Data collection

Data were collected between 2014 and 2017 across 14 sites in the south of England. The sites represented a range of habitat types. Most of the sites (n = 9) comprised agricultural land with dividing hedgerows. The remaining sites were rural sites with heterogeneous habitat including well established treelines, woodland, and/or watercourses (n = 3), or were green spaces within urban areas (n = 2). An Anabat Express bat detector unit (Titley Scientific, Ballina, Australia) was deployed at the study sites in rotation to record data across the sites for a total of 323 nights between sunset and sunrise. Deployment and positioning was carried out in a consistent manner at all sites with units mounted about 1.75 m above the ground adjacent to a suitable hedgerow or treeline to ensure detection of commuting and foraging activity along linear features. The units recorded data directly onto an SD card.

Post fieldwork, all data from the bat detectors were downloaded for sonogram analysis. The analysis was performed in AnalookW software (Titley Scientific, Ballina, Australia) developed specifically for Anabat detectors. Initially recordings were processed on a night-by-night basis and then data were subdivided into hourly units relative to sunset. This gave a total of 3424 hours of survey data over 323 nights, with each night of data being from a single site (i.e. sites were sampled independently not concurrently). Survey effort (number of survey nights per month and per site) is given in Supplementary material Appendix 1 Table A1. Species identification was carried out by assessment of the range and peak frequency, together with shape of each sonogram in terms of pitch and amplitude over time (Russ 2012).

Data relating to temporal and abiotic factors were collected for use as explanatory factors in statistical model-ling. The variables are explained in Table 1. Sunset, sunrise and lunar data were taken from time and date AS (< www.timeanddate.com>). Weather data were obtained via BBC (< www.bbc.co.uk/weather>) for the nearest town or using time and date AS using the nearest available weather station.

Table 1.

Temporal and abiotic data collected for use in statistical analyses. All data were hourly (n = 3242).

Statistical analysis

To examine whether there were significant deviations from a uniform distribution of bat registrations throughout the night, Kolmogorov–Smirnov two-sample tests were used as per Milne et al. (2005) for Australian bat species.

To explore the influence of spatiotemporal and abiotic factors on bat activity, generalised linear mixed models (GLMMs) were constructed. Models were developed for overall bat activity (total number of bat passes per hour regardless of species: n = 3424) and also the activity of each of the four most prevalent species/genus groups; common pipistrelle Pipistrellus pipistrellus, soprano pipistrelle Pipistrellus pygmaeus, Myotis spp. and Nyctalus spp. (specific bat passes per hour: n = 3424). In all cases, a full model was constructed whereby the factors listed in Table 1 were entered as continuous fixed factors (hour post sunset, temperature, wind speed) or categorical fixed factors (illumination, rainfall). Two random factors were also entered: site (coded 1–14 with no underlying rationale for the order and thus entered as a categorical random factor) and month (April–October). Because the dependent variable of bat activity (total or species-specific) used count data (number of bat passes per hour), a Poisson distribution was used with a log link function: this gave the lowest Akaike's information criterion (AIC) value (Akaike 1971) relative to other options for count data of Poisson with identity link, negative binomial with log link, and negative binomial with identity link. For the random factors, a scaled identity covariance type was specified as this covariance structure was associated with the lowest AIC score. To ascertain the effect of the fixed factors in explaining bat activity, marginal r² was calculated. To ascertain the effect of both fixed and random factors, conditional r² was calculated. The relative importance of the random factors can be inferred from the difference between conditional and marginal r².

Once full models had been computed for bat activity, reduced models were tested by dropping different combinations of fixed factors to establish whether the full model was optimal or whether a simpler model might better balance explanatory power and parsimony. Competing models were compared using delta (Δ) AIC on the basis that models with ΔAIC ≤ 2 had essentially have the same support and models with ΔAIC of 3–4 had strong support; models with AIC ≥ 5 were considered to have substantially less support and were discounted (Burnham and Anderson 2002). In all cases the full model was optimal and hence only full models are reported. All statistical analysis was carried out in IBM SPSS Statistics ver. 24.

Temporal distribution

Two sample Kolmogorov–Smirnov tests demonstrated that registrations of all species/genera differed significantly from a uniform distribution (p≤0.046 in all cases). The majority of bat registrations, regardless of species, occurred in the first hour post sunset and then decreased as the night progressed, with a small increase in activity towards dawn that made the overall activity distribution slightly bimodal (Fig. 1a). Both pipistrelle species also showed higher activity in the hours immediately following sunset (Fig. 1b–c), however, soprano pipistrelle alone showed an additional peak in activity towards dawn (Fig. 1c). The temporal distribution of Nyctalus registrations (Fig. 1d) was the most sporadic, with higher peaks in activity occurring haphazardly throughout the night. However, this species was recorded much less frequently than pipistrelle species or Myotis spp. and thus the variability in registrations, as shown by the standard error bars, was considerably higher. Myotis registrations were relatively infrequent in the first hour post sunset (Fig. 1e), but increased thereafter, peaking in the third hour post sunset and then decreasing, with a small peak in activity prior to dawn.

Figure 1.

Mean number of registrations in each hour post sunset for total bats and within species/genus. Error bars show the standard error.

Table 2.

Generalized linear mixed models exploring the influence of temporal and abiotic factors on overall bat activity (bat passes per hour) and activity for four specific taxa (n = 3242 survey hours across a total of 323 nights and 14 sites). The models used a Poisson distribution with a log link function. For the three continuous fixed factors – hour post sunset, temperature and wind speed – the gradient of any significant relationship with bat activity is given with the standard error below in parentheses. For the two fixed factors – illumination and rainfall – the estimated marginal mean (EMM) is given with the standard error below in parentheses. In all models, site (n = 14) and month (n = 7) were included as random factors. The importance of the fixed factors in explaining bat activity can be assessed using the marginal r squared value (r²_m), while the additional importance of the random factors can be assessed using the difference between r²_m and the conditional r squared value (r²_c). For more details of the variables, including the categories for illumination and rainfall, please see Table 1.

Spatiotemporal and abiotic influences on bat activity

Hour post sunset, temperature, wind speed, illumination and rainfall all had a significant effect on overall bat activity (bat passes per hour regardless of species) and the activity of the four focal taxa; the single exception was temperature for soprano pipistrelle, which was not significant (Table 2).

Hour post sunset was significantly negatively related to overall bat activity: bat passes per hour decreased by 0.147 per hour (± 0.002 SEM) as the night progressed. Similar negative relationships were seen for activity in all four specific taxa, with gradients varying between –0.051 ± 0.017 (Nyctalus) and –0.176 ± 0.004 (common pipistrelle). These relationships largely reflect the temporal pattern of bat activity decreasing throughout the night (Fig. 1), with the shallower gradients being for species with a notable pre-dawn peak in activity (soprano pipistrelle) or species whose activity was sporadic throughout the night (Nyctalus).

The relationship between temperature and overall bat activity was weakly positive, with bat activity increasing by 0.015 bat passes per hour (± 0.002 SEM) for each °C increase in temperature. A similar pattern was seen for common pipistrelle (0.022 ± 0.003), with a stronger positive relationship being found for Nyctalus (0.383 ± 0.017). A weak negative relationship was observed between temperature and Myotis (–0.029 ± 0.006). There was a significant positive relationship between bat activity and wind speed for overall bat activity and for activity of common pipistrelle, soprano pipistrelle and Myotis: all relationships were comparatively similar with bat passes per hour increasing by ∼0.067 (± ∼0.002 SEM) for each additional mile per hour in wind speed. The exception was Nyctalus where bat passes decreased by 0.111 (± 0.011 SEM) for each additional mph in wind speed. Bat activity, both overall and for each of the four focal taxa, was significantly lower in heavy rain. In the case of overall activity, bat passes per hour were fairly consistent in dry conditions and in light rain (5.668 and 5.114 bat passes per hour, respectively), but decreased substantially in heavy rain (2.628 bat passes per hour). This notable decrease in activity in heavy rain also occurred for both pipistrelle species. In contrast, Myotis declined linearly as rain intensified, while Nyctalus activity dropped substantially between dry conditions and light rain with activity levels light and heavy rain being approximately equal.

The impact of moon illumination on bat activity was more varied between taxa. Overall bat activity was significantly lower in instances of full illumination (2.934 bat passes per hour), than in partial or no illumination (4.002 and 4.838, respectively). The effect of moonlight on activity of both common and soprano pipistrelle was more gradual but remained negative. The effect on moonlight on Nyctalus and Myotis bats was less clear: partial illumination was associated with peak activity of Nyctalus and lowest activity of Myotis.

The random factors of site and month increased the amount of variance in total bat activity explained by the GLMM (r²_m = 0.369 versus r²_c = 0.474; a difference of 0.105). This demonstrates the importance of site-specific factors and seasonality on overall bat activity. For specific taxa, site and month varied in how much they influenced bat activity, with the difference between conditional and marginal r² values being negligible for Nyctalus (0.003) and low for soprano and common pipistrelle (0.055 and 0.061, respectively), but substantially higher for Myotis (0.231). It is important to note that the fixed factors (hour post sunset, temperature, wind speed, illumination and rainfall) together accounted for substantially more variation in bat activity than did the random factors of site and month in all cases.

Spatiotemporal factors

Bat populations differ across time and space suggesting that resource partitioning is important in facilitating the co-existence of multiple species (Arlettaz 1999). Within the United Kingdom, such partitioning has been observed previously between Pipistrelle species, which differ in foraging locations and feeding times (Nicholls and Racey 2006). Here, we found that random factors of site and month typically accounted for 12–22% of the variation in bat activity but were particularly important for Myotis, where they accounted for 45% of the variation explained by the GLMM model. This is likely driven by Daubenton's bats Myotis daubentonii, which are associated with water as they glean insect prey from the surface of lakes and ponds (Jones and Rayner 1988, Russ 2012) and were thus present in large numbers at some sites and absent from others. Seasonality (accounted for here by adding month as a random factor) is also likely to impact observed levels of activity. Temperate bats in the United Kingdom are most active during the summer months, foraging regularly to prepare for, or to recover from, the high energy demands of raising young (Racey and Speakman 1987, Ciechanowski et al. 2007). All bats remain active for the remainder of the summer and into early autumn to ensure they have sufficient energy reserves for winter hibernation (Speakman and Racey 1989). The time at which bats enter and emerge from hibernation is primarily dependent on temperature. They enter torpor when energy demands are higher than can be met by decreasing insect densities (Speakman and Racey 1989). Depending on ambient temperature, bats typically become active in April and seek out hibernation sites in late September as temperatures drop.

Bat activity was not uniform throughout the night, as shown by the clear patterns in the temporal distribution of activity across the night and reinforced by hour post sunset being significant in each GLMM. Bats emerge at different times (Russ 2012) and also commute different distances, at different speeds, between roosts and foraging areas. Some bats return to their roost part way through the night and then re-emerge for a pre-dawn feed and this likely explains the increase in bat activity shown towards sunrise shown here for soprano pipistrelle and Myotis. This has been seen to vary between nights and seasons (Anthony et al. 1981) and may be influenced by peaks insect densities at dusk and, to a lesser extent, at dawn (Rydell et al. 1996). Ultimately different temporal patterns in activity levels between species, as demonstrated here, plays an important role in niche partitioning in multi-species assemblages of insectivorous bats (Milne et al. 2005, Ciechanowski et al. 2007).

Weather

Temperature was weakly positively correlated with bat activity both overall and for common pipistrelle (i.e. more registrations in warmer conditions). This was expected given that bat activity tends to peak in the summer months, when temperatures are usually highest. The weak negative relationship between temperature and bat activity for Myotis and Nyctalus was more surprising. However, as surveys were undertaken between April and early October, when it is typically warm enough for bats and their insect prey to be active, one explanation is that temperature is important as a threshold, rather being linearly related to activity levels (Rydell 1989b). The notable pre-dawn peak for Myotis, when nightly temperatures are usually at their coldest, might also be a partial driver for this finding (and may also provide an explanation as to why there was no significant relationship between temperature and activity of soprano pipistrelle; the other species with a pre-dawn peak in activity). Rainfall was negatively correlated with bat activity in all cases (i.e. more registrations in dry conditions). This is consistent with previous findings that rain imposes an additional energetic cost and decreased prey abundance (Erickson and West 2002, Downs and Racey 2007). Wind speed was positively correlated with bat activity in all cases with the exception of Nyctalus. This finding is surprising given the potential for additional energetic costs posed by flight in strong winds (Norberg 1990). However, insectivorous bats, and specifically pipistrelles, are known to utilise linear features such as treelines and hedgerows to provide shelter when foraging in windy conditions (Verboom and Spoelstra 1999, Russ et al. 2003). This spatial shift in foraging activity might account for the increase in detected echolocation calls as the detectors used in this study were predominantly placed along linear features as is common in automated surveys (Collins 2016). In this way, automated survey results might be affected by the three-way relationship between the presence of linear features, detector placement and wind conditions.

Moon illumination

Previous studies on the effects of moonlight on bats have shown mixed effects. For example, Lang et al. (2006) found that activity of some insectivorous bats such as the whitethroated round-eared bat Lophostoma silvicolum in Panama to be lower on moonlit nights, while Appel et al. (2017) found bat activity was positively correlated with moonlight for Parnell's mustached bat Pteronotus parnellii and lesser sac-winged bat Saccopteryx leptura in Brazil. Here, we found that moon illumination was negatively related to bat activity. This agrees with work by Adam et al. (1994) on the Virginia big-eared bat Corynorhinus townsendii virginianus in the US but contrasts with previous work on non-British Myotis species, which did not find a link between activity and moonlight (Negraeff and Brigham 1995, Hecker and Brigham 1999) – although it is notable that neither of these studies included the modifying effect of cloud cover on illumination.

Although it has been suggested previously that bats at higher latitudes are exempt from the predation pressures that impact tropical species (Karlsson et al. 2002), predation risk on bright nights could still be an important modifier of activity in temperate species (Lima and O'Keefe 2013). It has also been suggested that temperate insectivorous bats may seek more enclosed (shaded) habitats when foraging in bright moonlight (Reith 1982, Erickson and West 2002), such that an apparent decrease in activity in open areas nights might actually be a repositioning of foraging activity spatially. We therefore suggest that the bats in our study might be avoiding bright moonlight conditions because of an increased risk of predation, either real or perceived. Moreover, it is notable that Myotis and Rhinolophus bats in Europe have previously been found to have an aversion to artificial illumination (Rydell 1992, Stone et al. 2009), which again was provisionally attributed to predation risk.

Implications and recommendations

Bats comprise an important, and legally-protected, part of mammal fauna in the UK. Surveying and monitoring is important in the contexts of scientific research, conservation, assessment of ecosystem health, monitoring progress towards sustainable development goals, and in compliance with legislation on development planning and infrastructure construction (Jones et al. 2009, Barlow et al. 2015, Collins 2016). It is thus vital that the factors which underpin bat activity, and thus detectability in acoustic surveys, are clearly understood. Bat surveys are notoriously difficult to standardise in terms of timing and the abiotic conditions under which they are conducted and only with robust understanding of optimal foraging conditions is it possible for this to be achieved.

We recommend that automated fixed-point surveys are undertaken throughout the night where possible (where this is not possible, they should be conducted for 4 h post-sunset and 2 h pre-sunrise to ensure peak activity times for all species are covered). As long as bats are active, temperature is largely immaterial but nights with heavy rainfall should certainly be avoided. Wind speed should also be taken into account, as linear features might be preferred habitat when shelter is sought from the wind, potentially increasing estimates of activity if detectors are placed close to such features. Surveying during high summer gives the simultaneous advantages of higher activity and greater concentration of activity as nights are shortest. We recommend that given increasing urbanization, the effect of light on bats should be further investigated, and that such research take cognisance of cloud cover as well as lunar phase (Stone et al. 2009, Russo et al. 2017). Given that the findings of this research indicate that overall bat activity decreases when the moon is unobscured by cloud, there remains potential for light from artificial sources to also impact bat activity. Passive monitoring of light levels in the field during surveys might be particularly helpful in such research.