Flying by the river side: Survey of bat distributions and environmental contexts along a 1000‐mile river corridor, Green and Colorado Rivers, USA

Emerging research shows how bioindicators, specifically bats, can serve as a means for monitoring conservation and management of riparian corridors for multiple taxonomic groups. To track changes in the composition or abundance of bioindicator species, researchers must attain a baseline in species presence and relative activity. We examined the spatial and temporal patterns of bat community composition and activity along a 1000‐mile river corridor to determine species diversity trends by latitude and habitat.


| INTRODUC TI ON
Riparian and river systems are important habitats for a wide range of aquatic and terrestrial organisms yet represent some of the most highly degraded habitats on Earth (Hughes, 2021;Vörösmarty et al., 2010).Human alterations of landscapes, via urbanisation, agricultural land conversion and energy development, partnered with increasing climate instability have reduced the ecological function, resilience, and integrity of riverine ecosystems.These have in turn affected wildlife distributions, population persistence and species survival (Comte et al., 2022;Harris et al., 2009;Russo & Ancillotto, 2015;Webb & Leake, 2006).
As contiguous habitat becomes fragmented, decreases in landscape connectivity can lead to disruptions to long-standing migration corridors.Some of the most prominent examples range from disruptions to large ungulate migrations (e.g., pronghorn migration barriers in Wyoming, Kauffman et al., 2018;Robb et al., 2022) to the increased likelihood of mortality of migratory birds due to power lines and collisions with wind turbines (Palacín et al., 2017).While these examples focus on terrestrial environments, it is important to acknowledge how natural corridors, such as riparian and river systems, have also been altered and what that may mean for wildlife (Ragan et al., 2022).Structural changes in riparian corridors significantly reduce overall biodiversity along rivers by altering resource availability for predators and prey, moderating extremes in daily air temperatures, and often, homogenising floral and faunal assemblages (Cubley et al., 2020).
Emerging research shows how bioindicators can serve as a proxy for ecosystem health and be used to monitor the conservation and management of riparian corridors for multiple taxonomic groups (Caruso et al., 2019;Cubley et al., 2020;Kennedy et al., 2016;Merritt & Bateman, 2012;Tonkin et al., 2021).In the traditional sense, bioindicators are individual species, or a community of organisms, whose reactions can be monitored and tracked to evaluate changes in an ecosystem (Gerhardt, 2002).Species deemed as bioindicators are organisms that react to anthropogenic changes in the environment.Specifically, species may change in abundance and detection, or alter their behaviours due to perturbations in their environment that exceed their preferred limits (e.g., noise and light pollution, habitat alteration via roost removal, or chemical spills).Jones and colleagues (Jones et al., 2009) state that there are two distinctly different types of global change that should be monitored and understood: (1) environmental disturbances caused by global climate change (i.e., large-scale changes in weather patterns or storm intensity) and (2) alteration of natural habitats (e.g., removal or degradation of intact ecosystems for the benefit of human consumption/ agriculture).Both types of ecosystem change described above have the capacity to alter the likelihood of survival for native organisms.
For both of these drivers, bats have been suggested as being excellent bioindicators due to their wide geographic range, trophic diversity, slow reproductive rate, and known ecosystem services (Jones et al., 2009;Russo & Jones, 2015).While the research into using bats as bioindicators are limited, the use of this taxonomic group in this key role is promising and warrants further research into how they can be used to assess specific contexts such as river quality, urbanisation and agricultural land conversion, bioaccumulation, and climate change (Russo et al., 2021).
Bats often key in on linear landscape features such as wind rows, roads, and riverine systems to facilitate movement and migration across a patchwork of landscapes (Cortes & Gillam, 2020;Henderson & Broders, 2008;Russo & Ancillotto, 2015).This behaviour is likely facilitated due to the presence or availability of prey (Holloway & Barclay, 2000;Jackson et al., 2020;Lin et al., 2022;Scott et al., 2010), as well as a behavioural adaptation to avoid predation (Lima & O'Keefe, 2013).In some regions, these landscape features are often the least changed, especially when compared to contiguous forests and open habitats, thereby retaining some semi-natural vegetation in highly degraded landscapes (Lundy & Montgomery, 2010).Forest obligate species may rely more heavily on riparian habitats along streams and rivers to travel from seasonal roosts or foraging locations (Henderson & Broders, 2008), whereas other species may not be as highly affected by large-scale habitat changes.
In the American West, habitat type has been documented as one of the most important factors influencing bat activity, with the greatest number of bat calls, or overall activity, recorded via acoustic detectors located along the edge and riparian habitat (Currier et al., 2006;Hagen & Sabo, 2014;Rogers et al., 2006;Williams et al., 2006).Habitat quality, as well as the diversity in vegetation species and forest stand age are also important factors when it comes to the amount of bat activity and the species of bats present (Currier et al., 2006); specifically, providing adequate prey and a multitude of roost types.River size, landscape and water quality may also affect bat species abundance and relative activity.Hagen and Sabo (2014) determined that reduced riparian habitat, high water temperatures, high stream primary productivity and stream flow (e.g., intermittent to flashfloods) can lead to changes in terrestrial arthropod availability, which can impact terrestrial predators.
If flows are dramatically altered, or the quality of water diminishes, taxonomic groups like bats may not be able to sustain populations along some riparian corridors.While there have been numerous studies focused on the activity of bats along river corridors, most occur along short segments of river (~5-50 miles), over long timescales (i.e., passive detectors at long-term monitoring sites for days to months) or via mobile driving transects.Methods for acoustically detecting and identifying bats have improved, and become more creative, over the years (Braun de Torrez et al., 2017;Jackson et al., 2020;Kennedy et al., 2014;Krauel et al., 2018;Whitby et al., 2014), allowing for unique opportunities to catalogue bat presence, habitat associations, and activity patterns in locations that are often hard to reach, or take great effort to access.Here we examined bat species presence and activity for 70 days along a 1000-mile river continuum (spatially robust with single-night samples).
The primary objectives of this project were to collect acoustic data to determine the spatial and temporal patterns of species presence and relative activity along the Green and Colorado Rivers from Green River, Wyoming to the confluence of the Colorado and Virgin Rivers, Lake Mead, Nevada.The acoustic bat survey was conducted opportunistically in association with the Sesquicentennial Colorado River Exploring Expedition (SCREE) in 2019 on the 150th anniversary of John Wesley Powell's first expedition along this same section.We use this broad survey to provide a baseline for species distributions over a large geographic range and identify regions of concentrated usage based on individual and species presence and relative activity.Specifically, we hypothesised that (H1) species composition will vary and diversity will increase along a latitudinal gradient.We predict that species, such as Eptesicus fuscus, Lasiurus cinereus, and Lasionycteris noctivagans will be detected throughout the survey period, however, species with more specific habitat associations, such as Parastrellus hesperus and Euderma maculatum will be detected less frequently due to habitat requirements (i.e., physiography).(H2) Relative activity of each species detected will vary based on environmental contexts such as temperature (i.e., at dusk and over the sample period [dusk to dawn]) and river features (i.e., reservoirs, slow-moving sections, and rapids).We predict that smaller-bodied species will be detected less frequently at lower temperatures (<15°C); thus, peaks in their activity will occur immediately following dusk.We also predict that bat activity would be lower on river sections with a high degree of ambient noise (i.e., rapids).

| Site description and data collection
In 2019 the Sesquicentennial Colorado River Exploring Expedition (SCREE; www.powel l150.org) of the Green and Colorado Rivers provided a platform for opportunistic sampling of the distribution of North American bats along a major river corridor.The purpose of the expedition was to examine socio-ecologic relationships based on modern river management 150 years after the John Wesley Powell expedition in 1869 (Powell, 2012;Stegner, 1992;Worster, 2002).While land usage and public policy was a large focus of the project, longitudinal surveys of various water conditions, such as water temperature, pH, and turbidity, as well as species composition (i.e., bats, birds, and flora) were incorporated into daily activities.
As major river systems, anthropogenic influences are omnipresent.Large reservoirs occur at the top, middle, and bottom of the survey route, starting at Flaming Gorge, Wyoming, to Lake Powell, Utah/Arizona and ending at Lake Mead Arizona/Nevada (Figure 1).River flow out of the upper two reservoirs has been regulated since the 1960's (Bestgen & Crist, 2000;Hundley, 2009), with water release from the dams following adaptive management plans to enhance ecological health of riparian systems (Cross et al., 2011;Hall et al., 2015;Pulwarty & Melis, 2001).Examples include spring flood releases from the Flaming Gorge Dam to match natural floods from downstream tributaries to activate floodplains to augment fish hatch habitat in the 425-mile section between Flaming Gorge and the slack water deposits at Lake Powell.Another example is reduced weekend hydropeaking from Glen Canyon Dam into the Grand Canyon section of the river to improve insect hatch (Abernethy et al., 2021).Despite the prolonged, millennial-drought in the Colorado River Basin, 2019 was an unusually wet year potentially creating a short-term reprieve from drought stress for the flora and fauna of the region.
The expedition covered 1000 river miles over 70 days, from 25 May to 2 August 2019.The Green and Colorado Rivers between Green River, Wyoming and the confluence of the Virgin River, Nevada, mostly trends north-south, until the Grand Canyon where the river flows predominantly west to the confluence of the Virgin River (Figure 1).Data collection was a hybrid of mobile and stationary transects (as defined by Braun de Torrez et al., 2017;Loeb et al., 2015) and enabled us to sample a large spatial area.
This method allowed us to maximize sample duration (i.e., each site sampled for an entire night), thus increasing the likelihood of truly detecting a species if they were present at a site during each sample session.We were not able to estimate bat abundance, as we could in a more traditional mobile transect survey.Bat surveys were spatially independent (between 7 and 30 river miles between camps) and conducted using a Wildlife Acoustics SM4BAT FS Full-Spectrum Ultrasonic Recorder™ with SMM-U2 microphones (Wildlife Acoustics, Maynard, MA, USA).According to the manufacturer's guide (Wildlife Acoustics, 2019), the SMM-U2 can cover up to 8x the amount of airspace as the SMM-U1 (maximum detection range of older units [SM3Bat + SMM-U1] estimated at ~40 m; Cortes & Gillam, 2020).In total, 63 surveys were conducted, with 4 repeat surveys during stopover days (7/11-7/12 and 7/21-7/22), totaling 61 unique locations.The recorder was placed outside of base camp and set to record from dusk to dawn.Placement of the recorder was typically within the riparian corridor, proximal to the river, but in an opening that had clear sky-views.Due to the nature of river travel, we were unable to maintain a strict 30-min before dusk to 30-min after dawn protocol.Sections of the river were categorised as open (basin), closed (canyon), or anthropogenic (reservoirs) in roughly equal proportions.Most of the camp locations were in "quiet" glide sections of the river.In canyon sections, riffles (rapids) were more common and introduced white noise to the acoustic space.

| Species identification
We identified species based on echolocation calls and call sequences using two methods: semi-automated identification software paired with manually vetting.Acoustic files were run through both Kaleidoscope Pro 5.1.9Analysis Software (Bats of North America classifier 5.1; Wildlife Acoustics, Maynard, MA, USA) and SonoBat version 4.4.5 (North America, Great Basin North, Great Basin, and Southwest regional classifiers; DNDesign, Arcata, California, USA) to compare and vet bat species identification.We then compared the two outputs using the 'comparedf' function in the compare package (Murrell, 2015) in the statistical software R (v. 4.3.1;R Development Core Team, 2023) using R Studio (v.2023.06.2+561) and manually vetted all calls that were identified as two different species in Sonobat and Kaleidoscope Pro (1948 calls) and 10% of calls that were identified in one software but left blank in the other (21,375 calls = 2138 vetted calls).We also cross-checked species identification with known records of each species distribution and state observations.

| Statistical analysis
Acoustic activity was quantified as the number of calls or "bat passes" recorded per night (i.e., sunset -sunrise the next day).A "bat pass" was defined as a file containing a search-phase echolocation sequence of ≥2 echolocation pulses (Gannon et al., 2003).We used number of call files as an index of overall activity, not to determine the true numbers of individuals per species at each location surveyed (Bernard & McCracken, 2017).For total activity per night, we used the total high and low frequency bat calls recorded per night as that included calls identified to species, as well as bat calls that were not clean enough to be identified via SonoBat & Kaleidoscope.All analyses were conducted in R (Version 4.1.1;R Development Core Team, 2023).

| Estimation of sampling effort
We generated sample-based rarefaction species accumulation curves using the function 'specaccum' in the vegan package (Oksanen et al., 2022), using the 'exact' method, which estimates the mean species richness across sites.The maximum expected species richness is estimated when the species accumulation curve reaches an asymptote, or when the probability of detecting new species approaches zero.

| Variation in species activity
To determine if activity varied by species over time (i.e., nightly activity (hour from midnight), as well as seasonal activity (Julian day)) we ran negative binomial generalised additive models (GAM) using the MASS package (Venables & Ripley, 2002) and the mgcv package (Wood, 2011) following methods described in Davison and Thomas (2017) and Thomas and Davison (2022).We allowed the mgcv package to automatically choose the value for k (degrees of non-linearity) needed to minimize AIC values.Model assumptions were assessed using the mgcv package.Previous studies have found that bat activity can vary based on the structural characteristics of riverine landscapes, such as channel geomorphology and riparian physiognomy (Hagen & Sabo, 2011).Therefore, we ran an Analysis of Variance (ANOVA) on total bat calls and river feature (i.e., glide, riffle, and still) and post hoc Tukey test (TukeyHSD()).To determine if bat activity varied with temperature, river feature, and physiographic division, we fit generalised linear mixed models using the glmmTMB package (Brooks et al., 2017) and kept bat day as a random effect (this was a proxy for site).We fit models using a Poisson distribution and assessed model assumptions using the DHARMa package (Hartig, 2021).Specifically, we tested how total bat calls (sum of high and low frequency calls) may vary with mean nightly temperature (averaged over 20:00-07:00), temperature at dusk and two categorical variables describing environmental contexts (river feature and physiography).

| Variation in species composition
To visually inspect the variation in species distribution across the landscape, we compared physiographic sections (Fenneman, 1917; Powell & National Geographic Society (U.S.), 1895) and total bat calls (i.e., those identified to species) using the ordination technique of nonmetric multidimentional scaling (NMDS) in the package vegan.
We used a Jaccard dissimilarity distance measure to calculate distance among call files (total call files identified to species were converted to presence/absence not abundance) and assessed the fit of the NMDS by observing the "stress" value.

| RE SULTS
We recorded 59,567 files of which 58,225 were identified as bat calls (>2 pulses), representing 19 species (Table 1, Figure 2).Of the total calls identified as bats, 18,490 were identified to species, all of which are known to be present in their respective states.Bat calls were recorded on all days we attempted to record (n = 63 days of the 70-day trip), ranging from 1 to 3217 (±681.23)calls per day (totals based on calls identified as high or low frequency).We reached the asymptote, or maximum number of bat species recorded, by day 20, indicating species composition varied both temporally and spatially (Figure 2).
Latitudinal gradients in species composition were evident, with diversity increasing north to south (Table 1, Figure 2).This result might be biased by physiographic complexity, with more open conditions away from the riparian corridor at the beginning of the transect (i.e., river flowing through the Wyoming Basin) in contrast to the end (Grand Canyon).Species composition also varied based on physiographic region (Figure 3; stress = 0.0754, k = 3), with these differences driven by the abundance of species and calls in the Grand Canyon and Canyon Lands (TukeyHSD p = .0037)and Middle Rocky Mountains and Grand Canyon (TukeyHSD p = .0005).Calls recorded in the Wyoming Basin were limited due to cool temperatures, high precipitation, and an early season start to the river trip.
We found that overall bat activity was best explained by temperature, where higher mean temperatures led to increased bat activity (β = .0779,p < .001; Figure 4, Table 2).Mean temperature during the sampling period remained below 12°C during the first 9 days, however, as the team continued to travel south, mean nighttime temperature increased, while precipitation decreased, allowing for longer periods for bat activity and foraging.We also determined nightly activity varied across species, with those recorded often (i.e., Antrozous pallidus, L. noctivagans, Myotis californicus, P. hesperus and Tadarida brasiliensis) increasing activity during different temperature thresholds (Figure 5a; p < .001).Lasionycteris noctivagans were more active at lower temperatures than the four other species, who increased in activity starting at 20°C.When looking at activity based on the temperature at dusk, L. noctivagans peaks in activity around 15°C, with the smaller bodied bats like M. californicus and P. hesperus more active at temperatures exceeding 25-30°C (p < .001, Figure 5b).Species activity also varied temporally, as noted in many other studies.Bat activity increased 2 h post sunset and remained consistent at lower levels throughout the night followed by one smaller peak in activity just before dawn (Figure 6; R 2 = .366,p < .001).
While this general pattern remains consistent across studies, we also found variation in activity levels by species (Figure 7; R 2 = .291,p < .001),with P. hesperus following the classic bimodal form of activity.However, M. californicus and T. brasiliensis increase their level of activity on the landscape 4-6 h post sunset, with the highest peak in activity occurring an hour before sunrise.Finally, we found that bat activity varied significantly when examining the total number of calls by river feature.More bats were recorded in glide and riffle sections than were over still, reservoir-like sections of river (Figure 8).Significant differences were driven between riffle and glide (TukeyHSD; p < .03)and riffle and still (TukeyHSD; p < .03),while there was no significant difference between glide and still.

| DISCUSS ION
Our research provides a critical baseline for bat activity along a vast and heavily pressured riverine ecosystem.In 2009, Jones and colleagues provided robust support for bats to be used as bioindicators in the habitats they reside (Jones et al., 2009).Since that time, numerous studies have provided additional evidence to support this suggestion (De Conno et al., 2018;Russo et al., 2021;Russo & Jones, 2015;Stahlschmidt & Brühl, 2012).Specifically, in places where increased drought and access to water can become limited, such as western ecosystems, changes in ambient conditions (i.e., anthropogenic habitat conversion, climate change, etc.) have been linked with a decrease in reproductive output of six vespertilionid species (Adams, 2010;Adams & Hayes, 2018).Changes in the timing and flow of water in the Colorado River system can dramatically alter the phenology, food web interactions, and biodiversity of an entire TA B L E 1 Acoustic call files identified to species for each physiographic section along a 1000-mile river corridor.system (Adams, 2010).Therefore, our results add another layer of evidence highlighting the importance of river corridors for insectivorous bats, not only as foraging hotspots (Hagen & Sabo, 2011;Metcalfe et al., 2023;Scott et al., 2010) but also as corridors for dispersal and commuting between daily and seasonal roosts (Lundy & Montgomery, 2010;Walsh & Harris, 1996).Our study offered the opportunity to sample bats in regions that may otherwise be hard to access via road or trail, allowing for quantifying bat species occupancy in areas that have never been surveyed before (Braun de Torrez et al., 2017).

Rocky
While our survey was a hybrid of passive and active acoustic surveillance (i.e., 8+ h of recording at one GPS location repeated along a 1000-mile river transect), this broad, slow-moving survey allowed for spatial overlap in habitat that may overcome point based precision attained from long-term stationary data collection (Braun de et al., 2017).Most deployment sites were ~20 river miles from each other, often passing through diverse habitat (Triedman, 2012).
However, the distance between sample sites is not necessarily be- F I G U R E 3 NMDS plot of bat calls identified to species.This analysis showed some degree of spatial partitioning among bat species according to physiographic sections.Stress metric = 0.0754, k = 3.
certain that each detector night is independently sampled, the likelihood of recording the same bat across multiple nights is low.
We were concerned that migration during the early part of the survey might skew early observations, but in the case of the upper reaches of our survey in Wyoming, most species are residents, with only 3 (L.cinereus, L. borealis, and L. noctivagans) considered migrants (Buskirk, 2016).Ultimately, we do not know how bats navigate when TA B L E 2 Akaike information criterion results for the seven generalised linear mixed models for total bat calls (i.e.sum of all high and low frequency calls) and several environmental variables: mean temperature from dusk to dawn, temperature at dusk, river feature (i.e., riffle, pool, glide) and physiographic division (i.e., Rocky Mountain System and Intermontane Plateaus).We decided to choose physiographic division for this analysis due to a more robust sample for each division.pristine, low-light environments (Hooker et al., 2022).As riverine ecosystems become more developed, artificial light at night (ALAN) or part-night light (PNL) can lead to fragmentation of foraging corridors, thereby leading to higher energy demands for bats needing to fly further to access dark environments for foraging opportunities (Cravens et al., 2018).The effect of ALAN or PNL on bats has been found to be strongest in rural habitats, where novel light is rare (Barré et al., 2022).forms that differ from the surrounding regions (Fenneman, 1917; Powell & National Geographic Society (U.S.), 1895).Physiographic regions often follow similar boundaries to ecoregions used in ecological mapping; therefore, species distributions and communities may be distinct geographically.We recorded the expected diversity gradient, with more species present at 36° N than at 41.5° N.
However, cooler temperatures related to late-spring conditions during the first week of the expedition, and the east-west orientation of the Grand Canyon increasing over all environmental complexity at the lower latitude of our transect may bias this gross pattern.
Reframing our first question into consideration of physiography, Our expectations for the second hypothesis: that warmer temperatures and river reach characteristics (pool, riffle, glide) would increase bat activity were in part met, with warmer air temperatures leading to greater call activity.In contrast, we identified differences in total bat activity when comparing river flow dynamics, contrary to our prediction.Levels of activity over still, reservoir waters was lower than that of glide and riffle, which is driven by surrounding vegetative structure and foraging behaviour and is consistent with previous studies (Ciechanowski, 2002).Reservoirs River District, where long-term ultrasonic detectors could be deployed to build upon the data collection we present here.
Our research provides a broad baseline for an already altered system and whose management will be further altered in response to ongoing drought in the western US.Given the propensity for anthropogenic change such as infrastructure buildout and climate change, which can lead to irreversible change within riverine ecosystems, long-term monitoring of bats along the river and terrestrial environment may allow for managers to better estimate how to mitigate and improve future outcomes for ecosystem health and function within this river system.While research into the use of bats as bioindicators is ongoing, tracking the relative abundance of some common bat species along the Colorado River corridor over time may assist managers in detecting deviations in river quality, the influence in urbanisation along river systems, as well as pesticide use in agroecosystems (Russo et al., 2021).

CO N FLI C T O F I NTE R E S T S TATE M E NT
We have no conflicts of interest to disclose.

PE E R R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/ api/ gatew ay/ wos/ peer-review/ 10. 1111/ ddi.13842 .

DATA AVA I L A B I L I T Y S TAT E M E N T
The complete dataset used for analysis is available in Dryad: https:// doi.org/ 10. 5061/ dryad.cfxpn vxcw.Dryad repository -https:// datad F I G U R E 8 Variation in the number of bat calls recorded based on river feature.Glides were defined as quiet, slow-moving sections of river; riffles were rapids and more common in canyon sections of the river; still were defined as reservoirs.

F
I G U R E 1 Map of the 2019 the Sesquicentennial Colorado River Exploring Expedition (SCREE) on the Green and Colorado Rivers.Over the course of the 70-day expedition, one bat detector was set to record from dusk to dawn.We recorded bats on each of the 63-nights the acoustic detector was active.Black dots represent the acoustic detector sites along the river corridor.Physiographic regions are labelled and identified in black polygons.
Calls were recorded nightly using a SM4BAT FS Full-Spectrum Ultrasonic Recorder™ with SMM-U2 microphones (Wildlife Acoustics, Maynard, MA, USA).SonoBat Autoclassification software (v 4.4.5) was used to identify call files to species.aSonobat identified all calls as Myotis lucifugus in regions where Myotis occultis may reside.Therefore, there may be an overestimation of M. lucifugus calls and an underestimation of M. occultus calls.
yond the realistic distance for commuting bats, especially if considering the likelihood of "shortcuts" used in open habitats.For example, assuming straight-line distance, female Leptonycteris yerbabuenae have been recorded flying over 80 km (~50 miles) round trip from caves to capture sites in saguaro forests where they forage (Medellín et al., 2018).Whereas, adult Corynorhinus rafinesquii in Kentucky, USA, were found to travel a maximum of 4334 m (2.6 miles) each night, suggesting high associations with their lepidopteran prey found in bottomland hardwood forests near their day roosts (Johnson & Lacki, 2013).Therefore, while we cannot say for F I G U R E 2 Species accumulation curve across survey nights showing temporal and spatial variation in the number of bats detected along the Green & Colorado River corridor.
migrating or how landscape features are used for long-distance movements.River systems and riparian habitats are likely frequently used corridors in otherwise open habitats, highlighting the need to determine how conversion of these landscapes may influence use and disturbance of existing migration routes.Habitat conversion, increased barriers, and light pollution on rivers may change bat flight behaviour, foraging attempts and or foraging success in previously F I G U R E 4 Total calls recorded varied according to temperature.However, this trend was largely driven by time of year, with a low number of calls recorded at the start of the trip (24 May 2019) when temperatures in Flaming Gorge, WY were below 10°C.

F
I G U R E 5 (a) Variation in activity of the five most recorded species compared to mean temperature at night (°C).(b) Variation in activity of the same species based on temperature at dusk (°C).Black line represents a smoothed trend line related to activity of each species over time (i.e., time since sunset) with the grey shading indicating the 95% confidence interval.Species: ANPA, Antrozous pallidus; LANO, Lasionycteris noctivagans; MYCA, Myotis californicus; PAHE, Parastrellus hesperus; TABR, Tadarida brasiliensis.The 1000-mile river corridor we sampled spanned 5.5° of latitude and flowed through six physiographic regions (Wyoming Basin, Middle Rocky Mountains, Uinta Basin, Canyon Lands, High Plateaus of Utah, and Grand Canyon) located within three physiographic provinces (Wyoming Basin, Southern Rocky Mountains and the Colorado Plateau).Physiographic provinces are geographic regions with characteristic geomorphology in which, over time, climate and geologic features have given rise to a variety of land- we found species composition varied by physiographic region (H1), with the greatest separation (or clumping) identified between Grand Canyon and Canyon Lands, and Middle Rocky Mountains and Grand Canyon.The Middle Rocky Mountain physiographic region includes the Bighorn and Wind River ranges in Wyoming, the Wasatch Range of southeastern Idaho and northern Utah, and the Uinta Mountains of northeastern Utah, of which the dominant type of mountains include volcanic and uplifted fault blocks.By contrast, the Canyon Lands are comprised of sandstone with canyons, basins and laccolithic mountains and structural domes.Finally, the Grand Canyon physiographic region is distinct from both as it is a river valley with significant uneven uplift from the Colorado Plateau.Due to the incredible diversity of ecotypes, the community composition of bats varies substantially over the course of the 1000-mile corridor, yet characteristic subdivisions of bat species assemblages occurred.While course, the community composition of bats within physiographic regions could allow for baseline comparisons and changes overtime.

F
Total calls recorded per species as a function of hours from sunset (cumulation of all calls recorded during the sample period).Black line represents a smoothed trend line related to activity of each species over time (i.e., time since sunset).Species are colour coded with each dot representing the total number of calls identified to species at a given time post sunset.Species: ANPA, Antrozous pallidus; COTO, Corynorhinus townsendii; EPFU, Eptesicus fuscus; EUMA, Euderma maculata; IDPH, Idionycteris phyllotis; LABL, Lasiurus blossevillii; LACI, Lasiurus cinereus; LANO, Lasionycteris noctivagans; MYCA, Myotis californicus; MYCI, Myotis ciliolabrum; MYEV, Myotis evotis; MYLU, Myotis lucifugus; MYTH, Myotis thysanodes; MYVO, Myotis volans; MYYU, Myotis yumanensis; NYSP, Nyctinomops species; PAHE, Parastrellus hesperus; TABR, Tadarida brasiliensis.had little-to-no riparian vegetation when compared to freeflowing sections.Narrower sections of river, where water is moving faster, may provide safer foraging corridors, preventing bats from flying out in the open (i.e., anti-predator behaviour; Hagen& Sabo, 2011).While turbulent sections of river (i.e., riffle and glide) produce more acoustic noise, we detected higher levels of bat activity, which is in contrast to previous research(Hagen & Sabo, 2011).One reason for this difference may be due to the size of the rivers being examined.Hagen and Sabo (2011) conducted their study along a 5-km stretch of the South Fork Eel River in California where summer base flow ranges from 2 to 4.5 m 3 /s, and peak discharge during winter storms can reach 56 m 3 /s(Power et al., 2004).In contrast, mean flow in summer 2019 was 409 m 3 /s at Lees Ferry, Arizona, USA (USGS, https:// water data.usgs.gov/ nwis/ wys_ rpt).Thus, the use of riffle versus glide and pool by bats may be context dependent.The scale of our survey may negate small-scale reach dynamics evident in more focused studies.Further, rivers with substantial water flow are likely wider than smaller order streams, thus the acoustic interference in riffle sections of wider rivers may be less disruptive than smaller rivers with tighter corridors and canyons.Again, while the level of bat activity in these sections of river may be context dependent, long-term acoustic sampling for bats may help natural resource managers better understand how anthropogenic-driven change in river flow may lead to changes in ecosystem function within the system.While a full basin repeat survey, or multiple surveys, would be ideal, it is an unreasonable expectation given the costs and logistical requirements needed for such an extensive and physically demanding resurvey.Seasonal movements and use of habitats by bats could be captured by greater utilisation of opportunistic sampling F I G U R E 7 General activity patterns of 19 bat species recorded along a 1000-mile river corridor.Species activity varied over time (i.e., hours post sunset), with some bat species becoming more active as time progresses.Black line represents a smoothed trend line related to activity of each species over time (i.e., time since sunset) with the grey shading indicating the 95% confidence interval.Each y-axis varies based on the total number of call files identified per species.Species: ANPA, Antrozous pallidus; COTO, Corynorhinus townsendii; EPFU, Eptesicus fuscus; EUMA, Euderma maculata; IDPH, Idionycteris phyllotis; LABL, Lasiurus blossevillii; LACI, Lasiurus cinereus; LANO, Lasionycteris noctivagans; MYCA, Myotis californicus; MYCI, Myotis ciliolabrum; MYEV, Myotis evotis; MYLU, Myotis lucifugus; MYTH, Myotis thysanodes; MYVO, Myotis volans; MYYU, Myotis yumanensis; NYSP, Nyctinomops species; PAHE, Parastrellus hesperus; TABR, Tadarida brasiliensis.through citizen scientists travelling through individual river sections.While conducting research on the river may be challenging, we suggest agency and citizen science partnerships for additional long-term passive monitoring of bats in this system.There are thousands of streamgages managed and maintained by agencies, such as the U.S. Geological Survey (Eberts et al., 2019) and Colorado There are too many to list individually.Thank you to the National Park Service Dinosaur National Monument and Grand Canyon National Park and Bureau of Land Management Desolation Gray Canyon and Stillwater Canyon for coordinating permit timings for the complete journey.Acoustic surveys were conducted along the Green and Colorado Rivers, which flow through the traditional lands of many Native peoples, including the Eastern Shoshone, Shoshone-Bannock, Ute, Pueblo, Nuwuvi, Havasu Baaja, Hualapai, Hopitutskawa, Dine, and Zuni.FU N D I N G I N FO R M ATI O N This material is based upon work supported by the U.S. Geological Survey: Youth Education and Science under Grant/Cooperative Agreement No. G19AC00119.The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the U.S. Geological Survey.Mention of trade names or commercial products does not constitute their endorsement by the U.S. Geological Survey.
Note: ΔAICc, the difference between the AIC of the best model compared to current model; AICc weight, the proportion of the total predictive power that can be found in the model; AICc, corrected AIC value of each model; Cumulative weight, cumulative sum of the AIC weights; k, # parameters in the model; LL, log-likelihood of the model.