High Frequency of Apodemus Mice Boosts Inverse Activity Pattern of Bank Voles, Clethrionomys glareolus, through Non-Aggressive Intraguild Competition

Simple Summary Ecologically similar animals can only co-occur if resources are used in a sufficiently diverse manner. The necessary separation can be made according to habitat, food choice and activity pattern. We investigated this question in an inner-alpine mixed forest for bank voles, Clethrionomys glareolus, and mice of the genus Apodemus, the latter being larger and stronger. In the presence of many Apodemus mice, bank voles were indeed more diurnal. We found no alternative explanation for this temporal partitioning, despite considering possible other influencing factors, such as exposure to lunar illumination, temperature, precipitation, the presence of predators or the availability of cover. Our video footage showed that this temporal separation hardly requires any physical interaction between the small forest rodents. We attribute this to the long common evolution of these animals. Abstract Sympatric animals with similar requirements can separate their ecological niches along the microhabitat, food and time axes. There may be alternative reasons for an interspecific different activity pattern, such as intraspecific social constraints, predator avoidance or physical conditions such as temperature, precipitation and illumination. We investigated the importance of intraguild competition in a 2-year study in an inner-alpine mixed forest, using small forest rodents as our model species. Apodemus mice were the physically superior, and bank voles, Clethrionomys glareolus, the inferior competitor. We predicted that bank voles would exhibit increased diurnal activity when frequencies of the almost exclusively nocturnal Apodemus mice were high during a seed mast year. To investigate this, we recorded 19,138 1 min videos. Controlling for confounding variables, bank vole diurnal activity was significantly related to the frequency of Apodemus mice. We assume that at high densities of Apodemus mice, a purely nocturnal separation in the niche dimensions of time, habitat and microhabitat is no longer sufficient, and therefore an inverse activity pattern by the bank voles is reinforced. Our videos showed, however, that this does not require persistent aggressive meetings and we explain this by the long co-evolution of the taxa under study.

Competition among species is a widespread phenomenon [13,14], and long-term coexistence is only possible if at least one important niche dimension differs. Dimensions of resource partitioning are mainly habitat, food and temporal niche [15,16]. For bank voles

Species Identification
While bank voles are fairly easy to identify visually on appropriate videos, for example by the ratio of body to tail length or the size of the ears [29], we have not attempted to distinguish members of the genus Apodemus. In the latter, identification problems can occur without calibration to the morphology of the local population at hand [57], and even genetic identification has proven difficult [58][59][60]. As Apodemus mice were typically silent, identification by means of the calls was also ruled out [61]. Due to the well-known habitat requirements [28,62], we expected the occurrence of the yellow-necked mouse, Apodemus flavicollis, and of the wood mouse, Apodemus sylvaticus, as most probable in our forested study area, in an unknown numerical ratio. The alpine mouse, Apodemus alpicola, is absent from the region according to current knowledge [63].

Study Area
The camera traps were set up on the mountain range of the Ossiacher Tauern (46,692 • N; 14,067 • E, 550 m a. s. l.) in Carinthia, Southern Austria. In the two years of the survey, the mean annual temperature was 8.88 • C and 9.39 • C, respectively, and the total annual precipitation was 1159 mm and 749 mm, respectively (meteorological station of Feldkirchen: https://data.hub.zamg.ac.at/ (accessed on 15 October 2022)). The study area was on the steep lower slope with a distinctly undulating relief and several small rocks. The concave areas have moist and humus-rich soils (sandy loams over clay of the pseudogley type), where a lime-maple noble deciduous forest (Tilio-Acerion) would be the primary vegetation. In the convex parts, beech forest types would predominate plant-sociologically (Melampyro-Fagetum, Galio odorati-Fagetum [64]). European beech, Fagus sylvatica, as the actual primary tree species has been pushed back, and today Norway spruce, Picea abies, dominates the forest. However, European beech, limes, Tilia platyphyllos and T. cordata, sycamore, Acer pseudoplatanus, European ash, Fraxinus excelsior, as well as silver fir, Abies alba, are mixed in, and therefore the study site is classified as a mixed deciduous and coniferous forest biotope type. Trees with a low trunk diameter dominate, but there are individual "achievers", as well as sections of pole wood and pioneer stages, the latter especially consisting of the European hazelnut, Corylus avellana. The study site is not located in any state-protected area.

Study Period
Camera traps were set up from September 2020 to September 2022.

Type and Model
Wild-Vision Full HD 5.0 with Black-LED flash was utilized for the study. The trigger speed was lower than 1 s. The passive infrared sensor (PIR) was designed for high sensitivity.

Modifications
To reduce the over-illumination that occurs when animals are close to the camera trap, half of the LED field was switched off. We changed to 64 MB cards in order to store the large amount of video data.

Recordings
We used 1 min videos in HD resolution of 1280 × 720 pixels. Only those recordings where there was no equipment failure, user error, or false trigger were used for further analysis.

Definition of an Event
All our specifications refer to the recorded 1 min videos. Several individuals of one species are only counted if they were visible on a single video and at the same time.

Bait
Apodemus mice and bank voles were attracted with unpeeled, black-and-white sunflower seeds. Each deployment equated to 0.5 kg per camera trap. The weight of the sunflower seeds presented was based on experimental results within the genus Apodemus. At high densities, several hundred grams can be consumed and transported from a single feeding site, especially during the intensively used first night [65]. In order to address the questions of relative frequency and activity pattern, it had to be possible to feed a maximum number of target species individuals for a sufficient period of time. The sunflower seeds were provided on the ground and not covered. This enabled both a quick triggering of the camera traps and the most reliable identification of small mammals, as well as other species.

Layout
According to a minimum convex polygon analysis of camera trap sites by means of QGIS 3.18.0, the study area had a size of 4.8 ha. A random allocation was carried out [66], with 83 sites being sampled over the course of the investigation.

Spatiotemporal Independence
In order to get a high degree of independence between trials, we set up camera traps every 12.07 ± 6.32 sd days [67], and sample sites from the last trial were not allowed to be repeated. Further, the analyses were standardised according to the first 24 h of video recording per sampling site [68], hereinafter referred to as a camera trap-night. To minimise the movement between two camera traps and the immigration of small mammals within one trial, a distance of 58.94 ± 42.47 sd m was maintained between typically 6 simultaneously active cameras. Distances were determined with QGIS 3.18.0 and corrected for the slope.

Camera Trap Placement
The bait was centred in the middle of the shot and sunflower seeds were spread over a maximum area of 50 × 50 cm. This distance from the focal point to the camera (1-2 m) and the height of the camera (typically from 0.25-0.75 m), and thus the angle, depended on the different slopes in our study area. Camera traps were only placed in locations where minimal vegetation change was required for a clear line of sight to the bait.

Data Treatment
For the analysis of day-night activity pattern, we achieved a high degree of independence of the data by keeping only the single recording with the highest number of Apodemus mice or bank voles from all videos of the same camera trap within a 30 min period [69]. The 48 periods of 30 min periods within a calendar day were assigned to either day or night, separated by sunrise or sunset. The two twilight periods of a day were assigned to the time of day in which more of the 30 min were located.

Predictors
For the multivariate regression analysis of the effects of extrinsic factors, we tested 6 additional independent variables besides mast year/non-mast year: (1) lunar illumination: % illumination in the camera trap-night (https://moonphases.co.uk/ (accessed on 10 December 2022)); (2) minimum temperature: average of the minimum values of the two calendar days of one camera trap-night (https://data.hub.zamg.ac.at/ (accessed on 12 December 2022)); (3) precipitation at night: precipitation total (mm) from 19:01-7:00 CET per camera trap-night (https://data.hub.zamg.ac.at/ (accessed on 12 December 2022)); (4) predators: number of 30 min periods per camera trap-night with at least one visible predator on the available videos. In our dataset, red fox, Vulpes vulpes, stone marten, Martes foina, European polecat, Mustela putorius, and tawny owl, Strix aluco, were classified as mandatory predators of small mammals. Moreover, (5) cover: for our year-round comparison, we recorded the % of cover with lying deadwood, snags and rocks in a 10 m radius of the camera trap, because numerous studies have shown a positive correlation between the abundance of voles of the genus Clethrionomys and the amount of woody debris and brushwood coverage (so-called thigmotaxis [70][71][72]); and (6) Apodemus spp.: Σ of Apodemus mice from the pseudo-replication-limited 30 min periods per camera trap-night.

Statistical Analysis
The data were statistically analysed by the program package R 4.2.2 (https://www. R-project.org/ (accessed on 14 December 2022)). Mann-Withney U tests were used to examine the frequency differences between two groups, and χ 2 tests to examine the relationship between nominal variables. The dichotomous day-night activity of bank voles was investigated by multivariate logistic regression. Activity patterns were analysed with Hermans-Rasson tests, using the R-package "CircMLE" [73,74]. Overlap coefficient (∆) between the temporal activity patterns of Apodemus mice and bank voles was estimated with the R-package "overlap" [75]. This coefficient ranges between 0 (no overlap) and 1 (complete overlap, [76]). Since the number of cases for the comparisons was greater than 75, the ∆ 4 estimator was calculated [76]. In addition, 95% confidence intervals for ∆ 4 using 10,000 bootstraps were determined. The overlap was rated as "medium" when ∆ 4 was between 0.50 and 0.75, as "high" with ∆ 4 > 0.75 and as "very high" if ∆ 4 was larger than 0.90 [77]. Statistical significance was set at p < 0.05, and marginal significance at p < 0.10.

Frequency of Small Forest Rodents
On 215 camera trap-nights, 19,138 analysable videos (319.0 h) were taken within the two years of survey. 102 camera trap-nights were conducted in the first mast year, and 113 in the non-mast year. At least one small forest rodent was detected on 15,931 1 min videos (265.5 h). Using the maximum number of either simultaneously visible bank voles or Apodemus mice per camera trap-night as a measure of frequency, population trends between the mast year and the non-mast year could be compared ( Figure 1). The frequency of bank voles did not differ between the mast year and the non-mast year (U = 5965, p = 0.63, r g = 0.035). Bank voles showed a tendency to be more frequent after the reproductive phase from spring into summer in both survey years, but three, or in a single case four, individuals visible on the same video always represented an exception. In contrast, Apodemus mice were significantly more frequent in the mast year compared to the non-mast year (U = 9297, p < 0.001, r g = 0.612). In the mast year, Apodemus mice overwintered in high numbers and reached their population peak the following summer. Regularly up to five, and exceptionally up to eight, individuals could be counted on the videos at the same time. In the non-mast year, the population collapsed totally and recovered only slowly towards the end of the study, in the second summer. In winter and spring of the non-mast year, Apodemus mice were not detected on many camera trap-nights and were rarer than bank voles during this phase (Figure 1).

Figure 1.
Population development of small forest rodents in an inner-alpine mixed forest. The frequency of Apodemus mice increased significantly in the mast year, whereas this was not the case for bank voles, Clethrionomys glareolus. In the non-mast year, the population of Apodemus mice collapsed.

Temporal Niche of Bank Voles
A multivariate logistic regression was used to investigate the dichotomous activity pattern of Clethrionomys glareolus against seven extrinsic predictors ( Table 1). The probability of nocturnal activity was significantly increased in the non-mast year (p < 0.001, OR = 2.329). Overall precipitation during the night significantly reduced activity (p < 0.001, OR = 0.932). A marginal significant effect was caused by lunar illumination (p = 0.051, OR = 0.997) and the activity of Apodemus mice (p = 0.060, OR = 0.992), both of which led to a tendency towards increased daytime activity.

Temporal Niche of Bank Voles
A multivariate logistic regression was used to investigate the dichotomous activity pattern of Clethrionomys glareolus against seven extrinsic predictors ( Table 1). The probability of nocturnal activity was significantly increased in the non-mast year (p < 0.001, OR = 2.329). Overall precipitation during the night significantly reduced activity (p < 0.001, OR = 0.932). A marginal significant effect was caused by lunar illumination (p = 0.051, OR = 0.997) and the activity of Apodemus mice (p = 0.060, OR = 0.992), both of which led to a tendency towards increased daytime activity. In the mast year, univariate comparisons revealed significantly more diurnal activity of bank voles, with an increased number of Apodemus mice (U = 41,794, p = 0.01, r g = 0.125), whereas this was not the case in the non-mast year, with a much lower absolute number of interspecific competitors (U = 76,515, p = 0.09, r g = 0.072). Nocturnal precipitation induced a significant increase in diurnal activity of bank voles in both the mast year (U = 39,366, p = 0.045, r g = 0.059) and the non-mast year (U = 78,993, p < 0.001, r g = 0.106). Increased lunar illumination resulted in a significant increase in nocturnal activity in the mast year (U = 32,337, p = 0.008, r g = −0.130), but in a significant decrease in nocturnal records of bank voles in the non-mast year (U = 86,762, p < 0.001, r g = 0.215).

Intraguild Aggression
Apodemus mice and bank voles were only seen together on 350 videos out of 15,931 recordings, with at least one of these small forest rodents filmed (2.20%). Encounters were about equally frequent in the non-mast year and in the mast year, although in the latter Apodemus mice (but not bank voles) were much more frequent ( Table 2, comp. Figure 1). In the mast year, Apodemus mice were less aggressive, whereas in the non-mast year bank voles more often avoided Apodemus mice or were actively chased away from the food by them (χ 2 (2) = 7.240, p = 0.027, Cramérs V = 0.144).

Discussion
Rodent communities are considerably shaped by interspecific competition for limited resources and temporal partitioning is one possible dimension of niche separation [16,78,79]. We studied whether bank voles show an inverse activity pattern compared to the superior Apodemus mice. There are important, but contradictory, studies on this question [4,20,21,25,[80][81][82][83][84][85][86][87][88][89], and there is still no consensus on the influence of interference competition on the chronoecology of the bank voles due to the use of different methods (e.g., capture data, cafeteria experiments, and camera-trapping), the incomplete inclusion of confounding variables, the lack of year-round studies and the limited knowledge on the frequency, as well as nature of direct confrontations with Apodemus mice. We tried to approach this issue with a large sample size of videos from camera traps in an inner-alpine mixed forest, which were baited to increase the number of observations [20]. We assume that our smell or handling hardly delayed the use of the bait, because (diurnal) bank voles regularly appeared in the first half hour after setting up the camera trap (19.2 ± 7.66 sd min, minimum 3 min, n = 10).
Assumption 1: An important requirement to test competition among species is met if at least one species is particularly common [9]. We expected the massive increase of small forest rodents in the seed mast year. In agreement with many other studies [10], this first assumption was fulfilled in Apodemus mice. In contrast, bank voles remained at low population levels throughout the study. Such relatively stable populations have been found, for example, in the transition zone between the northern cyclic and the southern outbreak populations in the so-called "limes norrlandicus" in Scandinavia [40], in undergrowth-poor forest stands of the central European lowlands [42] and on low-productivity study areas of the inner Alps [90]. For our research question, this ideally meant a quasi-experimental variation in the frequency of the dominant competitor, with the frequency of the inferior species remaining approximately constant. The small number of bank voles also renders (largely unknown) intraspecific effects unlikely to outweigh interspecific competition [91].
Assumption 2: Albeit Apodemus mice were more diurnal in the mast year, they were almost exclusively nocturnal overall. This is in agreement with numerous findings on activity patterns within the genus Apodemus [4,21], which we could not distinguish at a species level in our study. Bank voles showed more variability, with diurnal activity predominating in the mast year, but nocturnal activity predominating in the non-mast year. Accordingly, the overlap in activity rhythms between the two small rodent taxa was much smaller in the mast year, when Apodemus mice were common. This suggests an avoidance strategy, which, however, remains a vague assumption without the inclusion of other possible explanatory variables. Prediction 1: We predicted an inverse activity pattern of bank voles with an increased presence of Apodemus mice, controlling for a set of other possible explanatory variables. The analyses revealed a close relationship between the frequency of Apodemus mice and the diurnal activity of Clethrionomys glareolus, particularly and significantly in the mast year. This indirect measure strongly suggests a temporal partitioning effect due to intraguild competition, because none of the numerous additional explanatory variables describe the same relationship: overnight precipitation reduced bank vole nocturnal activity, regardless of the frequency of Apodemus mice. This is consistent with findings in other studies [3,82], but contrary evidence exists [2] and causes are not fully understood (e.g., thermoregulatory costs, interspecific concurrence, and predation risk). Lunar illumination led to an overall marginally increased diurnal activity of the bank voles, but interestingly with different effect directions in the mast and the non-mast years. While bank vole night activity increased significantly with lunar illumination in the mast year, it was significantly reduced in the non-mast year. We suggest an association to intraguild competition in this case as well, because Apodemus mice are specifically known for their lunar phobia [3,4,19]. Therefore, bank voles might take advantage of the relatively competition-free period of bright lunar nights, as Apodemus mice reduce their activity even at high densities due to bright illumination. At low Apodemus mice densities, bank voles may use even the darkest hours to avoid visual predators [92,93]. This would explain the contrasting findings on this physical factor in other studies [3,4,19].
For the non-significant predictor cover, we assume too much uniformity between camera trap locations. To the bank voles' disadvantage [9,18,26,42,[94][95][96], our study forest was generally poor in cover below the tree layer, which is probably reflected in the comparatively low frequency of this species, even in the mast year [62,97]. We detected very few predators on our video recordings (n = 7). All were nocturnal, almost exclusively mammalian predators (85.7%) and occurred primarily in the non-mast year (87.5%), where a time-delayed population increase of these generalist, non-migratory, and territorial predator species can be expected as a consequence of dispersal. Therefore, it is even unlikely that indirect cues from predators, such urine, faeces and anal gland secretions [8,93,[98][99][100][101], led to increased daytime activity in bank voles in the mast year. In addition, there was also no discernible correlation between the minimum temperature and the behaviour of the bank voles. In other studies, the effects of temperature were small as well or counteracting [2,3,82].
Finally, it is particularly important to note that the independent variable mast year/nonmast year itself is significantly correlated with the diurnal activity of bank voles. Since a test for multicollinearity showed that predictors of the multivariate logistic regression model were not redundant (each variance inflation factor < 2), there must be an influence on bank vole diurnal activity pattern beyond the mere frequency of Apodemus mice. Having excluded the most common alternative variables, we surmise an additional influence of the other two niche axes: habitat and food. It is well known that the dietary niche between these small forest rodents overlaps much more in mast years, and high densities of Apodemus mice make spatial separation more difficult for the bank voles, even at the microhabitat level [41,102,103]. We therefore assume that under such ecological preconditions, the diurnal activity becomes particularly important for Clethrionomys glareolus. Due to a similar food choice and the "omnipresent" Apodemus mice, raised fitness costs of diurnal activity (e.g., increased predation risk [104,105]) can no longer be compensated for at night. Prediction 2: The two taxa encountered each other very rarely in front of the camera traps. The prediction that a higher number of Apodemus mice would lead to increased intraguild aggression towards bank voles was not confirmed. In the non-mast year, despite increased aggressiveness of Apodemus mice, no inverse activity pattern of bank voles analogous to the mast year was observed. Thus, the greater number of Apodemus mice had a stronger negative impact on bank voles in the mast year than could be compensated by improved food supply. As the number of encounters did not differ between the mast year, with numerous Apodemus mice, and the non-mast year, we assume an active avoidance strategy by the bank voles. Given the long co-evolution of these small forest rodents, it would be unlikely that dominance relationships need to be actively re-established permanently. Other authors, too, could not find evidence for persistent aggressive meetings between Apodemus mice and bank voles [20,25,27]. In the non-mast year, the few adult Apodemus mice (as far as assessable by body size) were particularly aggressive, probably in the struggle for scarce food resources.

Conclusions
Our studies showed that a high frequency of the almost exclusively nocturnal Apodemus mice led to increased diurnal activity of the bank voles, and this could not be explained by an extensive set of alternative predictors. We assume that at high Apodemus mice densities, increased separation along the time axis by bank voles must occur, as at night, there is no longer sufficient relief of intraguild competition in all niche dimensions: time, microhabitat and diet. The "omnipresence" and intimidation by Apodemus mice is satisfactory for the bank voles' inverse activity pattern, without the need for frequent aggressive meetings. For rigorous quantification of the exclusive influence of Apodemus mice on the activity budget of bank voles, studies in areas where the two genera do not live syntopically, and extensive, capture-mark-recapture or removal experiments simultaneous to the camera trap survey, would be purposeful.