Data: Snorkel survey

METHODOLOGICAL INFORMATION Description of methods used for collection/generation of data: To test hypotheses regarding fish use of submerged macrophyte-based microhabitats across multiple scales, we conducted snorkel surveys at each reach to characterize fish microhabitat preferences at the micro-scale (~1 m), and concurrently conducted physical habitat surveys meant to identify physical habitat availability at the reach scale (~100 m). We sampled in summer (July) and fall (October) of 2020, and spring (May) of 2021 and across the Headwaters, Middle, and Tailwater locations to capture fish habitat use and availability across spatial and temporal gradients in submerged macrophyte coverage. In July 2020, above-average flows (&gt;1500 cfs) and turbidity (&gt;5 NTU) precluded safe and effective use of snorkel surveys and physical habitat surveys at the Last Chance reach. Our July 2020 data for Last Chance is therefore derived from radio-telemetry-based habitat selection data and randomly-selected physical habitat data from June 2013 and June 2014 (Kuzniar et al. 2017), collected at Harriman Ranch. We were also unable to collect physical habitat data for one Headwaters reach, Buffalo River, and one Middle reach, Flat Rock, during the summer of 2020 due to sampling constraints associated with the onset of the COVID-19 pandemic. The decision not to collect at Buffalo River and Flat Rock was not deliberate or pre-planned. Since each location is made up of two sampling reaches, we are assured that missing a single sampling event at a single reach did not impact data availability or our results. <br> The snorkel-based individual fish habitat use survey used a matched-pairs design in which fish habitat preferences were deduced by comparing physical habitat characteristics between fish focal points located through snorkeling or wading and paired, randomly-selected “nearby” points which represented points where we assumed fish could have been but chose to avoid. Fish focal points and metabolic parameters (size, species, and temperature) at each focal point were estimated through snorkeling surveys (Thurow 1994). Water clarity was good at all sampling events, ensuring high quality data (&lt;5 NTU, &gt;1 m visibility). If submerged macrophyte growth compromised visibility, we used visual wading surveys with polarized sunglasses, which have been shown to be effective at counting and observing fish in relatively shallow, clear streams (Hankin and Reeves 1988). Wading surveys consisted of 3-5 surveyors with polarized sunglasses walking from the downstream end of the sampling reach to the upstream end in a zig-zag pattern. For each fish focal point, we randomly selected a paired nearby sampling point which represented a point where the observed fish could have inhabited. A random-number generator in R version 3.6.2 (R Core Team, 2019) defined a random compass direction (0-360° from N, in 10° increments) and distance (6-12 m, in 0.1 m increments) from the fish focal point for the nearby point. Kuzniar et al. (2017) found fish focal points in the Henrys Fork to be approximately 3 meters in diameter, and so selecting a nearby point requires moving at least 6 m away from the fish’s observed focal point. We ignored young-of-the-year fish in our snorkeling survey since their high densities would have exhausted resources available for sampling. <br> Next, we measured physical and biotic habitat characteristics at each of the paired points. We estimated submerged macrophyte coverage (%) with a 0.25 m2 plexiglass viewbox with a 1/16 grid. We used a decimeter top-set wading rod to measure water depth (m) and maximum submerged macrophyte growth height above the substrate (m) visible in the viewbox window. We later calculated relative submerged macrophyte growth height as a percentage of depth (%). We measured flow velocity (m·s-1) at 20%, 60%, and 80% depth with a Hach Corporation Marsh-McBirney electronic flowmeter or an anemometer-style analog flowmeter. We estimated substrate size (mm) by measuring a single particle on a gravelometer from a random grab. <br> We estimated invertebrate drift density (mg⋅m-3) for each reach within each season using standard equipment and methods (Baxter et al. 2017). We collected three replicates from three equally spaced 250 μm mesh, 0.25 m2 opening drift nets at each reach for a total of 9 samples per reach and season. The National Aquatic Monitoring Center at Utah State University identified invertebrates to genus and sized to the nearest mm. Invertebrate genus and sizes were converted to biomass using genus- and family-specific length-biomass curves in Benke et al. (1999). We estimated the volume of water sampled by each net at each sampling event by recording and linearly interpolating water velocity in each net every minute using a Marsh-McBirney flowmeter. Time to 70% filtration efficiency determined maximum sampling soak times for each replicate net, which were 15 minutes in the Tailwater locations and 30 minutes at the Middle and Headwater locations (Baxter et al. 2017). We calculated invertebrate drift density (mg·m-3) for each season and reach from biomass (mg) and water volume (m3). We used the BioenergeticsHSC program to estimate average Net Rate of Energy Intake (NREI) by combining reach-scale invertebrate drift density with measures of water velocity, water temperature, turbidity, water depth, and fish size and species from each paired point in the snorkeling survey (Naman et al. 2020) Methods for processing the data: Pre-processing involved calculating the water velocity at the actual vertical position in the water column where the fish was sighted (20%, 60%, and 80% corresponded to "T", "M", and "B", respectively wherein T = Top, M = Middle, and B = Bottom). Next, we calculated a "roughness" measure for input into the BioenergeticsHSC program. Roughness corresponded to the size of macrophyte growth or substrate, whichever was greater. <br> NREI values (j·s-1) were calculated using the BioenergeticsHSC program (Naman et al. 2020). The BioenergeticsHSC program has batch processing methods that allowed us to feed the program the necessary data from the UHF_Snorkel.csv dataset, the Henry's Fork Foundation data website: https://henrysforkdata.shinyapps.io/scientific_website/, and from the HSCinput.csv dataset, which contains necessary invertebrate drift data. Data required for NREI calculations included: temperature (C), dissolved oxygen (mg·l-1), water depth (m), velocity (mg·l-1), substrate size or macrophyte growth height (whichever was larger, mm), turbidity (NTU), and invertebrate drift information. Invertebrate drift information is contained within the "HSCinput.csv" dataset, and is organized into rows by invertebrate genus and size (1 mm bins). Each row contains the density in mg·m-3, and the regression equations from Benke et al. 1999 to calculate dry mass. Energy density remains at the preset default of 5200. This data is fed in various forms to the BioenergeticsHSC GUI, manual available by Naman et al. (2020). <br> To assess which habitat variables fish preferred as habitat and whether reach-scale submerged macrophyte coverage affected habitat selection, we modeled microhabitat selection using conditional, matched-pairs logistic regression, with velocity, depth, submerged macrophyte cover, median substrate size, and NREI as potential predictors of amalgamated fish locations, following Kuzniar et al. (2017). We included reach-scale average submerged macrophyte coverage as an interaction term across all variables to measure whether seasonal or locational differences submerged macrophyte coverage modulated fish habitat preferences. The response variable in matched-pairs logistic regression is a change in the log-odds of organism presence per unit change in each predictor variable. Each habitat observation is expressed as the difference in a particular habitat-variable measurement between the focal point and the nearby random, unoccupied point. Matched-pairs logistic regression models are fit without an intercept so only relative odds, and not absolute odds or probabilities, can be calculated. Paired differences of habitat-variable measurements have symmetric distributions and require no data transformations. <br> We made a total of 327 pairs of observations across each of the nine season-location combinations; 48 pairs of observations in the Tailwaters in the summer by Kuzniar et al. (2017), and 24 to 31 pairs of observations, depending on the number of fish sighted during snorkeling surveys, in the remaining season-location combinations. We fit 243 candidate models of the additive combinations of the five predictor variables: depth, submerged macrophyte cover, substrate size, velocity, and NREI plus the five interaction terms with reach-scale average submerged macrophyte coverage. After establishing a list of candidate models we calculated log likelihood, AICc scores, and model weight for the group of candidate models (Burnham and Anderson 2002; Gelman and Hill 2007). These AICc-based model weights are then used to compute averaged model coefficients and standard errors using the R package MuMln (Burnham and Anderson 2002, Barton 2022). AICc scores, Wald’s Z scores (Sokal and Rohlf 2012), and the averaged model coefficients inform which variables are most likely to explain the relative odds of fish presence. <br> More information, including citations, author information, and more is available in the metadata document "McLaren_SnorkelSurvey_Readme.txt"