Understanding how wildfires affect food web structure and function remains an important challenge, especially at high elevations that historically have burned infrequently. In particular, fires may alter the magnitude of reciprocal cross-ecosystem subsidies, leading to indirect effects on aquatic and terrestrial consumers. We quantified characteristics of high-elevation (2500 to 3000 m) stream-riparian food webs at 10 locations in the southern Rocky Mountains less than one year following high-intensity, stand-replacing wildfires. Using a paired 'burned-unburned' stream survey design, we assessed benthic periphyton, aquatic macroinvertebrate community structure, trout population characteristics, trout stomach contents, inputs and emergence of insects to and from streams, and abundance of predatory riparian spiders that consume aquatic insects. Benthic macroinvertebrate density, flux of emerging aquatic insects, and riparian spider abundances were lower at burned sites. Fluxes of insect inputs entering the stream did not differ with burn status, despite the loss of riparian vegetation due to fire. Trout were somewhat less abundant, but larger on average at burned sites and did not differ in body condition. These results suggest mortality of smaller trout from fire disturbance and/or recolonization of burned sites by larger individuals. Trout showed subtle changes in diet composition with burn status, but no change in biomass or number of prey consumed. In general, burned sites showed greater variation in community characteristics than unburned sites, which may reflect differences in the timing and magnitude of post-fire flooding, erosion, and scouring of the stream bed. Taken together, our results suggest that short-term effects of fire disturbance strongly altered some food web responses, but others appeared relatively resilient, which is notable given the high severity of the wildfires in the study area.

Stream surveys – We surveyed five pairs of streams in late July and August of 2021 in the northern Colorado Rockies within the Arapaho and Roosevelt National Forests.  Three of the burned sites were within the perimeter of the Cameron Peak Fire, one was within the East Troublesome Fire, and one was within the Williams Fork Fire. Each burned site was paired with an adjacent unburned site outside of the fire perimeter. At each site, we calculated the percentage of upstream watershed area that had been burned. We also visually estimated local tree mortality along the riparian corridor (% of trees killed by fire). To characterize the stream environment, we measured stream widths, depths, discharge, chlorophyll a from periphyton, canopy cover, adjacent burn severity, water chemistry (dissolved oxygen, pH, water temperature, conductivity), and turbidity at each site at the time of the survey. We attempted to select sites that were similar in elevation, aspect, stream size, and valley classification (Carlson 2009) between the burned and unburned pairs. To test this assumption, we assessed whether the burned and unburned sites differed in elevation, wetted width, maximum depth, or discharge using paired t-tests. There were no significant differences in these environmental variables (all p > 0.05). Trout species composition at all sites included nonnative brook trout (Salvelinus fontinalis) and/or brown trout (Salmo trutta), except for one site that supported cutthroat trout (Oncorhynchus clarkii).

Macroinvertebrates were sampled using benthic Surber samplers, emergence traps, and pan traps. Five Surber samples (0.09 m² in area, 248 mm mesh each) were evenly spaced along each reach in approximately the center of the stream, focusing on riffle habitat. We deployed three emergence traps and three pan traps within each reach on the same day as surveys were conducted. Traps of both types were collected after approximately 70 hrs (range = 66 to 72 hrs). Emergence traps (Cadmus et al. 2016) were 0.37 m² in area and incorporated an opening at the top that directed insects into a Nalgene bottle of 80% ethanol. Emergence traps were situated in relatively calm water, typically below a riffle. For pan traps, we used shallow plastic bins 0.30 m² in area supported on frames of PVC plastic tubing. Pan traps were situated in shallow water at the edge of the streams and were filled with stream water and approximately 5 mL of unscented dish soap. Insects were removed from pan traps using a handheld dipnet (750 mm mesh) and preserved in 80% ethanol.

To survey riparian spiders, we followed methods described in Benjamin et al. (2011). Spiders were surveyed along a 50-m reach that was undisturbed by electrofishing, either immediately upstream or downstream of the electroshocking survey reach. Because the focal spiders are nocturnal, spider surveys were initiated at night after 2100 h. Two observers wearing headlamps, one on each side of the stream, recorded all visible spiders within approximately 1 m of the wetted stream edge and up to 2.5 m in height. Based on their morphology, spiders were identified as Tetragnathidae, Araneidae, or "other" for individuals that could not be identified (<2% of all spiders). Prior studies indicated that detection probabilities using these survey methods were >90% (Benjamin et al. 2011). Because spider abundances may be related to habitat availability for web-building, we also quantified the density of branches along each stream reach (focusing on branches <5 cm diameter and >50 cm in length). Branch density was estimated in categories (0, 1-5, 6-25, 26-50, >50 branches) for every 2 m along both sides of the 50 m reach (Benjamin et al. 2011). We used the midpoint of each category as the estimated branch density. For the highest branch density category, we used 75 as the midpoint. Branch density data for a single site (South Fork Poudre) was lost, resulting in a comparison between four unburned and five burned sites.

We quantified trout relative abundance, population size structure, body condition, and stomach contents. At each site, we conducted a single-pass survey using a backpack electroshocker within a designated 50 m stream reach. The selected reaches were composed primarily of riffle and run habitats, which helped standardize habitat conditions between sites. All trout caught were anesthetized with Aqui-S, nonlethally lavaged for stomach contents, measured for total length (nearest mm) and weight (nearest 0.1 g), and then released after recovery from anesthesia. Fish were lavaged using either a laboratory wash bottle with an affixed plastic straw (for larger fish) or a 60 ml syringe with a blunt 18 gauge needle (for smaller fish). Fish stomach contents were preserved in 80% ethanol until processing in the laboratory. At some sites, we collected fewer than three trout in the designated reach, so we continued electroshocking upstream and downstream of the reach until we obtained 20 to 30 trout for diet analyses (mean = 27). We calculated trout relative abundance within each reach as the number of fish caught per minute of the electroshocker running. The number of people electroshocking (n=4) and number of backpack electroshockers (n=1) were standardized within the reach at each site. At one site (Corral Creek), problems with the electroshocker affected shocking efficiency, so we omitted this site from estimates of trout relative abundance but included it for all other analyses.

Sample processing – Macroinvertebrates from Surber samples were identified to family, while macroinvertebrates from trout stomach contents, emergence traps, and pan traps were identified to either family or order (Ward et al. 2002, Merritt et al. 2008). Macroinvertebrates from trout stomach contents, emergence traps, and pan traps were also assigned as either 'aquatic', indicating they have one or more aquatic life stages, or 'terrestrial', indicating they have no aquatic life stage. Invertebrates identified to order generally included adult life stages of certain taxa (e.g,, Diptera) and invertebrates from trout stomachs that were partially digested and had lost key features. A subset of 10 intact invertebrates of each taxon from each sample were measured for body length to the nearest 0.5 mm. We then used published length-to-mass regressions to estimate dry biomass for individual invertebrates (Benke et al. 1999, Sabo et al. 2002), which was then summed for each Surber to obtain biomass per unit area.