allow comparison of invertebrate prey consumption among study reaches. Dry mass values of fish stomach contents were standardized for fish length to minimize fish-size bias. Arcsine-square-root transformation of percentages was used to standardize variance and improve normality of the data for statistical analysis. One-way analysis of variance (ANOVA-SPSS v.24) of transformed values was used to evaluate the difference between study reaches for invertebrate prey categories using a robust test of equality of means (Welch) and post-hoc Games-Howell tests where equal variances are not assumed (SPSS v.24) were used to conduct pairwise comparison of means between study reaches. Values were considered significant at a < 0.05. Lack of replication in treatment categories limits inference to the three study reaches.

Vegetation type
The composition of vegetation types adjacent to sample sites differed substantially between study reaches ( Figure 4). Grass/sedge represented 89%, 12% and 8%, shrubs 8%, 55% and 43%, and trees represented 3%, 33% and 49% at sample sites on BC, PC and RR, respectively. Riparian vegetation types along study reaches were similar to vegetation type found at sampling sites ( Figure 4).

Invertebrate contributions to juvenile Salmon diet
Stomach contents of 405 juvenile Coho Salmon were analyzed during the 2015 (n ¼ 210) and 2016 (n ¼ 195) field seasons. In 2015, 76 distinct combinations of taxa and life stage were identified among 2,481 total diet items while in 2016, 96 distinct combinations of taxa and life stage were identified among 6,822 total diet items. The dominant aquatic taxa ingested by juvenile salmon were similar for all study reaches and consisted primarily of larval life forms of aquatic winged adults including: fly larvae (Diptera), caddisfly larvae (Trichoptera), mayfly larvae (Ephemeroptera) and stonefly larvae (Plecoptera)) ( Table 2). These taxa represented 92%, 88% and 95% of the aquatic biomass in juvenile salmon diet for BC, PC and RR. Wasps and ants (Hymenoptera) were the largest contributors of terrestrial invertebrate biomass for each of the study reaches providing 39%, 39% and 31% of the total terrestrial contribution for   BC, PC and RR. Butterfly/moth larvae (Lepidoptera) represented a larger percentage by weight of total invertebrates consumed by salmonids within PC (20%) and RR (9%) as compared to BC (0%), however, they were infrequent in stomach contents (n ¼ 17) during the 2015-2016 sampling periods. The combined aquatic winged adult contributions to juvenile salmon diet by adult fly (Diptera), adult mayfly (Ephemeroptera) and adult caddisfly (Trichoptera) represented 100% and 96% within BC and PC respectively for those study reaches. Those taxa represented 74% of aquatic winged adult invertebrates consumed by juvenile salmon in the RR, and adult stonefly (Plecoptera) represented 26%. The contribution of aquatic winged adult invertebrate prey to juvenile salmon diet was 5%, 9% and 15% for BC, PC and RR, respectively ( Figure 5 and Tables 3 and 4) which was statistically different among study reaches (F 2, 234.7 ¼ 4.431, P ¼ 0.013). Pair-wise comparison of BC and RR showed significant differences for aquatic winged adult prey contributions (P ¼ 0.013). Adult Diptera of unknown origin contributed 17%, 12% and 26%, respectively, to total invertebrate prey consumed by juvenile salmon in BC, PC and RR which was statistically significant among study reaches (F 2, 260.8 ¼ 6.079, P ¼ 0.003). Pairwise comparison between PC and RR showed significant difference for adult Diptera of unknown origin (P ¼ 0.002). The proportion of juvenile salmon diet represented by aquatic invertebrates was significantly different among study reaches (F 2, 260.4 ¼ 13.707, P < 0.001) representing 62%, 55% and 37%, respectively, for BC, PC and the RR. Post-hoc comparison showed significant differences between RR and BC (P < 0.001) and PC (P ¼ 0.001). Terrestrial invertebrate prey contributed 15%, 22% and 21% for BC, PC and RR, respectively; however, these differences were not significantly different (F 2, 252.6 ¼1.548, P ¼ 0.215). Contributions by all cross-ecosystem sources of invertebrate prey were 38%, 45% and 63%, respectively, for BC, PC, and RR ( Figure 6) which differed significantly among study reaches (F 2, 260.5 ¼12.453, P < 0.001). Pair-wise comparison Cross-Ecosystem Figure 6. Percent of total prey dry mass ingested by juvenile Coho Salmon for cross-ecosystem and aquatic invertebrate prey categories within each study reach.
showed significant difference in cross-ecosystem invertebrate prey consumed by juvenile salmon in RR as compared to PC (P ¼ 0.003) and BC (P < 0.001). Seasonal contribution of invertebrate prey to juvenile salmon diet varied among study reaches (Figure 7). Aquatic invertebrates provided the majority (>50%) of prey resources to juvenile salmon in BC and PC throughout the year. In RR, aquatic invertebrate contributions decreased below 50% in Spring which was coincident with increased fractional consumption by juvenile salmon of Diptera of unknown origin and aquatic winged adult invertebrate prey and again in Fall when terrestrial and Diptera of unknown origin prey provided increased fractional contributions to diet. During Spring in BC, a reduction in the proportion of aquatic prey consumption was coincident with an increase in the proportion of Diptera of unknown origin consumed by juvenile salmon.

Discussion
Our results support the prediction that the fractional contribution to juvenile salmon diet of aquatic winged adult and cross-ecosystem invertebrates would be lower in streams with  more open grass/sedge riparian habitats than those in shrub or tree riparian habitats. We found the proportion of juvenile Coho Salmon diet provided by cross-ecosystem invertebrate prey (terrestrial, aquatic winged adult and Diptera of unknown origin) was significantly lower in grass/sedge (BC) and shrub (PC) study reaches than was observed in the tree (RR) study reach (38%, 45% and 63%, respectively). Observed differences in crossecosystem prey contributions among study reaches resulted in part from variation in the proportion of aquatic winged adults and Diptera of unknown origin in juvenile salmon diet. Aquatic winged adult and Diptera of unknown origin together contributed 22%, 21% and 41%, respectively, to juvenile salmon diet in grass/sedge, shrub and tree study reaches. The immature form of these taxa represented the dominant aquatic resident taxa consumed by juvenile salmon in each of the study reaches; however, the proportion of aquatic winged adult invertebrates that returned to the stream, and were consumed by juvenile salmon was significantly lower in the grass/sedge (5%) vs. tree (15%) reaches and Diptera of unknown origin provided a greater proportion of prey contributions to diet for juvenile salmon in the tree study reach (26%) as compared to the grass/sedge (17%) and shrub reaches (12%). The observed differences between study reaches for terrestrial invertebrate contributions to juvenile salmon diet were not statistically significant. High variability in the size and frequency of terrestrial invertebrate prey observed in juvenile salmon diet contributed to the lack of significance in comparisons between study reaches. Understory vegetation within the tree reach was similar to the shrub reach and similar terrestrial invertebrate contributions were observed (21% and 22% respectively). The percent contribution of terrestrial invertebrates in the grass/sedge study reach was generally lower (15%). Prior studies have found greater terrestrial invertebrate in-fall to streams in areas of deciduous riparian shrub and tree habitat as compared to coniferous forest (Wipfli 1997;Allan et al. 2003;Inoue et al. 2013), pasture (Edwards and Huryn 1995) and grasslands (Kawaguchi and Nakano 2001), consequently greater observed terrestrial prey contributions from tree and shrub vegetation types as compared to grass/sedge would be expected.
Observed consumption of aquatic winged adult prey varied seasonally and aquatic winged adults represented approximately 30% of spring invertebrate consumption in the tree study reach. Seasonal consumption of Diptera of unknown origin (30%) was observed during spring (grass/sedge and tree study reaches) and fall (tree study reach). Increased consumption of terrestrial, aquatic winged adult and Diptera of unknown origin was generally associated with decreases in aquatic invertebrate consumption in grass/sedge and tree study reaches and most likely were the result of emergence or hatching events. Opportunities to measure invertebrate consumption associated with emergence or hatching events were constrained by the gastric evacuation rates of juvenile salmon (Allan et al. 2003;Sweka et al. 2004;) and frequency of sampling.
Our results suggest that increased dispersal of aquatic winged adult insects and Diptera of unknown origin may reduce in-fall of those taxa as prey in open riparian habitats. Abundance of aerial insects has been found to be high in areas of open riparian habitats such as grass or clear-cut (Hetrick et al. 1998;Albertson et al. 2018); however, there is no strong correlation between aerial abundance and in-fall to streams (Hetrick et al. 1998;Inoue et al. 2013). Open areas expose insects to greater wind velocities that facilitate downwind drift and dispersal of weak-flying insects while taller riparian vegetation such as shrubs and trees can serve as habitat (Kuusela and Huusko 1996;Petersen et al. 1999;Delettre and Morvan 2000) that collects flying insects at stream margins (Helle and Muona 1985;Whitaker et al. 2000), and increases opportunities for in-fall (Cadwallder et al. 1980;Baxter et al. 2005;Saunders and Fausch 2007).
Within the Kenai watershed, late seral stage boreal forests are predicted to be replaced by persistent grasslands (Wolken et al. 2011;Hansen et al. 2016) as mean annual temperatures continue to increase, total annual precipitation decreases (Wolken et al. 2011;Bauret and Stuefer 2013;Schoen et al. 2017), and the frequency and intensity of disturbance events such as wildfire and insect outbreaks increase (Lynch et al. 2002;Klein et al. 2005;Berg et al. 2006). In addition, human development along the lower Kenai River is resulting in increased clearing within riparian areas (Schoen et al. 2017). While change in riparian vegetation from tree/shrub to grass/sedge can be expected to reduce contributions of cross-ecosystem invertebrates to juvenile salmon diet, research has shown that fish response to changing climate and habitat can be complex (Wainwright and Weitkamp 2013;Naman et al. 2018). Localized reduction of cross-ecosystem contributions to streams may result in reduced fish biomass in stream reaches due to movement of fish to more prey-rich areas (Kawaguchi and Nakano 2001;Kawaguchi et al. 2003) as well as increased predation of resident aquatic invertebrates (Nakano et al. 1999). In freshwater habitats where productivity is low, juvenile salmon are often dependent on the movement of prey resources from the terrestrial environment to streams (Allen 1951;Huryn 1996;Richardson et al. 2010). Widespread reduction in the availability of cross-ecosystem prey for resident fish may therefore result in reduced fish growth, fitness and density when and where fish are food-limited (Sweka and Hartman 2008;Fischer et al. 2010;Inoue et al. 2013). Alternatively, a shift to more open riparian canopy could raise stream temperatures, increasing in-stream production of autotrophic aquatic invertebrate prey and providing thermal habitats that promote fish growth and influence fish distribution (Hartman and Scrivener 1990;Tschaplinski and Pike 2017). In addition, increased predation risk due to lack of cover can lead to a shift in foraging behavior and diet that can result in altered growth for stream fishes (Dill and Fraser 1984;Reinhardt 1999;Allouche and Gaudin 2001).
While riparian areas generally represent a small portion of a total watershed, they can have a large effect on stream fish (Naiman et al. 2005;Wipfli and Baxter 2010;Wipfli and Richardson 2015). Our results document decreased aquatic winged adult and cross-ecosystem prey consumption that is correlated with more open riparian canopy. These results allow us to better understand how environmental or anthropogenic caused change in riparian vegetation might affect invertebrate prey contributions to juvenile salmon diet within the Kenai watershed. Monitoring change in riparian habitat using remote sensing data such as LiDAR can help us identify the extent of those changes, and management of riparian buffers along streams is likely to help mitigate potential impacts of those changes.