The moth fauna is more diverse in the understorey than in the canopy in a European forest

Abstract

The canopy of forests as the ‘last biotic frontier’ has often been neglected in insect biodiversity studies because it is harder to access compared to the understorey, even in relatively well-known temperate ecosystems. We investigated the diversity, abundance, and body size patterns of macromoths (Lepidoptera) in the canopy and understorey in a central European deciduous forest. We collected moths at two sites during 19 trapping nights and three lunar phases between July and September 2021 using a weak ultraviolet light emitting diode (LED) lamp (LepiLED mini). Overall, we captured 4368 individuals (165 species) from 11 families. Based on a number of metrics, richness and diversity was significantly lower in the canopy than in the understorey. Non-metric multidimensional scaling ordinations show that communities largely overlap, but the proportion of species that only occur in the understorey was higher. While Noctuidae and Erebidae species were abundant in both strata, Geometridae species were most frequently observed in the understorey. We identified 16 indicator species for the understorey but only three for the canopy. Forewing length of moths in the canopy was on average 1.7 mm longer than of those in the understorey. Overall, the understorey is far more important for moths than the canopy in a temperate forest. The canopy is dominated by fewer and larger species and probably has a higher proportion of dispersers.

Introduction

Erwin (1983) described the canopy of tropical rainforests as the ‘last biotic frontier’. Canopies are ecologically important and structurally complex systems, which contain all leaves, twigs, and branches that are out of reach of the ground (Parker, 1995; Moffett, 2000; Ødegaard, 2000). The canopy fauna includes all organisms harboured in the canopy of trees with a breast height diameter of at least 10 cm (Ødegaard, 2000). The canopy has been studied less than the understorey (Basset et al., 2012), probably because the canopy is much more difficult to access than the ground-level understorey.

The canopy insect fauna was expected to be twice as diverse as the understorey (Erwin, 1982), but this estimate has been critically debated (e.g. Ødegaard, 2000; García-Robledo et al., 2020). Available studies on tropics show that the canopy indeed tends to have a greater insect species diversity and higher abundance compared to the understorey, at least seasonally (DeVries and Walla, 2001; Neves et al., 2014). However, geometrid moths tend to be more diverse in the understorey compared to the canopy (Beck et al., 2002; Brehm, 2007).

Lepidopterans are well suited for comparative studies because they are easy to survey using standardised methods such as light trapping (Wölfling et al., 2016; Infusino et al., 2017). Furthermore, they can be indicators of the vegetational community composition (see Erhardt, 1985; Holl, 1996). Vertical stratification in various Lepidopteran clades has been observed at sites of various elevations and latitudes (Ashton et al., 2016). The degree of vertical stratification varies considerably between the different clades of Lepidoptera (see Intachat and Holloway, 2000; Schulze et al., 2001; Brehm, 2007). These stratifications likely differed due to biogeographical differences across the locations and variation in resource availability. Thus, existing studies suggest taxon-specific small-scale vertical diversity patterns in the tropical regions.

Temperate forests differ from tropical forests in many aspects, including their significantly lower species diversity (Ricklefs, 1989). A major difference is that tropical canopies have a high diversity of epiphytic plants and lianas (Hegarty and Caballéley, 1991). Although many epiphytes tend to be low in nutrients, they are important microhabitats for a diverse fauna, and different epiphyte species host ecologically diverse arthropod communities (Stuntz et al., 2002). A comparable diversity of epiphytes is not present in forests of central Europe north to the alps where epiphytes are rare and restricted to particular sites, e.g. with thick moss cushions (Zotz, 2002, 2005). Thus, central Europe lacks both the structural diversity and food resources that epiphytes can provide.

Previous studies in temperate forests show that the canopy is not as important for biodiversity as the canopy of tropical rainforests. In Missouri, le Corff and Marquis (1999) found no differences in densities of herbivorous insects on two oak species (Quercus spp.) between the strata, but species diversity was significantly higher on both tree species in the understorey. In Mordovia, Russia, Ruchin (2023) showed the highest abundance of Coleoptera and Diptera in the understorey. Hymenoptera, Dermaptera, Neuroptera, and Trichoptera were most abundant in the canopy. For Hymenoptera, the species richness between canopy and understorey was the same in New York State, USA (Urban-Mead et al., 2021). But they had a higher diversity as well as a higher proportion of female bees.

For lepidopterans, the species diversity in deciduous forests has been shown to be higher in the understorey compared to the canopy, for example in studies in Virginia and Japan (Hirao et al., 2009; Seifert et al., 2020). In Europe, several studies found contrasting patterns in abundance in different families. Geometridae were more abundant in the understorey in a Belgian mixed deciduous forest (de Smedt et al., 2019), but in Saxonian deciduous floodplain forest, Germany, they occurred more frequently in the canopy (Fröhlich et al., 2007). In Belgium, Noctuidae were similarly diverse in both canopy and understorey (de Smedt et al., 2019), whereas in Saxony they occurred only in the canopy (Fröhlich et al., 2007). In Saxony, Noctuidae dominated all strata with respect to species richness and diversity, but they were less abundant in the understorey. Other studies did not find a distinct moth communities or indicator species in the canopy in Bavaria, Germany (Hacker and Müller, 2008) or detect a maximum activity for any moth species in the canopy of Hainich National Park, Thuringia, Germany (Erlacher et al., 2009).

Causes for the stratification of herbivorous insects might include adaptations to differences in the nitrogen content of leaves (le Corff and Marquis, 1999). Other possible causes include differential exposure to sunlight (Basset, 1991) and microclimatic conditions (e.g. Rytteri et al., 2021) as well as seasonally varying levels of secondary metabolites (Murakami et al., 2005).

The body size and shape of Lepidoptera can also show an adaptation to different conditions between the canopy and understorey: for example, more vigorous, fast-flying butterfly species were found in the canopy (Graça et al., 2017; but see Schulze et al., 2001). In contrast, slimmer and slower species tend to prefer the understorey. DeVries et al. (2010) showed a significant change in wing shape for morpho butterflies at the transition to the canopy. However, insect size has overall rarely been compared between the forest canopy and understorey so far.

Here we studied diversity, abundance, and body size of moths at two sites in the canopy and understorey in a forest in Central Germany. We tested the following hypotheses:

• • In the canopy, we expect a lower diversity of adult moths than in the understorey, because caterpillars of most species find their food plants in the shrub and herbaceous layers.

• • In the canopy, we expect a higher proportion of larger moths than in the understorey, because their wing loading is lower, and this allows them to be more robust in the windier conditions in the canopy (Gabey et al., 2010).

Materials and methods

Study area

We conducted our study in the nature reserve TH-No. 452 ‘Jenaer Forst’, west of the city of Jena. The area covers 541 ha and forms the eastern border of the flora-fauna-habitat area No. 127 ‘Jenaer Forst’. The subsoil is a highly decomposed shell limestone plateau with low water storage capacity (Heinker et al., 2019). The vegetation is dominated by Fagus sylvatica L. and Quercus robur L. belonging to the oak-hornbeam forest vegetation type, but semi-arid grasslands of anthropogenic origin occur as well (Heinker et al., 2019; Singh et al., 2022). The study area is in the temperate climate zone in the transition from oceanic to continental climate (Hiekel et al., 2004). Precipitation is 550–600 mm year⁻¹ and the annual average temperature is 9–10°C. We collected moths at two locations (Supplementary file S1, fig. S1): at Bismarckturm (N 50.9293, E 11.5595, 328 m), and at Forstturm (N 50.9227, E 11.5548, 348 m). The Bismarckturm is 20 m high and the Forstturm is 21 m high; the surrounding forest was a maximum of ca. 16 m high. Collection sites were 800 m apart in a direct line. We chose these sites with the towers because they provided easy access to the canopy (see de Smedt et al., 2019). Both towers are more than 100 years old and surrounded by a dense forest cover. The sites were not directly affected by artificial light at night, but the nearby city emitted sky glow at night, which we did not expect to significantly affect our results as the city is 500–600 m away from our sites.

Traps and sampling

The traps (Insects & Light, Jena, Germany) consisted of a roof, three vanes, and a funnel, as well as a net to collect the moths (fig. S2). The trap was made of white 1.5 and 3 mm thick polypropylene, respectively, and the vanes were 30 cm long (fig. S2). This basic trap design was described by Brehm (2007), among others. The same type of vane trap was used by Singh et al. (2022), who tested different trap types at the same time and place.

We used ultraviolet (UV) radiation to attract moths (Fabian et al., 2023) and specifically a LepiLED mini lamp (Insects & Light) that is attracting moths only from short distances (Niermann and Brehm, 2022). The lamps are equipped with eight power light emitting diodes (LEDs) with emission peaks at 368 nm (UV), 450 nm (blue), 530 nm (green), and 550 nm (cold white) (Brehm, 2017). Thus, it covers those parts of the spectrum to which the photoreceptors of moths and other insects, in general, are particularly sensitive (van der Kooi et al., 2021). We used 26 Ah power bank batteries as the power source for the lamps and timers (Lucstar, USA) controlled the switching.

We conducted trapping from July to September 2021. We chose three phases around the new moon in the lunar cycle to minimise the negative effect of a full moon on catches (Nag and Nath, 1991). Each of these lunar phases consisted of three sampling events at the two sites. We always conducted trapping at each site simultaneously in the canopy and understorey, resulting in a pair of samples collected in parallel. In total, we collected 38 samples, which included two additional samples, accounting for a technical problem with one trap in one night (Singh et al., 2022).

A catch started after dusk, at 21:30 in July (lunar phase 1), at 21:00 in August (lunar phase 2), at 20:30 in September (lunar phase 3), and lasted between 7 and 9 h. In the canopy, we always trapped at the same position on the towers, namely Bismarckturm and Forstturm. In the understorey, there were three positions per site: U1, U2, and U3. These were arranged in a triangular shape (fig. S1) and matched the positions of Singh et al. (2022). The samples were identical to those of the white vane traps tested by Singh et al. (2022). In the understorey, all traps hung between 1.5 and 2 m in height. The trap in the understorey hung first at position U1, then at U2, and on the third night at U3. We chose this rotation to exclude a site effect during the nights. The positions were each 20–25 m apart.

We emptied the traps in the morning after a night of trapping and released most of the immediately identifiable species in the field. We generally collected at least one voucher specimen of each species from each sample. In total, we prepared, labelled, and permanently integrated 588 individuals (14% of all captured moths) into the collection of the Phyletisches Museum Jena (PMJ). We determined these individuals with Steiner et al. (2014) and in comparison with the reference collection in the PMJ.

Body size measurements

From the collected moths, we measured the forewing length (FWL) from the wing base to the apex (Miller, 1977). We carried out measurements on spread individuals with a digital caliper (Powerfix Profi+) with 0.1 mm accuracy. For species with up to ten individuals, we measured all individuals; for more common species, we measured ten randomly selected individuals. We then calculated the arithmetic means of each species. The inclusion of all species, in particular those with a few or even only one individual is strongly recommended (Beck et al., 2023).

Statistical analyses

We performed statistical analyses in R (v 4.1.2, R Core Team, 2021) and R Studio (v 1.4.17, RStudio Team, 2020). We generated the species accumulation curve using the iNEXT package (Hsieh et al., 2016). Indicator species analysis was performed using the indicspecies package (de Cáceres and Legendre, 2009), to identify species strongly associated with either canopy or understorey. The multipatt function was used to calculate the indicator value for each species in the dataset, based on its relative frequency and abundance in each stratum. We used the pruneindicators function and permutation (n = 999) test to filter erroneous results and allow a suitable comparison of paired data. For diversity analyses, we used the vegan package (Oksanen et al., 2020). We calculated observed species richness per strata (canopy vs. understorey) and four species richness estimators: Fisher's α (Fisher et al., 1943), Chao estimator (Chao, 1987), Jackknife 1-index (Zahl, 1977), and bootstrap estimator (Burnham and Overton, 1979). We performed ordination analyses using non-metric multidimensional scaling (NMDS) and tested the ordinations for differences between strata using analyses of similarities (ANOSIM). We adjusted the P values of these analyses by using a false-discovery rate control by Benjamini and Hochberg (1995). We calculated beta diversity using the adespatial package (Dray et al., 2022) and considered the components ‘nestedness’ and ‘turnover’ (Baselga, 2010) of the Sørensen index of dissimilarity (Sørensen, 1948).

We used a generalised linear mixed model (GLMM) to test for differences in the average body size of moths across strata. Firstly, to test for normality in the distribution of count data and FWLs, we used the bestNormalize package (Peterson, 2021) and conducted a square root transformation. However, the transformed data were identified to be non-normal by a Shapiro test. Thus, a generalised linear model with identified overdispersion and non-normal distribution could not be used. We examined the QQ-plot of standardised residuals, tested overdispersion, and validated the model properties using the MASS package (Venables and Ripley, 2002). Finally, we used a GLMM with a negative binomial distribution, using the lme4 package (Bates et al., 2015). The negative binomial distribution is often used with a log-link function, which helps with data transformation and also accounts for the overdispersion parameter. The emmeans package (Lenth, 2022) was used to look at the different contrasts and back-transformed values from the model summary. Our GLMM model included body size (FWL) as the response variable and stratum levels, location, and moon phase as fixed effects. The trapping date was added as a random effect. A test with the DHARMa package (Hartig, 2022) did not show overdispersion. We compared this approach with a model that used gamma distribution which is usually more appropriate for continuous variables. Both models showed the same result for the comparison between the strata. The final model was chosen based on AIC (Akaike Information Criterion) values and after visualising the effect plots to evaluate the model fit.

To visualise the distribution frequency of FWLs at the moth family level, we created two dual-axis stacked histograms per stratum, for both individuals and species. The bin size in the resulting plots was calculated following Freedman and Diaconis (1981). We visualised all results using the ggplot2 package (Wickham, 2016) and Affinity Designer (version 1.10.21).

Results

Moth abundance and diversity

In 38 nightly catches we captured a total of 4368 individuals of 165 species belonging to 11 families. In total, 2749 individuals (126 species) were collected at Bismarckturm: 1552 individuals (74 species) in the canopy and 1197 individuals (110 species) in the understorey. At Forstturm we caught 1619 individuals (129 species): 726 individuals (65 species) in the canopy and 893 individuals (99 species) in the understorey. The number of individuals caught per sample ranged from 20 to 502 (fig. S3). In the first lunar phase we caught a total of 1965 individuals, in the second 652, and in the third 1751 (fig. S3). In the canopy, we captured 2278 individuals (89 species), and in the understorey 2090 individuals (141 species, fig. 1A). The species accumulation curve also shows significant differences between the strata, with the understorey being more species-rich than the canopy (fig. 1A). This is also true for the Shannon index (fig. 1B), which gives less weight to singletons. Furthermore, the sample coverage was high for both strata (canopy: 98.9%, understorey: 97.6%, fig. S4). Species richness (observed and using different estimators) was consistently higher in the understorey than in the canopy (table 1).

Figure 1. Species accumulation curve for canopy and understorey samples based on (a) species richness (q = 0) and (b) exponential Shannon index (q = 1). The curve represents sample size (solid line) and extrapolated species numbers (dashed line) with 95% confidence interval (shade).

Table 1. Species richness (observed and estimated) and diversity measures of moth assemblages in the canopy and the understorey for all sites and sampling nights combined

In each lunar phase, moth diversity was typically higher in the understorey than in the canopy (table S3). We detected more species in the canopy than in the understorey in only seven out of 19 pairs of samples (table S4).

Twenty-four species were exclusively found in the canopy, 225 species were exclusive to the understorey and 65 species were present in both strata. There were 13 singletons in the canopy and 37 in the understorey. We detected a total of 19 indicator species for one stratum: three for the canopy and 16 for the understorey (table 2).

Table 2. Indicator species for the canopy and the understorey for all sites and sampling nights combined

Significant P-values are shown in bold.

The three families with the most collected individuals were Noctuidae (2545 individuals), Erebidae (1375 individuals), and Geometridae (303 individuals, tables S1 and S2). Across all samples, the five most common species were Eilema lurideola (Zincken) (Erebidae, 1039 individuals), Noctua pronuba (L.) (Noctuidae, 1021 individuals), Xestia c-nigrum (L.) (Noctuidae, 549 individuals), Noctua fimbriata (Schreber) (Noctuidae, 238 individuals), and Eilema complana (L.) (Erebidae, 155 individuals). A complete species/site/stratum table is provided in the Supplementary file S2 and our raw data are available in Supplementary file S3.

Moth community composition

The NMDS ordination contains 38 data points representing all samples from both sites of both strata (fig. 2A). It shows that the samples are primarily temporally partitioned, with the second lunar phase samples standing out from those of the first and third lunar phases. Along the second axis, samples were differentiated according to the stratum. The canopy communities of the two different sites were more similar than the communities of the canopy and understorey of a given site, respectively. Bismarckturm and Forstturm did not differ in ANOSIM with grouping by site (R = −0.015; P adjusted = 0.570). The two strata, on the other hand, differed significantly (R = 0.112; P adjusted = 0.039). Data separated by Bismarckturm and Forstturm showed no significant differences in ANOSIM with grouping by stratum (Bismarckturm: R = 0.035; P adjusted = 0.301; Forstturm: R = 0.104; P adjusted = 0.171). Each lunar phase showed a separation by stratum as the polygons are separated (fig. 2B). However, only the first and the second lunar phases differed significantly in ANOSIM with grouping by the strata (lunar phase 1: R = 0.540; P adjusted = 0.003; lunar phase 2: R = 0.401; P adjusted = 0.007). The third lunar phase did not show significant differences in ANOSIM with grouping by strata (lunar phase 3: R = 0.265; P adjusted = 0.071).

Figure 2. NMDS ordinations of the samples from the Jenaer Forst (a) for the combined dataset and (b) separated by lunar phases (1–3). Blue, canopy, green, understorey, circle, Bismarckturm, triangle, Forstturm, star, centroid, arrow, post-hoc vector of dates. The date given in each case represents the start date of the sampling event in 2021.

Considering the combined data of both sites, beta diversity measured by the Sørensen index between the canopy and understorey was 0.25 (fig. 3). Turnover accounted for a much larger portion compared to nestedness. Beta diversity values of the individual sites ranged from 0.27 to 0.34 with the lowest measured turnover at Bismarckturm. Beta diversity values for the three lunar phases ranged from 0.19 to 0.35 with the largest measured turnover at lunar phase 2.

Figure 3. Beta diversity as the Sørensen index of dissimilarity between the canopy and understorey, classified by nestedness and turnover for all samples and separated by site and lunar phases.

There was a change in the composition of species communities at the family level with time (fig. 4), with the understorey following a different trend than the canopy. Noctuidae dominated the canopy in terms of number of individuals (table S1). They formed almost 50% of the individuals found in the first two lunar phases and almost the entire community in the third lunar phase (fig. 4). This dominance in lunar phase 3 was caused by the extremely abundant species N. pronuba (735 individuals) and X. c-nigrum (512 individuals). In the understorey, both the proportion of Noctuidae (fig. 4) and their number of individuals (table S2) also increased with time. However, the number of individuals was roughly four times greater in the canopy than in the understorey during the first lunar phase (tables S1 and S2).

Figure 4. Composition of species communities at the family level in the three lunar phases separated by individuals and species as well as canopy and understorey. Families are colour coded.

Erebidae occupied almost the same proportion as Noctuidae in the canopy in lunar phases 1 and 2 (fig. 4). In the understorey, Erebidae dominated with a share of more than 50%. Here E. lurideola was particularly abundant (1024 individuals in moon phase 1). The number of individuals of Erebidae decreased with time in both strata (tables S1 and S2). Geometridae were represented in the canopy by a very small proportion (fig. 4). In the understorey, the number of individuals was five times greater in lunar phase 1 and nearly 11 times greater in lunar phase 3 compared to the canopy (tables S1 and S2).

Body size

Across all samples, the most common FWL size class of moths was 24.2–25.3 mm in the canopy (fig. 5) and was dominated by Noctuidae. This size class showed the largest difference in abundance when compared to the understorey. Conversely, the understorey peaked in abundance for the 14.3–15.4 mm class, with a dominance of Erebidae. The abundances in both strata for this size class were roughly equal. Noctuidae and Geometridae covered the entire size spectrum except for the upper end (>26.4 mm).

Figure 5. Dual-axis stacked histograms representing the FWLs of moth species (right) and individuals (left) in the canopy and understorey. FWL are shown in mm. Dashed lines represent the estimated marginal means for canopy (19.4 mm) and understorey (17.7 mm). Different colours represent the moth families. Bin size = 1.1 mm.

A Mann–Whitney U test between the two strata showed different trends in the strata (W = 87,012; P < 0.001). The test for contrasts between strata showed a difference of 1.69 mm in body size (FWL) (P < 0.0001, table S5). We were able to visually confirm these results by a boxplot of the body sizes of all captured individuals (fig. S5).

The mean values of the individual trap pairs converged over time (fig. S6). With few exceptions, this was also true for the medians. The interquartile ranges of the FWL of each sample show that the variation in body size decreased only slightly within the observation period in the canopy (fig. S7). In the understorey, on the other hand, the variation was very small at the beginning but approached a similar value as the canopy over time (fig. S8).

Discussion

Our study shows that the canopy of a forest in the temperate region had a significantly lower richness and diversity of moths than the understorey. The low species richness in the canopy confirms previous studies, e.g. by Fröhlich et al. (2007) or Hacker and Müller (2008) who found fewer individuals and species in the canopy than in the understorey. In our study, we found fewer species in the canopy, but some were extremely common such as N. pronuba or E. lurideola.

The study shows that there is a pronounced vertical stratification in moths. For example, Geometridae and Noctuidae showed contrasting distribution patterns among the strata, observed in other studies before (Brehm, 2007; Hirao et al., 2009; de Smedt et al., 2019). We detected high turnover values of beta diversity, indicating that many species in the understorey were replaced in the canopy. The species accumulation curve shows the impoverishment of species in the canopy compared to the understorey.

Geometridae were the dominant group among the understorey specialists. They represented nine of 16 species highly associated with the understorey and accounted for a large proportion of individuals in the lower stratum. This supports previous observations of Geometridae as understorey specialists. For example, Brehm (2007) found a significantly higher diversity in the understorey of geometrid moths in Costa Rica than in the canopy. Only three indicator species were found in the canopy – and all were Noctuidae. Interestingly, two have caterpillars mostly feeding on host plants in the understorey (or in more open habitats) (table S6). This clearly suggests that the observed adult moths of these species (and probably others found in the canopy) most likely dispersed between other sites (and habitats) or performed mating flights in search of females. For dispersing moths, it is certainly easier to fly above the canopy than in the forest.

How can the observed body size patterns be explained? All three canopy specialist species are above-average body sized and can certainly be regarded as better and faster fliers than most of the understorey species (Nieminen, 1996; Graça et al., 2017). N. fimbriata (FWL 24.4 mm) is capable of flying long distances and is generally known as a migratory moth (e.g. Fröhlich et al., 2007); the same is true for Agrotis exclamationis and for the common N. pronuba (L.) (Hächler et al., 2002). Slade et al. (2013) measured dispersal distances in N. fimbriata of up to 13.7 km within 2 months. For Noctuidae in general, mobility depends partly on hostplant specialisation (Betzholtz and Franzen, 2011). On the other hand, not all large species and strong flyers were primarily observed in the canopy. The pine hawk moth Sphinx pinastri L. is an excellent flyer and thus a strong potential candidate species expected to be associated with the canopy. The caterpillar hostplants are tree species such as Pinus spp. and Picea spp. (Robinson et al., 2023), thus with abundant food resources in the canopy. However, S. pinastri turned out as an indicator species of the understorey. This could be attributed to the occurrence of their important sphingophilous nectar plants such as Lonicera spp. and Platanthera spp. which grow in the understorey and which S. pinastri can locate olfactorily even in dark nights (Tinbergen, 1969).

Species communities changed over the course of the three lunar phases and different species, or families dominated each phase (fig. 4). Within each lunar phase, each NMDS ordination showed a separation of species communities by canopy and understorey. This separation was confirmed by the ANOSIM only for the first and the second lunar phases. The ANOSIM also shows that the differences in species communities between strata became weaker with time. The weaker differentiation could possibly be explained with the decreasing abundance of Geometridae, many of which appear to be potential understorey specialists.

Our study was restricted to a period between July and September. We acknowledge that we have missed certain species with early and late flight periods. For example, several notodontid moths fly in spring and early summer while some noctuid and geometrid species are on the wing in autumn and winter. Furthermore, we exclusively used light for moth attraction. Thus we missed moth species that can only be sampled by other techniques (e.g. Kalinova et al., 1994). Additionally, Preisser et al. (1998) have shown that malaise trapping can also be used for testing stratification patterns of arthropods in general in temperate forests.

The canopy sampling was done slightly above the crown and not in the canopy; potentially attracting a higher proportion of dispersing moths as opposed to the understorey. Future studies may also include the full season, and consider other abiotic factors such as wind strength or temperature in the modelling, because they can have an impact on the number of individuals caught per night (e.g. Fröhlich et al., 2007). Our data showed considerable temporal variation; it is therefore important to conduct similar studies over as long a period as possible in habitats with strong seasonal change. Analyses of additional morphological traits such as proboscis length could provide further insight into the ecology of vertical stratification in forests.