Habitat usage of Daubenton's bat (Myotis daubentonii), common pipistrelle (Pipistrellus pipistrellus), and soprano pipistrelle (Pipistrellus pygmaeus) in a North Wales upland river catchment

Abstract Distributions of Daubenton's bat (Myotis daubentonii), common pipistrelle, (Pipistrellus pipistrellus), and soprano pipistrelle (Pipistrellus pygmaeus) were investigated along and altitudinal gradient of the Lledr River, Conwy, North Wales, and presence assessed in relation to the water surface condition, presence/absence of bank‐side trees, and elevation. Ultrasound recordings of bats made on timed transects in summer 1999 were used to quantify habitat usage. All species significantly preferred smooth water sections of the river with trees on either one or both banks; P. pygmaeus also preferred smooth water with no trees. Bats avoided rough and cluttered water areas, as rapids may generate high‐frequency echolocation‐interfering noise and cluttered areas present obstacles to flight. In lower river regions, detections of bats reflected the proportion of suitable habitat available. At higher elevations, sufficient habitat was available; however, bats were likely restricted due to other factors such as a less predictable food source. This study emphasizes the importance of riparian habitat, bank‐side trees, and smooth water as foraging habitat for bats in marginal upland areas until a certain elevation, beyond which bats in these areas likely cease to forage. These small‐scale altitudinal differences in habitat selection should be factored in when designing future bat distribution studies and taken into consideration by conservation planners when reviewing habitat requirements of these species in Welsh river valleys, and elsewhere within the United Kingdom.

Foraging activity in bats is not uniform, with individuals favoring certain environmental characteristics over others. For example, Warren, Waters, Altringham, and Bullock (2000) and Todd and Waters (2017) found that foraging of both Daubenton's bats (Myotis daubentonii) and common pipistrelle (Pipistrellus pipistrellus) in upper Wharfedale (Yorkshire Dales National Park, UK) preferred river sections with smooth water and trees on both sides. This is in agreement with Lundy and Montgomery (2010), who also found that M. daubentonii favor smooth water areas.
Elevation is also thought to impact bat abundance and foraging efficiency.  found that foraging activity of bats in a riparian area of British Columbia decreased with increasing elevation, as did insect availability. Todd and Waters (2007) found that, at the same location as this study (Lledr River, Conwy, North Wales, UK), M. daubentonii switched feeding strategy with elevation.
At the lower elevation site, bats' primary hunting method was aerial hawking, regardless of aerial prey density, but at higher elevations, a switch to gaffing was observed. Dunn and Waters (2012) also found that the activity of P. pipistrellus and Pipistrellus pygmaeus declined with increasing elevation. Elevation has been found to play a significant role in bat community structure and could potentially be used to predict assemblage structure due to relationships between plants and elevation (Capaverde et al., 2018). Elevation may also influence sexual segregation-with male M. daubentonii recorded at higher elevations (Nardone et al., 2015); consequently, even smallscale changes in the environment can influence bat foraging ability (Salvarina, 2016).
Habitat modifications have been correlated with bat population declines on a number of occasions (e.g., Millon et al., 2018;Russo & Ancillotto, 2015), and identification of factors that influence bats substantially requires collection of quantitative data on the significance of each habitat type. Difficulties of studying flying bats at night in nonlinear habitats mean that published studies are limited in both size and number; consequently, there are numerous knowledge gaps for many vulnerable species. Fortunately, M. daubentinii, P. pipistrellus, and P. pygmaeus are known to forage along linear landscape elements, such as rivers and tree lines (Downs & Racey, 2006;Verboom & Huitema, 1997), rendering them relatively easy species to study through the night.
Data presented here were collected 20 years ago and so provide a useful baseline for which future studies can be compared to investigate potential effects of a changing climate. Alterations to bat detection, habitat use, reproduction, and migrations have been reported with increasing effects of climate change (Adams, 2010(Adams, ,2018Rebelo, Tarroso, & Jones, 2010;Sherwin, Montgomery, & Lundy, 2013); therefore, understanding environmental drivers linked to detections of species is vital for the informed development of management strategies. Results of this study can have significant management implications for these species in this region, and elsewhere in upland riparian valleys. This paper investigates the effects of three main factors thought to affect bat distributions: (a) water characteristics, (b) bank-side trees, and (c) elevation. We examine detections of M. daubentonii, P. pipistrellus, and P. pygmaeus in relation to these factors along a river gradient, as well as their time of emergence. Based on previous studies of habitat usage (Todd & Waters, 2017) and detections along altitudinal gradients (Walsh & Harris, 1996), we predict that bats are more likely to be detected along smooth water sections of river with trees along the banks and also that detections will decrease with increasing elevation.

| Study area
Data were collected between 14 July and 3 August 1999 along an 8.65 km stretch of the river (Afon) Lledr which runs through the Lledr Valley, North Wales, UK. The study area was located between Fairy Glen (53º03′N 3º48′W) and Pen-y-rhiw (53º02′N 3º56′W; Figure 1).
The upper reaches of the Lledr are characterized generally by smooth water with few trees. Beyond Roman Bridge, the river is bordered mostly by trees (mainly alder, Alnus glutinosa), thick riparian vegetation, and littered with large boulders and rapids. At Dolwyddelan, the river smooths out again and trees are scarce (similar to upper regions), but beyond the bridge at Pont-y-Pant it narrows, flowing through several gorges bordered by trees and riparian shrubs all the way to the Conwy confluence beyond Fairy Glen.
About 50% of the Lledr is bordered by riparian woodland, especially in the mid-and lower reaches. The Lledr is generally of similar width (ca. 6-10 m) along its entire length, with a few narrow stretches between 3 and 4 m. Water depth in summer rarely exceeds two meters, typically being 0.3-0.8 m deep.

| Habitat mapping
Length of the Lledr was mapped for the nine different habitat types presented in Table 1, as in the studies of Warren et al. (2000) and Todd and Waters (2017). Where possible, length of each habitat was paced and converted to meters; however, for some areas, pacing was not possible due to the difficulty of climbing over boulders and fences; therefore, for these areas, individual habitats were measured to the nearest millimeter from the 1:25,000 map enlarged to 1:12,000.
Water was classified as smooth, rapid, or cluttered. Smooth water was defined as lacking white water or riffles, rapid water had white water and heavy riffles, and cluttered water contained projecting rocks and riffles.

| Bat detections
Due to the precarious nature of the river borders, scattered usually with large boulders, barbed wire fences, and impenetrable scrub barriers, it was impossible to select transects of a set length.
Consequently, eight transects of varying length were selected which could each be walked in 30 min (Table 2).
Transect three (Rhiw Goch and Pont-y-pant) was divided into two 15-min transects (3i and 3ii) because part of the area was bisected by a forest which forced the observer too far away from the river to detect reliable signals with the bat detector. For treatment of the data and comparison of numbers of bat passes on the outgoing and return transects, transect 3i was treated as the outgoing transect (30 min total comprised of 15 min outgoing and 15 min return) and transect 3ii (30 min total) as the return.
Each transect was walked by the same trained observer twice commencing 14 July and terminating on 3 August 1999 totaling 16 nights for eight transects. Observations approximately covered bat lactation period of species concerned (Swift, 1980). At the transect location, the bat detector was switched on from sunset onwards (21:30 BST), to identify the first bat species to arrive at the transect location. Transects were selected randomly and were walked at a constant pace, commencing 30 min after sunset, on evenings with wind ≤2 ms −1 , with temperatures ranging from 10-21°C without heavy rain, when bat activity may be suppressed (Turner, 2002).

Elevation (m)
TA B L E 1 Major physical features of the Afon Lledr according to Warren et al. (2000) and Todd and Waters (2017)  Note. Trees were more than 5 m high with branches touching each other, that is, no gaps between adjacent trees; each tree was more than 5 m in width including branches.
The bat detector (Tranquillity II, Cheltenham) was always held facing the river to ensure all bat passes were recorded. Detection of Myotis and Pipistrellus bats using this detector are usually limited to 20 m or less (Adams, Jantzen, Hamilton, & Fenton, 2012), which ensures bats are only detected in the transect areas and not adjacent habitats. The highly directional response of similar designs of detector (Waters & Walsh, 1994) ensured that bats foraging over the river would not be detected at neighboring locations, and only bats using the river would be recorded. The detector, configured to record 40 ms in time expansion mode, was connected via the right channel to a stereo cassette recorder (Genexxa ® SCR-59). The highfrequency output from the detector was time expanded (10×) for computer analysis as recommended by Jones, Vaughan, and Parsons (2000). The bat detector microphone had a frequency response of ±20 dB, 12-200 kHz and the cassette recorder roughly ±3 dB, 400 Hz-12 kHz (Genexxa were unable to provide exact specifications).
Recordings were made on 90-min normal position tapes (TDK ® , Luxembourg D-IEC/type I). The tape recorder was held in the other hand together with the map of the transect for the identification of habitats. Each new habitat passed (together with time) was dictated into the internal microphone on the left channel of the tape recorder using identifiable distance markers such as houses, walls, and bridges.
When the transect was completed in one direction, it was walked in reverse commencing 1 hr after sunset. This method investigated the influence of start time on bat distributions. The transect was then replicated the following night.

| Data analysis
Outgoing and return transects were not pooled for analysis, as there may have been a danger of recording the same bat on the return transect as on the outgoing, if for example, some bats were territorial or had fixed feeding sites; consequently, outgoing and return transects were treated separately. Since length and representation of habitats Comparisons of numbers of bats on outgoing and return transects were tested using paired t tests. Non-normally distributed data (which included most bat pass counts) were normalized by (n + 1) log 10 T no.  (Neu, Byers, & Peek, 1974) to determine which habitats were used significantly more or less in proportion to availability as in the studies of Walsh, Harris, and Hutson (1995), Warren et al. (2000), and Todd and Waters (2017). Bonferroni confidence intervals were calculated on a custom written DOS program.

| Habitat mapping
With possible exceptions of cluttered water with trees on one side or no trees (habitats eight and nine respectively), all habitats along the Lledr were well represented (Table 3). Smooth water without trees (habitat three) closely followed by smooth water with trees on one side (habitat two) and then smooth water with trees on both sides (habitat one) were the most well-represented habitat types in the study area with rapids and cluttered areas covering fewer river stretches.
On examination of the proportion of habitats available to bats at different elevations (Figure 2), no single transect contained the full range of habitat types, the most representative being transect five (Ty isaf), and the least representative transect one (Fairy Glen).
There were more trees present lower down the river valley than higher up, and more varieties of water state with more rapids and cluttered water (Figure 2). In the midriver section, trees were present mainly on one side, with many smooth water and few cluttered and rapid habitats. At transect seven, there was still a fair proportion of smooth water with trees on either one or neither bank, but at the highest point along the river (transect 8), there was a high proportion of smooth water, but an almost complete absence of trees and cluttered and rapid water habitats.  Figure 3). The midriver section (transects 3-6) contained the most constant times recorded after sunset for M. daubentonii with bats being recorded the earliest at transects three and five.

| Habitat selection
More detections of each species were recorded over smooth water in both outgoing and return transects, particularly in areas with trees on one or both sides of the river (Figure 4). Pipistrellus pygmaeus was detected in the highest density (passes km −1 ) in smooth water habitats, but of the three species, it was also most commonly detected in rapid and cluttered areas.
A chi-square analysis was performed on bat passes in each habi-  (Neu et al., 1974). These determined whether each habitat was selected more, less, or in proportion to availability than would be expected from a random distribution.
Myotis daubentonii showed a significant (p < 0.05) preference for stretches of river with a smooth water surface with trees on one bank (  (Arlettaz, 1999).
Pipistrellus pipistrellus showed a mixture of preferences, generally selecting for smooth water with either trees on one or both sides and avoiding or showing no preference toward rapid and cluttered water stretches (Table 4). Pipistrellus pygmaeus again showed a mixture of preferences for habitat mainly selecting smooth water with trees (and in one case without trees) and avoiding rapid and cluttered water (Table 4).

| Bat detections
All three species of bat (M. daubentonii, P. pipistrellus and P. pygmaeus) were detected over the Lledr at some point in the early evening and the proportion of time spent in different habitat types varied between species.
Detection of M. daubentonii was reasonably high (n = 556 counts). Siemers, Stilz, and Schnitzler (2001) propose that M. daubentonii forage low over the surface of smooth water, as it enhances acoustic detection of prey. Todd and Waters (2017) also report that M. daubentonii from the same region as this study are detected most often over smooth water in areas with trees on one or both sides, and slightly less over smooth water with no trees.  (Jones & Rydell, 1994) and 40 min after sunset (Warren et al., 2000). (1) Fairy Glen (4) Dolwyddelan (2) Tan aeldroch F I G U R E 3 Time of first bat recorded after sunset at each transect, one (low altitude) to eight (high altitude). Transects include replicates n = 2 nights per transect (n = 16 nights) emergence (Racey & Swift, 1985;Rydell, Bushby, Cosgrove, & Racey, 1994;Todd & Waters, 2017).
Pipistrellus pygmaeus was the most frequently detected species (n = 808 counts) indicating that riparian habitat in this area may be important for this species. This result complies with findings of Barlow (1997), who suggests that this species is associated with water and insects which have aquatic larval stages. Habitat preferences for P. pygmaeus are considered to be different than those of P. pipistrellus, with the former detected more frequently in riparian habitats and the latter detected in a wider variety of habitats, as also reported by Walsh and Harris (1996). This study has revealed that the Lledr valley is a potentially suitable foraging area for P. pygmaeus, which in Wales, may be at risk from climate change, since higher winter temperatures between January and March have been linked to previously a smaller population size (Andrews, Crump, Harries, & Andrews, 2016).
The roughly equal activity of P. pipistrellus and P. pygmaeus on both outgoing and return transects is likely related to emergence time from roosts. For pipistrelles, mean emergence times after sunset have been reported to be between 21 min (Jones & Rydell, 1994) and 33.5 min (Bullock, Combes, Eales, & Pritchard, 1987). Transects in this study were begun 30 min after sunset, which suggests that the lack of difference in detections on outgoing and return transects is because the bats have already emerged and may already be at their foraging sites at the beginning of the transect. This is common for pipistrelles, which are often reported to roost close to major rivers (e.g., Speakman et al., 1991). Arrival times of P. pygmaeus at the river showed little variation (range 11 min), suggesting that this species might head straight for the river upon emergence. Pipistrellus pipistrellus arrival times were variable (range 48 min) suggesting that they either emerge later than P. pygmaeus or P. pipistrellus may head to other areas after emergence and not necessarily the river.
Previous studies by Walsh et al. (1995) and Walsh and Harris (1996) have measured mean number of bat passes per km in marginal upland areas to be 5.5 and 6, respectively. This study, which used the same methodology as Walsh and Harris (1996), recorded a much higher number of passes per km (mean ± SD of 39.6 ± 62.67) of all species using the return transect data (pooled for all transects). Figures as high as 160 bat passes per km for M. daubentonii (14.5 ± 28.96 bats/km) and 160 (5.8 ± 11.95) and 227 (18.5 ± 37.13) passes per km for P. pipistrellus and P. pygmaeus, respectively, were recorded in the area. All highest counts were recorded within the categories of smooth water with trees on both or one side, which further places importance on these habitat types within a marginal land class. These counts are comparable to those given by Warren et al. (2000) of a total mean of 20.1 passes per km for all bats, 8.9 passes per km for M. daubentonii, and 11.2 passes per km for P. pipistrellus, although their five passes of P. pygmaeus are too low to present as a mean value. Indeed, numbers of bat passes per km of all land classes in the Walsh et al. (1995) and Walsh and Harris (1996) studies are far lower than those recorded here and by Warren et al. (2000) in Wharfedale, Yorkshire, UK. In comparison with Wharfedale, the Lledr valley is more heavily wooded. There may potentially be more roosts available for bats, and therefore, the area may support a larger population of individuals than Wharfedale (if roosts, rather than food are a limiting factor), though no population estimates can be made from bat passes alone.

| Habitat selection
Distribution of all three species reflected a mixture of habitat availability, type, and elevation. Along the river Lledr, the only habitat types that bats selected for were the smooth water habitats and for each species, smooth water with trees on one bank accommodated a higher number of bat passes than would be expected in a random distribution (Table 4). Pipistrellus pygmaeus was the only species to select for smooth water with no trees on the return transect; however, this same habitat was selected against on the outgoing transect. Rapids and cluttered water categories were either used as available or selected against by each species. These findings are similar to the findings of Warren et al. (2000), Todd and Waters (2007), and Todd and Waters (2017).
In Suitable habitat (smooth water with trees on one or both sides) was readily available in transect seven, yet there were far fewer detections of M. daubentonii and P. pygmaeus than in transect 6 which was nearly 30 m lower in elevation. It may be that although sufficient habitat is available at higher elevation, other factors such as lower temperature and less predictable insect availability restrict bats in these areas (Capaverde et al., 2018;Dunn & Waters, 2012;Nardone et al., 2015;Todd & Waters, 2007). Indeed, foraging activity of bats and insect availability has been shown to decrease with increasing elevation , as has bat species richness (Patterson, Stotz, Solari, Fitzpatrick, & Pacheco, 1998); however, detections of P. pipistrellus were higher in transects six and seven than in lower transects.
de Jong and Ahlén (1991) found the distribution of pipistrelles to be correlated with local insect abundance, but there are conflicting reports of numbers of insects in relation to water smoothness. Warren et al. (2000) reported that numbers of insects increase with the smoothness of the water, whereas Todd and Waters (2017) found that insect distributions were independent of water surface state. Rydell, Miller, and Jensen (1999) reported that aerial insects were more abundant over ripples, so insect distribution could not explain why bats in this study avoided those habitats. Insect abundance has also been found to be unrelated to the presence of bank-side vegetation (Todd & Waters, 2007), which might explain why P. pygmaeus also selected for the smooth water category with no trees. Selection of trees during the twilight period may permit earlier emergence to exploit the dusk insect peak. Bats may use trees as part of a predator avoidance behavior when they may be particularly susceptible to predation by crepuscular birds (Mikula, Morelli, Lučan, Jones, & Tryjanowski, 2016;Speakman, 1991). Observations suggest that bats are most frequently attacked by raptors just after evening emergence and as they travel to and from the roost (Fenton et al., 1994;Mikula et al., 2016). This would explain pure selection of trees on outgoing transects (30 min after sunset) when it is lighter and selection of areas with no trees once the threat of predation has diminished.
Transect eight (the highest elevation) had plenty of habitat three (smooth water, no trees), but there were very few P. pygmaeus. Again, effects of physical variables associated with elevation (high wind, temperatures that are more variable, and less vegetation cover) might be overriding bats' ability to reliably utilize suitable habitat types.

| CON CLUS ION
A changing climate has been reported to cause alterations to bat detection, habitat use, reproduction, and migrations (Adams, 2010(Adams, ,2018Rebelo et al., 2010;Sherwin et al., 2013); therefore, while data presented in this study were collected 20 years ago, they provide a useful snapshot of the detection of bats in the Lledr valley during the summer period. This can be used as a baseline to which future data can be compared to investigate potential effects of a changing climate and or land development on these species. All species were recorded more in areas with moderate elevation and smooth water; however, there was suitable smooth water habitat available at higher elevations which was unoccupied. It is possible that bats could occupy these higher sites in future, especially if anthropogenic pressures, such as land development in other sections of the river, marginalize bats into using suboptimal habitats.
In marginal upland river valleys (up to ca. F I G U R E 5 Box plots of bat passes per km of the three bat species at each transect. Box represents interquartile range, and horizontal line is median

ACK N OWLED G M ENTS
The authors gratefully acknowledge the constructive comments of three anonymous reviewers and the assistance of landowners, farmers, and river bailiffs for providing access to the river. This study was performed under a three-year Natural Environment Research Council (NERC) grant with a Council for Advancement and Support of Education (CASE) stipend from Countryside Council for Wales (CCW) through the University of Leeds.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N
Victoria Todd involved in data collection, interpretation, and manuscript preparation. Laura Williamson involved in manuscript preparation.