Bioenergetic responses of a stream food web to habitat restoration: interactions between Brown trout and invertebrate prey resources

Stream habitat restoration has the potential to rehabilitate degraded stream communities, but the bioenergetic impact of restoration on prey resources and utilization by salmonids has been understudied. We measured the responses of Brown trout (Salmo trutta) populations, aquatic and terrestrial prey resources, and prey resource utilization by trout, before (2008–2014) and after (2015–2020) large‐scale habitat restoration in the Upper Arkansas River, a previously listed US EPA superfund site located in the western United States. We observed significant increases in Brown trout populations following restoration, as well as complex alterations to invertebrate prey resources that generally resulted in increased benthic biomass but reductions in adult aquatic insects and terrestrial invertebrates. Population‐level metabolic demand by trout increased 25% after restoration, and prey utilization shifted with reduced consumption of adult aquatic insects and greater consumption of terrestrial invertebrates, relative to their availability. We observed reduced biomass of prey resources consumed by trout with increasing trout population size, and while the total number of invertebrates consumed increased after restoration, actual invertebrate biomass consumed decreased. Our results suggest that restoration was effective in increasing targeted Brown trout populations, but prey resources were altered by physical changes in habitat and increased resource utilization by trout. Habitat restoration projects that are primarily designed to increase trout populations may benefit by focusing on improving prey resources and utilization to support the bioenergetic needs of the restored fishery.


Introduction
Montane stream communities can respond rapidly to water quality remediation in mining-impacted watersheds Kotalik et al. 2021), but many of these streams also have significant habitat degradation from historical mineral extraction and associated land use practices. Channel widening, bank-erosion, reduced riparian vegetation, and degraded instream habitat are often observed in mined watersheds (Mebane 2001;DeNicola & Stapleton 2002;Maret & Mac-Coy 2002). While remediation to improve water quality is often prioritized because of the well-established relationship between elevated metal concentrations and biological impairment, physical habitat restoration can offer additional benefits to stream communities recovering from mining impacts. For many streams in the western United States, restoration is targeted to improving populations of salmonids, and "success" is commonly quantified by measuring changes in habitat quality or suitability and responses of the target species (Roni et al. 2008). However, changes in habitat can affect stream invertebrate communities (Miller et al. 2010), and the relationship between trout and aquatic and terrestrial prey resources are rarely considered, despite the bioenergetic significance of invertebrates in sustaining and increasing trout populations (Baxter et al. 2005;McCarthy et al. 2009).
The "Field of Dreams" hypothesis proposes that because of the well-established relationship between habitat heterogeneity and species diversity (Lepori et al. 2005), habitat improvements can enhance ecological recovery in disturbed watersheds (Palmer et al. 1997). In a global review of stream rehabilitation projects (n = 345), Roni et al. (2008) found that restoring stream habitat is effective in increasing local fish abundance under many circumstances. However, the responses of benthic macroinvertebrates (i.e., prey resource) to habitat restoration is equivocal, with some studies reporting increases in species richness and abundance (Moerke et al. 2004;Nakano & Nakamura 2006;Nuttle et al. 2017), while other studies show negligible effects (Friberg et al. 1998;Palmer et al. 2010). Variation in macroinvertebrate responses among habitat restoration projects has been attributed to numerous factors, including the high variability of metrics (Miller et al. 2010), background physical and chemical disturbance (Lake et al. 2007), colonization attributes (Mackay 1992), and watershed-scale processes (Larson et al. 2001). In addition, monitoring studies for restoration effectiveness often fail to estimate responses of macroinvertebrates, rather focusing on physical habitat metrics and responses of target fish species (Bernhardt et al. 2005;Wohl et al. 2005). Given that the primary goal for many habitat restoration projects is to improve fish populations (Wohl et al. 2005;Roni et al. 2008), uncertainty in the responses of macroinvertebrates, a critical prey resource in streams, complicates our ability to establish and evaluate restoration goals.
Habitat suitability models for drift-feeding salmonids often rely on correlative habitat associations based on fish abundance and habitat characteristics applied at different spatial scales (e.g. microhabitats, reach scale) (Raleigh et al. 1986;Rodriguez 1995;Rosenfeld 2003). These models are exclusively based on physical habitat parameters that do not account for the availability and production of prey resources. Prey resource production may be directly affected by restoration (Palmer et al. 2010) and increased fish populations may potentially alter prey availability through competitive interactions and resource suppression (Fausch 1988;Keeley 2001). Bioenergetic habitat models have the potential to integrate invertebrate production with habitat-specific salmonid growth rates (Fausch 1984;Railsback & Rose 1999), but these models are most often used to inform habitat restoration treatments prior to installation.
The Upper Arkansas River (UAR) is a fourth order Rocky Mountain stream located 110 km southwest of Denver, Colorado. Historical mining for gold and silver in the watershed in the late nineteenth and early twentieth century resulted in significant water quality impacts and degraded stream habitat. Remediation by the US Environmental Protection Agency (EPA) began in the early 1990's with installation of water treatment facilities, in situ treatment of fluvial tailings, and removal of 150,000 m 3 of tailings from the watershed. Significant improvements in water quality, macroinvertebrate communities, and Brown trout populations were observed after completion of these remediation activities in 2000 (Clements et al. 2010). Following these improvements, a large-scale habitat restoration project was conducted in the UAR from 2013 to 2014. The project goals for biological recovery included increases in riparian vegetation, prey resources, and Brown trout (Salmo trutta) population size (Stratus 2010). Richer et al. (2022) conducted a before-after control-impact (BACI) analysis of stream restoration on trout populations in the UAR that included this present study's river reach (4 km) in addition to six other stations located upstream and downstream. The authors found that changes in trout biomass were significantly greater at treated stations relative to controls, and that trout density and relative weight increased across all sites regardless of treatment type. However, changes at individual sites were less evident, with only one treatment site showing significant increases in trout population metrics when analyzed independently. In addition to these findings, Wolff et al. (2022) reported that restoration treatments in the UAR had limited effects on invertebrate prey resources and hypothesized that increased predation by Brown trout may have dampened invertebrate responses. In this paper, we focus on the bioenergetic impacts of habitat restoration on invertebrate prey resources and Brown trout populations in a 4 km habitat restoration reach of the UAR.
To characterize bioenergetic responses to restoration, the primary objectives for this research were to: (1) quantify biomass of aquatic and terrestrial prey resources before and after habitat restoration; (2) evaluate responses of Brown trout population size to habitat restoration; and (3) determine Brown trout metabolic demand and prey utilization before and after habitat restoration. We annually sampled prey resources (i.e., benthic macroinvertebrate standing stock, emerging adult aquatic insects, and streamside invertebrate inputs), and linked these to spatially and temporally concurrent trout population surveys and diet sampling. Field data were collected before (2008)(2009)(2010)(2011)(2012)(2013)(2014) and after completion of restoration (2015)(2016)(2017)(2018)(2019)(2020). Because the locations of fish population and invertebrate monitoring sites overlapped designated treatment and control reaches, we were unable to employ a BACI statistical approach for this analysis. However, we draw on multiple lines of evidence obtained from spatially (four stations) and temporally (13 years) replicated biotic (i.e., aquatic and terrestrial invertebrates, trout) and abiotic (i.e., water chemistry) assessments before and after restoration to test for restoration effects. We test the hypothesis that stream habitat restoration increases invertebrate prey resources and trout population size.

Study Site
The UAR is a snowmelt driven stream with peak discharge typically occurring in mid-June (USGS gage: 07081200), and baseflow conditions are generally reached by October. Mean annual Restoration Ecology July 2023 discharge during the study period (2008-2020) ranged from 0.95 to 3.32 m 3 /s. Most of the 29.5 cm annual average precipitation is in the form of snow, but monsoon events can increase stream discharge in mid-to late-summer. Substrate at all stations is dominated by medium to large cobble. Streamside riparian vegetation is a mixture of woody plants including willow (Salix spp.), sagebrush (Artemsia tridentata) and currant (Ribes laxiflorum), various grasses, sedges (Carex spp.), and rushes (Juncus spp.) (Cubley et al. 2022).
Restoration activities among stations included the installation of instream structures (e.g. log-vanes, boulder clusters, large woody debris) to increase habitat heterogeneity, bank stabilization to reduce erosion, channel narrowing to increase sediment transport, and planting of streamside riparian vegetation (Richer et al. 2019;Richer et al. 2022). Four sampling stations (AR4C, AR4E, AR4G, AR4H) were used to evaluate the effects of habitat restoration on prey resource availability, Brown trout populations, and prey utilization. These stations were located along a 4 km reach, with the most upstream station (AR4C) located 5 km downstream of California Gulch, the historical source of metals contamination (Fig. S1). The locations of these stations were selected based on their proximity to active habitat restoration treatments and where other stream monitoring research was ongoing. All monitoring stations were located within the extent of a larger 17.7 km restoration project that spanned both private and public lands. Detailed descriptions for the restoration approach and treatments, including as-built drawings, are available in Richer et al. (2017Richer et al. ( , 2019Richer et al. ( , 2022. Routine water quality characteristics (conductivity, pH, water temperature, water hardness) were measured annually (late-summer) among all stations throughout the study period. Water samples (15 mL) collected for analysis of trace metals (Cu, Cd, and Zn) were filtered (0.45 μm) and preserved with analytical grade nitric acid (three to five drops per 15 mL sample). Metal concentrations were determined by (flame or furnace) atomic adsorption spectrophotometry or inductively coupled plasma spectroscopy. To estimate metal exposure risk among the three metals of concern, we calculated cumulative criterion units (CCU) using US EPA hardnessadjusted chronic aquatic life criteria. CCU were calculated as CCU = P M i /C i , where M i is the measured concentration of each metal, and C i is the respective chronic aquatic life criterion value that is intended to protect freshwater aquatic life from adverse effects.

Benthic Macroinvertebrates, Emergence, and Aquatic and Terrestrial Streamside Inputs
To estimate the biomass of benthic standing stock, we sampled stream benthic macroinvertebrates annually during late-summer (August through early October) from 2010 to 2019. Quantitative benthic samples were collected at all stations using a modified Hess sampler (0.1 m 2 ; 350 μm mesh net). Replicate (n = 5) samples were gathered at each station, with specific sampling locations chosen based on the presence of riffle or run habitat, water depth between 0.1 and 0.25 (m), and cobble-sized substrate that was representative at all stations. Substrate contained within the Hess sampler was scrubbed of all detrital material and macroinvertebrates. Samples were rinsed through a 350 μm mesh sieve in the field and retained organisms and detrital material were preserved in 80% ethanol. In the laboratory, benthic macroinvertebrates were sorted from detrital material and subsampled using a 300-count protocol (Moulton et al. 2000). Samples were identified to the lowest practical level of taxonomic resolution (genus for most aquatic insects; family for chironomids [midge]) using regional (Ward et al. 1992) and North America keys (Merritt et al. 1996). Total benthic macroinvertebrate biomass was estimated for each replicate, and because replicates were subsampled, total sample biomass was calculated based on the number of grids sorted.
Emerging adult aquatic insects and streamside invertebrate inputs were sampled concurrently with benthic macroinvertebrates. Sampling was conducted in summer because streamside inputs of invertebrates and emergence of adult aquatic insects are generally greatest during this period (Saunders & Fausch 2012). Emerging aquatic adults were sampled using 0.3 m 2 floating emergence traps (Cadmus et al. 2016) deployed in scour pools, with three replicate nets at each station sampled for two consecutive 24-hour periods. Adult insects were preserved in 80% ethanol in the field, identified in the laboratory (order for all aquatic insects except for chironomids that were identified to family), and separated into the following groups for biomass measurements: Ephemeroptera (mayfly), Trichoptera (caddisfly), Chironomidae, and "other." The grouping "other" was comprised of a diversity of non-chironomid Diptera (true flies) and Plecoptera (stonefly) taxa that were combined to obtain sufficient weights for biomass estimates.
Streamside inputs of adult aquatic and terrestrial invertebrates were estimated using pan traps (100 cm Â 41 cm Â 15 cm). Three replicate pan traps were deployed at each station for two consecutive 24-hour periods. To quantify inputs of aquatic and terrestrial invertebrates from different streamside habitat types, three dominant habitats were sampled: grasses, gravel bars, and willows. Pan traps were placed in these habitats based on the homogenous presence of these streamside habitat types for a least 5 m upstream and downstream. Each pan trap was filled with approximately 4 L of stream water and 5 mL of unscented biodegradable surfactant to reduce surface tension. Aquatic and terrestrial invertebrates were removed from the pan traps using small aquarium nets and preserved in 80% ethanol. In the laboratory, invertebrates were identified to order-level taxonomic resolution and separated into the following groups for biomass measurements: Ephemeroptera-Plecoptera, Trichoptera, aquatic Diptera, and terrestrial invertebrates. Invertebrates that spent any part of their life cycle in the aquatic environment were classified as "aquatic," while invertebrates that completed their entire life cycle in the terrestrial environment were classified "terrestrial." Biomass groupings were chosen a priori to ensure that sufficient dry weights were available to make comparisons among groups. For all samples, biomass was converted to mg/m 2 , and because emergence and aquatic and terrestrial streamside inputs were sampled twice over two consecutive 24-hour periods, an average between days was used.

Brown trout Population Estimates and Diet Collection
Brown trout population surveys and diet collection were conducted within 1-2 weeks of invertebrate sampling. Trout were sampled using bank electrofishing with a five-electrode array, and abundance estimates were based on a two-pass removal method (Seber & Le Cren 1967). Brown trout total length (mm) and weight (g) were measured for each fish. Estimates of trout density (no./ha) and biomass (kg/ha) were used for trout equal or greater than 1 year of age (≥100 mm, identified by length-frequency histograms; Fig. S2). Electrofishing survey extents for each station are shown in Figure S3, with estimates of trout density and biomass per hectare extrapolated based on the respective electrofishing survey stream area. Diet samples (n = 17-25 fish per station) were collected using gastric lavage of living fish (Waters et al. 2004;Saunders & Fausch 2007) from 2012 to 2016, and in 2018. Trout between 120 and 350 mm total length were selected for diet analyses because: (1) this size range of trout predominately consume invertebrates in riverine environments (Sanchez-Hernandez 2020); (2) smaller fish were difficult to sample without causing harm; and (3) larger fish can become piscivorous (Budy & Gaeta 2018) and population estimates of trout fry (i.e., prey) were unavailable, preventing estimates of fry prey utilization. After sampling, trout were placed in a net pen to recover and released back into the river. Stomach contents were rinsed through a 350-μm sieve and preserved in 80% ethanol in the field. In the laboratory, stomach contents (including partially digested invertebrates) for each trout were identified to the same taxonomic groupings as described for aquatic and terrestrial streamside inputs and consolidated into three groups for biomass measurements: aquatic larvae, aquatic adults, and terrestrial invertebrates.
Biomass of invertebrates (benthic macroinvertebrates, emerging aquatic adults, aquatic and terrestrial invertebrates in pan traps, and trout diets) was measured by drying samples for 48 hours at 60 C. Detrital material and rock cases of larval caddisflies (e.g. Glossosomatidae and Oecetis spp.) were removed during laboratory sorting and identification. We did not separate diet samples based on extent of prey digestion and assumed that rates of digestion were similar among trout sampled.

Brown trout Prey Utilization and Bioenergetics
The utilization of prey resources (i.e. aquatic larvae, aquatic adults, and terrestrial invertebrates) by Brown trout was calculated as the ratio of prey biomass consumed to the prey biomass available. To ensure that estimates of prey resources were representative of reach scale availability, biomass of benthic macroinvertebrates and emerging aquatic adults was converted to per hectare wetted stream area. Biomass of aquatic and terrestrial inputs were estimated using linear streamside distances extrapolated to per hectare stream area (i.e. streamside distance = 10,000 m 2 /wetted channel width in meters) based on the average wetted channel widths for each station (10.2-16.5 m; Richer et al. 2019). Average prey biomass consumed per trout at each station was multiplied by the measured trout density to estimate population-level prey consumption per hectare. Because the minimum length cutoffs for trout diet collection (≥120 mm) was slightly greater than for adult trout population sampling (≥100 mm), estimates of prey consumed per hectare may be slightly inflated. Importantly, diet sampling coincided with stream invertebrate sampling, and we assumed that aquatic and terrestrial prey were representative of their availability when trout were sampled. The purpose for these analyses was to determine if habitat restoration affected the utilization of aquatic and terrestrial prey by trout, and to test for a relationship between consumed prey biomass and trout population size.
We estimated energy density of aquatic and terrestrial streamside inputs and the metabolic energy demand of trout populations before and after habitat restoration. The goal for these bioenergetic estimates was to compare energy of prey resources from streamside inputs to the energy required to sustain trout populations among the sampling stations. Importantly, this bioenergetic comparison was possible because stream invertebrate and fish sampling were conducted within the same 1-to-2-week period annually. Aquatic and terrestrial energy content was estimated from published proportion dry mass (pDM) and energy density data. For aquatic adults, we used pDM of 0.18 for pooled Ephemeroptera, Plecoptera, Trichoptera, and Diptera adult biomass (McCarthy et al. 2009). Proportion dry mass for terrestrial invertebrates collected in our study ranged from 0.22 to 0.58 pDM (Table S1). Because terrestrial invertebrate biomass was pooled across taxa, we conservatively estimated 0.30 pDM for these samples. We used James et al. (2012) generalized model for estimating energy density of invertebrates, which estimated energy density for aquatic adults and terrestrial invertebrates at 3,958 and 6,713 J/g wet mass, respectively.
Total metabolic demand required to sustain Brown trout populations was estimated for each station and sampling event from 2008 to 2018 using Fish Bioenergetics 4.0 (Deslauriers et al. 2017). Hourly temperature data were averaged over a 2-week period bracketing each fish sampling event and used to develop the bioenergetic models. All fish sampling events reported in this study took place within a 10-day period, from 9 to 19 August. Average water temperatures associated with fish sampling events in the month of August ranged from 11.1 to 12.6 C (Table S2). Fish weights were assumed to be constant because trout growth data were unavailable. Approximate age classes based on total length were obtained from previous studies aging Brown trout in the UAR (Nehring & Policky 2003), and these were used to calculate average weight of each age class of fish observed at each station-sampling event (Age-1 through Age-8+). Fish below 120 mm in length were removed from the analysis due to large variation in capture efficiency of Age-0 fish; therefore, our estimates of total metabolic demand are conservative.
Estimates of respiration (g O 2 g fish À1 day À1 ) were obtained from the species-specific activity multiplier for stream dwelling Brown trout (Dieterman et al. 2004). Respiration estimates were calculated based on the average weight of each age class at each station-sampling event and summed for the number of individuals of each size class represented in the observed trout population. Respiration rates were converted to Joules (i.e., energy) using an oxycaloric conversion factor of 13.56 kJ/g oxygen consumed (Elliott & Davison 1975). Details on the bioenergetic models (i.e. consumption, respiration, egestion, and excretion), and associated parameter values, are summarized and reported by Deslauriers et al. (2017). These estimates of trout metabolic Restoration Ecology July 2023 demand should be considered a "snapshot" of expected demand that corresponds with our assessment of available aquatic and terrestrial invertebrate prey energy.

Statistical Analyses
All statistical analyses were conducted using the R statistical computing language (R Development Core Team 2013; v4.2.1). Data were log-transformed to meet the assumptions of parametric statistics and to improve fit for all statistical models. Generalized linear mixed models (GLMM) ("lmer" function: package "lme4"; family = "Gaussian"; Bates et al. 2015) were used to test for statistically significant differences in biomass of benthic macroinvertebrates, emerging aquatic adults, streamside aquatic and terrestrial invertebrate inputs, and Brown trout population metrics and diet. We used the Satterthwaite's method to estimate p values for all predictor variables (package "lmerTest"; Kuznetsova et al. 2017) and defined statistical significance for all tests as p < 0.10 to protect against Type II error. Conditional variance (R 2 ) for the GLMMs was calculated using the "MuMln" package (Nakagawa & Schielzeth 2013).
For all GLMMs, fixed effects were habitat restoration (before vs. after) and mean annual discharge (m 3 /s) (treated as continuous). Sampling year (treated as categorical) and station were included as random effects in all models. For biomass of aquatic and terrestrial streamside inputs, differences among habitats (i.e., gravel, grass, willow) were treated as fixed effects and used in the full models. Because Brown trout body size can influence feeding habits, we included trout length (categorically summarized as 120-199, 200-299, and 300-350 mm) as a fixed effect covariate in the trout diet models, and we tested for a significant length Â restoration interaction to determine if effects of restoration on diet varied by fish length. The effects of metals (as CCU) on all response metrics were tested separately using simple linear regression. Finally, to test the hypothesis that the biomass of prey items in the diet of Brown trout was influenced by Brown trout population density and biomass, we used simple linear regression.

Results
Physiochemical characteristics were relatively consistent throughout the study period, with marginal increases in average water hardness, alkalinity, pH, and conductivity from before to after restoration (Table S3). Average annual concentrations of dissolved zinc, copper, and cadmium were below the US EPA hardness-adjusted chronic water quality criteria for aquatic life protection before and after restoration; however, small decreases in concentrations of these metals were observed after restoration was completed, resulting in a lower CCU value from before (1.52 AE 0.17) to after (0.91 AE 0.18) restoration. CCUs were not significantly associated with the responses of any prey resources ( p ≥ 0.201) (Table S4).
Variation in the biomass of benthic macroinvertebrates (F 1,35.6 = 10.50; p = 0.002) and the streamside inputs of aquatic invertebrates (F 1,6.0 = 4.12; p = 0.092) was significantly influenced by restoration (Tables S5 & S6; Fig. 1). The biomass of all aquatic and terrestrial prey resources showed relatively little spatial variation and was remarkably consistent among stations (Table S7). Annual average discharge was an important predictor of benthic macroinvertebrate biomass (F 1,35.7 = 4.91; p = 0.033) and adult emergence (F 1,4.4 = 5.76; p = 0.067), with higher discharge resulting in lower biomass for benthic macroinvertebrates, but higher emerging adult biomass.
The significantly higher biomass of benthic macroinvertebrates after habitat restoration was driven primarily by a large increase immediately after restoration followed by a gradual return to pre-treatment conditions (Fig. 1A). In contrast, we observed a 51% decrease in the streamside inputs of aquatic adults (Fig. 1B) after restoration, which was consistent across taxonomic groups. The relatively high biomass of streamside inputs of aquatic adults observed in 2013 was from large-bodied aquatic dipterans (e.g. Tipula spp.) and trichopterans (Fig. S4). The streamside inputs of terrestrial invertebrates varied among years and was 22% lower after restoration; however, the difference between pre-and post-restoration was not significant (F 1,6.1 = 0.81; p = 0.402). The biomass of emerging adults followed a similar trend as for streamside inputs and was generally lower after restoration (Fig. 1C). The GLMMs for adult emergence and streamside inputs of aquatic adults explained 45 and 81% of the conditional variation, respectively.
In addition to lower streamside inputs after restoration, we observed highly significant differences in the inputs of aquatic (F 2,88.4 = 42.39; p ≤ 0.001) and terrestrial (F 2,88.6 = 34.43; p ≤ 0.001) invertebrates among habitat types (Table S5; Fig. 2). Streamside inputs of terrestrial invertebrates were consistently lower compared to aquatic invertebrates, but patterns among habitat types were generally similar between these groups. Grass and willow habitats contributed much higher biomass compared to gravel habitats, with vegetated areas providing approximately twofold to threefold greater total biomass of terrestrial prey resources (Table S8).
The total biomass of prey in Brown trout diets did not significantly change after restoration (F 1,3.0 = 0.01; p = 0.934), but moderate reductions in biomass of aquatic larvae (28%) and adults (9%) were observed. The restoration Â trout length interaction term was not significant for any of the prey group metrics assessed ( p ≥ 0.55); therefore, the interaction term was not included in the Brown trout diet GLMMs. Aquatic-sourced invertebrates were the dominant prey in the diets of Brown trout, both before and after restoration, with aquatic larvae contributing 82-89% of the total biomass. Average biomass of terrestrial invertebrates in Brown trout diets increased by 39% after restoration, especially in larger fish (Fig. S5), but this difference was not significant (F 1,3.1 = 0.22; p = 0.669).
Results of regression analyses showed that the average total biomass of prey consumed by individual trout significantly decreased with increasing trout population density and biomass  ( Fig. 4). A weak relationship between total biomass of invertebrates consumed and CCUs was also observed (F 1,70 = 2.61; p = 0.110) (Table S4). Among the three prey groups consumed by trout, reduced consumption of terrestrial invertebrates with increasing CCUs was highly significant (F 1,70 = 11.29; p < 0.001), but consumption of aquatic larvae (F 1,70 = 1.13; p = 0.291) and adults (F 1,70 = 1.55; p = 0.217) was not significantly influenced by metals.
Based on the availability of aquatic and terrestrial prey, we estimated utilization of these resources by Brown trout before and after restoration. Relative to availability, Brown trout consumption of terrestrial resources was much greater than for aquatic larvae and adults, and this difference increased following restoration (Fig. 5). After restoration, utilization of terrestrial prey resources relative to availability increased by 62% and was approximately 17-fold and 11-fold greater than for aquatic larvae and adults, respectively. Notably, while the biomass of terrestrial prey consumed by trout increased after restoration, we observed reduced streamside inputs of terrestrial invertebrates.
The estimated metabolic demand of Brown trout populations significantly increased (F 1,29 = 3.81; p = 0.061) by 25%, from 3,206 kJ ha À1 day À1 (AE281 SE) to 4,007 kJ ha À1 day À1 (AE298 SE) after restoration (Fig. 6). Energy from terrestrial prey was considerably lower than from aquatic prey before and after restoration, but still provided a substantial amount of energy to trout, accounting for approximately half of their metabolic requirement. Before habitat restoration, available energy from streamside inputs of aquatic prey were approximately 2.5-fold greater than the energy required to sustain Brown trout populations. After restoration, energy available from aquatic prey was reduced by 52% and was similar to the energy required for trout metabolic demand. Although total available energy from streamside inputs were reduced by 46% after restoration (5,804 kJ ha À1 day À1 ), these inputs still exceeded the amount Figure 4. Relationship of Brown trout (Salmo trutta) average prey consumption to trout density and biomass. Statistical results were obtained from simple linear regression models. Dotted lines are 95% confidence intervals. Fish diet samples were collected from four stations located in the habitat restoration reach of the Upper Arkansas River, before and after restoration. Figure 5. Brown trout (Salmo trutta) prey utilization of aquatic larvae, aquatic adults, and terrestrial invertebrates. Prey utilization was calculated as the ratio of available prey resources to the consumption of prey resources by trout populations, per stream hectare, respectively. Fish diet samples and measurements of prey availability were collected before and after habitat restoration. Figure 6. Comparison of mean (AE SE) energy of aquatic and terrestrial streamside prey inputs in the Upper Arkansas River compared to the mean (AESE) metabolic energy demand of trout populations, before and after habitat restoration. of energy required to sustain Brown trout populations in the UAR in late-summer.

Discussion
Concurrent evaluation of Brown trout populations, prey resources, and prey utilization suggests that stream habitat restoration was effective in increasing Brown trout populations, but restoration directly and indirectly altered prey resources and prey utilization by trout. Prey resources are responsible for sustaining and growing trout populations (Railsback & Rose 1999;Hayes et al. 2000), but with few exceptions (Bellmore et al. 2017), most restoration monitoring programs do not address the important linkage between prey resources and utilization. Traditional biological metrics (e.g. population surveys, invertebrate composition) are informative and serve as a proxy for stream bioenergetics, but directly quantifying the biomass and energy of prey resources, and relating these to trout utilization, provides more quantitative estimates of changes in stream communities resulting from habitat restoration.
Recovery of benthic communities following instream disturbances is influenced by factors such as the magnitude and duration of the disturbance, the return to pre-disturbance habitat conditions, and the proximity to communities that supply organisms for recolonization (Yount & Niemi 1990;Mackay 1992). For this restoration project, heavy machinery was used to move material and install instream structures, which caused short-term physical disturbance to benthic habitat, as well as elevated turbidity and sediment deposition. These direct and indirect effects of construction may have caused mortality and increased emigration of organisms through behavioral drift (Jones et al. 2012). Given the relatively short duration of this disturbance, proximity to a robust upstream colonization source, and potential for improved benthic habitat through increased sediment transport after restoration, biomass of macroinvertebrates quickly recovered. Numerous field studies have documented the importance of natural physical disturbance in maintaining diversity of benthic communities (Townsend et al. 1987;Poff et al. 1997), and previous research in this system showed that abundance of macroinvertebrates quickly returned to prerestoration levels (Wolff et al. 2022).
We predicted that increases in habitat heterogeneity (Richer et al. 2019) and riparian vegetation (Cubley et al. 2022) would increase prey resources, but we observed reductions in biomass of emerging adult insects and terrestrial prey resources after restoration. One hypothesis is that the significant increase in Brown trout populations, which resulted in increased population-level metabolic demand, increased overall prey consumption. Reductions in aquatic adults may have been the result of increased predation during emergence (Henschel et al. 2001) or greater predation on benthic larvae prior to emergence that prevented larvae from maturing to their adult life stage (Knight et al. 2005;Wesner 2010;Wesner 2019). In addition, while the biomass of benthic prey significantly increased after restoration, this was primarily driven by an increase immediately after restoration was completed, which was followed by a gradual decline in benthic biomass to pre-restoration levels by the end of our study. This may indeed suggest that benthic prey immediately responded to restoration but were suppressed by increasing trout populations. These potential ecological feedbacks between prey resources and trout indicate that biological responses to restoration are interdependent and that accounting for these relationships is an important consideration when setting restoration goals.
Our observation of reduced biomass of prey consumed by trout was quite different from the total number of individual prey consumed, which increased significantly after restoration (Wolff et al. 2022). Trout foraging models derived from optimal foraging theory (Pyke 1984) have demonstrated that energy expenditure is balanced with caloric energy gain from prey resources, and that trout adjust their foraging strategies to maximize net energy gain (Bachman 1984;Fausch 1984). After restoration, we observed significant decreases in biomass of aquatic adults but increases in the biomass of benthic macroinvertebrates. Because aquatic adults are developmentally mature and have reached their maximal body size, the number of adults required to meet trout energetic demands is likely to be lower than the number of immature aquatic insects that must be consumed. We hypothesize that alterations in prey availability influenced trout foraging strategy, whereby the total number of prey items consumed increased, but prey quality decreased due to greater reliance on smaller-bodied, immature benthic larvae. These findings have important bioenergetic implications for trout populations because it is less efficient for fish to forage on many small prey items versus fewer large prey (Bannon & Ringler 1986).
Despite the overall low biomass of terrestrial invertebrates in trout diets, they were disproportionately utilized by trout compared to available aquatic resources, a finding supported by previous field studies. Baxter et al. (2005) showed that salmonids preferentially consumed terrestrial invertebrates because of their larger size and higher food quality (i.e., energy density) relative to in situ aquatic resources. After restoration, there was also a notable increase in consumption of terrestrial prey by larger Brown trout (≥300 mm). This response may be due to larger fish excluding smaller fish from optimal foraging habitats where drifting terrestrial prey are more available to larger fish, whereas smaller fish are more restricted to benthic foraging (Dineen et al. 2007). Fausch (1984 proposed that microhabitat selection and optimal foraging position were related to intraspecific hierarchies, with larger fish holding optimal stream positions over smaller individuals. We hypothesize that improved instream habitat and foraging positions (Richer et al. 2019) increased terrestrial prey availability that was usurped by larger and competitively superior trout.
Quantitatively linking prey utilization with availability requires that researchers recognize spatial and seasonal variation in prey resources and trout populations (Marcarelli et al. 2020). For example, we observed increases in Brown trout populations and altered prey resources after restoration, but this variation was not consistent with specific restoration treatments implemented at individual stations. Richer et al. (2022) also found that changes in trout populations at individual stations were less evident between treated and untreated stations. Trout are highly mobile fish, and while they can be seasonally sedentary Restoration Ecology July 2023 (i.e., during low flow conditions in winter), they can also travel great distances (Meyers et al. 1992), which may confound sitespecific assessments of restoration treatments. In addition, the type, quality, and availability of prey resources varies seasonally; therefore, the utilization of prey resources by trout also varies seasonally (Nakano et al. 1999). For this study, annual biomonitoring occurred in late-summer, in part, because we wanted to measure available terrestrial resources, but also because of logistical constraints (e.g. high flow conditions, access to sampling stations) associated with sampling a subalpine stream. While aquatic adults and terrestrial invertebrates are primarily available to trout in summer, metabolic demand is also the highest due to elevated water temperatures, and therefore greater prey consumption is required. In contrast, prey utilization is considerably different in winter, when trout are strictly consuming benthic larvae or other fish, but metabolic demand is much lower due to low water temperatures (Cunjak et al. 1987). Because the sampling regime for this research occurred in late-summer, the patterns of prey utilization reported are temporally specific and would likely be very different during other seasons (e.g. winter).
Water quality remediation prior to the implementation of habitat restoration was critical for restoring macroinvertebrate communities and Brown trout populations in the UAR. Elevated metal concentrations have direct effects to trout (Farag et al. 1995), and indirect effects via reduced quantity and quality of aquatic prey resources (Clements & Rees 1997). While US EPA aquatic life criteria exceedances for Zn and Cd are occasionally observed during spring run-off (Clements et al. 2010), water quality in the UAR has dramatically improved from historical conditions. Throughout our restoration assessment period (2008-2020), CCUs approximated values considered protective of aquatic life . A marginal reduction in metal concentrations was observed after restoration that may have been associated with restoration activities that treated in situ fluvial tailings, reduced erosion of fluvial tailings due to improved bank stabilization, or a continuation of the gradual decline in metals concentrations observed in the UAR watershed since remediation was completed in the late 1990s. Importantly, these results suggest that metals have a relatively minor influence on the UAR aquatic food web, and the observed bioenergetic changes were predominantly the result of habitat restoration. Proposed habitat restoration projects in miningimpacted streams would benefit from pre-restoration analyses to determine if remediation has sufficiently improved water quality for aquatic life (e.g. prey resources) to the extent that habitat restoration can provide additional benefits.
This research demonstrated that habitat restoration following improvements in water quality benefited Brown trout populations, and that responses of prey resources and trout to restoration were interdependent. Multiple lines of evidence indicated that while prey resources responded to changes in habitat, increased trout population density and biomass may have altered available prey resources and patterns of prey utilization. This suggests that habitat restoration projects that are primarily designed to increase trout populations may benefit by focusing on improving prey resources and utilization to support the bioenergetic needs of the restored fishery.
Our results also suggest that it is unlikely that the UAR has reached an equilibrium between resource production and consumption by trout (i.e. carrying capacity), and that continued monitoring will be required to determine the long-term effects and sustainability of restoration. Biological monitoring programs that assess changes in available prey resources, alongside prey resource utilization by target fish species, will provide a more complete characterization of restoration effectiveness on stream food webs.

Acknowledgments
We would like to thank the dozens of work-study students in the Aquatic Ecotoxicology Laboratory over the past decade for their time processing invertebrate samples that were used for this research. This study would not have been possible without the contribution of the many Colorado Parks and Wildlife (CPW) researchers, biologists, technicians, and volunteers that supported electrofishing surveys. Funding was provided by Natural Resource Damage Assessment provisions of the Comprehensive Environmental Response, Compensation, and Liability Act via the Colorado Department of Public Health and Environment. CPW also provided funding for this project, in part, through the Federal Aid in Sport Fish Restoration Program (Project F-161-R, Stream Habitat Investigations and Assistance). Comments by D. Walters, C. Saunders, and two anonymous reviewers on earlier drafts of this manuscript are greatly appreciated.

Supporting Information
The following information may be found in the online version of this article: Table S1. Proportion dry-mass (pDM) values reported for terrestrial invertebrate in the literature. Table S2. Average (AE s.d.) water temperatures in the month of August from 2008 to 2018 for all years that electrofishing surveys were conducted. Table S3. Mean (AE s.e.) water chemistry measurements collected in the habitat restoration reach of the Upper Arkansas River, 2008-2020. Table S4. Results of linear regression analyses testing for the effect of trace metals (as Cumulative Criterion Units (CCU)) on the biomass of prey resources, Brown trout diets, and Brown trout populations in the Upper Arkansas River. Table S5. Results of the generalized linear mixed models testing for the effects of restoration (before/after), annual stream discharge, habitat (gravel/grass/willow), and fish length on the biomass of prey resources, Brown trout diets, and Brown trout populations in the Upper Arkansas River. Table S6. Degrees of freedom and F-statistics for Generalized Linear Mixed Models testing for the effects of restoration (before and after completion of habitat restoration), average annual discharge, habitat (grass/gravel/willow), and fish length on biomass of aquatic and terrestrial prey, Brown trout (Salmo trutta) diets, and Brown trout population metrics in the Upper Arkansas River. Table S7. Mean (AE s.e.) biomass of benthic macroinvertebrates (mg m À2 ), emerging aquatic adults (mg m À2 ), streamside inputs (mg m À2 ), Brown trout (Salmo trutta) diets (mg fish À1 ), and Brown trout population metrics (biomass, kg ha À1 ; density, no. ha À1 ) before and after restoration, and among stations. Table S8. Variation in mean (AE s.e.) biomass of streamside inputs (mg m À2 ) among habitat types and biomass of Brown trout prey (mg fish À1 ) among size classes in the Upper Arkansas River. Figure S1. Map showing the locations of the four sampling stations located in the NRDAR habitat restoration segment of the Upper Arkansas River, Colorado USA. Figure S2. Length-frequency histogram of all Brown trout captured among stations during the study period. Figure S3. Aerial imagery showing the locations of the specific habitat restoration treatments and the approximate upstream and downstream sampling extents for fish and invertebrates for each station: AR4C (a), AR4E (b), AR4G (c), and AR4H (d). Figure S4. Trends in mean (AE s.e.) biomass separated by taxonomic groupings for (A) streamside inputs of aquatic and terrestrial insects, and (B) emerging aquatic adults measured at four sites located within the restoration reach of the Upper Arkansas River, before and after physical habitat restoration. Figure S5. Mean (AE s.e.) Brown trout (Salmo trutta) diet biomass of aquatic larvae, aquatic adults, and terrestrial invertebrates among the three trout size classes 120-199 mm, 200-299 mm, and 300-350 mm before and after habitat restoration.