Temporal Activity Patterns of Sympatric Bornean Carnivore Species: Implications for Niche Partitioning and Conservation

To propose proper conservation measures and to elucidate coexistence mechanisms of sympatric carnivore species, we assessed their temporal activity patterns using 37,379 photos collected for more than three years at three study sites in Borneo. We categorized activity patterns of nine carnivore species (one bear, three civets, two felids, one skunk, one mustelid, one linsang) by calculating the photo-capturing proportions at each period (day, night, twilight). We then evaluated temporal activity overlaps by calculating the overlap coecients. We identied six nocturnal (three civets, one felid, one skunk, one linsang), two diurnal (one felids, one mustelid), and one cathemeral (bear) species. Temporal activity overlaps were high among the nocturnal species. The two felid species possessing morphological and ecological similarities exhibited clear temporal niche segregation, but the three civet species did not. Broad dietary breadth may compensate for the high temporal niche overlaps among the nocturnal species. Despite the high species richness of Bornean carnivores, almost half are threatened with extinction. By comparing individual radio-tracking and our data, we propose that a long-term study of at least three years is necessary to understand animals’ temporal activity patterns by camera-trapping and to avoid diverting conservationists away from effective protection measures.


Introduction
Approximately 20% of the world's mammal species face the risk of extinction, and this rate has become rapid and worse.
Particularly, the status of mammals in the Indomalayan region is the worst among the world's biogeographic realms 1 .
The leading causes of this issue are anthropogenic factors such as hunting, habitat degradation, and invasive species 1 . Regardless of this time-sensitive issue, we still await effective and realistic solutions because of the scarcity of basic ecological information on mammals. Information regarding the temporal activity patterns of animals is crucial for assessing responses to anthropogenic disturbances and will allow the implementation of proper conservation measures 2 . Moreover, understanding the temporal activity patterns of animals may contribute to elucidate their coexistence mechanisms, which is one of the major themes in ecology 3 .
From the ecological perspective, investigating the temporal activity patterns of sympatric closely related species is critically important to understand their coexistence mechanisms in relation to interspeci c competition and/or niche separation, especially among species in the same guilds. Closely related species usually have similar morphology, physiology, behavior, and ecology, and therefore competition among them, primarily inhabiting in the same area is intense 4 . In most cases of those species, to avoid interspeci c competition among these species, one or more differences in temporal and spatial activity patterns, and/or diet is present 4 . Mammalian carnivores are a typical taxon that adapts to a speci c diet, and their internal and external morphologies are suitable for carnivorous diets 5 . Therefore, competitive interactions with food resources may occur among sympatric carnivore species. To reduce the negative effects of competition, such as interspeci c killing and to increase effective access to food resources. They exibly change the temporal activity patterns for their temporal niche partitioning between 2 ≤ species with similar body size and/or utilizing similar-sized prey 6,7,8 . In addition, considering that their temporal activity patterns change in response to environmental conditions, such as disturbance level 9 , temporal activity patterns should be evaluated at the site level.
Asian rainforests possess a far larger number of sympatric carnivores than to other tropical regions, such as Neotropics and Afrotropics 10 . Among the Asian rainforests, Carnivore species diversity is reported to be high in Borneo 11 . The taxonomy of some carnivore taxa is still controversial, but for now Borneo has one bear, ve felids, four mustelids, four otters, one linsang, at least eight civets, and two mongoose species, and at least three of them are endemic 12 . Despite these species-rich communities, almost half of the Bornean carnivore species are threatened with extinction 13 . Some of them are among the top predators, for example, Sunda clouded leopards, in each community 14 . Others are important seed dispersers, such as civets 15,16 ; therefore, they have high value in being protected due to being ecologically important key species that maintain their living ecosystem. Nonetheless, the current information is too limited, and sporadic to understand their basic behaviors, such as temporal activity patterns, which may affect the progress in evaluating and improving the threatened status. This could be because many species of terrestrial carnivores are elusive and di cult to detect in general because of their naturally low density 17 .
Camera-trapping is the most effective method for studying cryptic animals, such as carnivores 17 . Consequently, many studies on the temporal activity patterns of Bornean carnivores have been conducted. However, these are mainly based on small sample sizes, collected in one site during the limited periods, and focusing only on one or a few species. A notable exception is a study 8 which reported the spatio-temporal interactions among Bornean felids by camera-trapping for seven years in 10 sites. However, other taxa such as civets and mustelids are still scarce. To evaluate the threatened status and species interactions, assessing the temporal activity patterns of multiple species in multiple study sites is inevitable.
In this study, we assessed daily activity patterns of sympatric carnivores using our comprehensive photo dataset, including not only felids but also other carnivore species collected for more than three years at three study sites in Sabah, Malaysian Borneo. The objective of this study was to investigate the differences in temporal activity patterns among the study sites and species.

Recorded species and preparation for analysis
We recorded 753, 218, and 290 photos capturing carnivores, totaling 1,261 in the DVCA, LKWS, and TWR, respectively. We recorded one bear, i.  (Table 1). We excluded records of the bay cat, Sunda clouded leopards, atheaded cats, binturongs, and Malay weasels due to their small sample size (< 10). We also omitted the data of the mongooses and otters from analyses because of the di culty in identifying these taxa at the species level based on photos that captured only a part of their body. However, we used mongoose spp. data to t a circular kernel density. The sample sizes of the three civets (banded civets, common palm civets, Malay civets) and sun bears were more than 10 in each study site, and we compared their activity levels among the study sites. We found no signi cant differences in the activity levels of these four species among the study sites (all p > 0.016) (Fig. 1); therefore, we pooled photos taken from the three study sites for all the independent carnivore species records.

Determination Of Temporal Activity Patterns
We applied GLMMs to determine the highest activity period for the carnivore species that were photographed more than 50 times: Malay civets, banded civets, common palm civets, and sun bears. Sun bears had no signi cant differences in the recorded periods (Wald χ 2 = 3.71, p = 0.16). Banded civets were not recorded during daytime, and were recorded signi cantly more at night than during twilight (Wald χ 2 = 6.47, p = 0.01) ( Table 2). There were signi cant differences in the recorded periods in Malay civets (Wald χ 2 = 18.0, p < 0.01) and common palm civets (Wald χ 2 = 26.2, p < 0.01). Both of them were more active at night than twilight (z = -3.46, p < 0.01 in Malay civets; z = -4.93, p < 0.01 in common palm civets), and there were no differences in the recorded photo numbers between daytime and twilight (z = 1.43, p = 0.31, in Malay civets; z = 0.04, p = 0.99 in common palm civets) ( Table 2). Malay civets were more active at nigh than during the day (z = 3.22, p < 0.01), but the signi cance was marginal in common palm civets (z = 2.19, p = 0.06), probably due to the small sample size of their daytime activity (n = 3) ( Table 2). Table 2 The proportion of independent photo-capture of carnivore species during nighttime, twilight, and daytime. Data of species in parenthesis indicate the number of independent photo-capture. All the data are derived from pooled photocaptures of the three study sites. Values in bold letters indicate the signi cantly most active period of each species.
From these data, we determined that Malay civets, banded civets, common palm civets, Sunda stink badgers, leopard cats, and banded linsangs were strongly nocturnal (Fig. 2). Marbled cats were strongly diurnal, and yellow-throated martens were diurnal and crepuscular (Fig. 2). Note that these species were also active during twilight, so the activity pattern here indicates their tendency. The sun bears were cathemeral (Fig. 2).

Discussion
To the best of our knowledge, this is the rst study of the temporal activity pattern of multiple carnivores other than felids in several study sites in Borneo. We nd that six species (three civets, one skunk, one felid, one linsang) are nocturnal, two species (one felid, one mustelid) and mongoose spp. are diurnal, and one species (bear) is cathemeral. We successfully obtain substantial sample sizes of the three civet species (banded civets, common palm civets, Malay civets) and sun bears, and we con rm that their activity patterns do not differ among the three sites. Therefore, this study would be a thorough reference for the primary activity patterns of these four species. The results of the activity patterns for the other ve species (leopard cats, marbled cats, Sunda stink badgers, yellow-throated martens, and banded linsangs) should be interpreted with caution because we could not distinguish individuals and pooled data from the three study sites, which may have introduced some pseudo-replications 18 . However, given the limited amount of data available on some of these species in general, our data would still contribute to understanding their activity patterns.
Temporal niche partitioning among some species with morphological and/or ecological similarities is observed in this study. First, we nd a clear separation of activity patterns between two felid species; leopard cats are strongly nocturnal, while similar-sized marbled cats present diurnal behavior (Table 3, Fig. 3). These results corroborate the previous study in Sabah 8 . Second, yellow-throated martens and common palm civets on Borneo also have several similarities such as body sizes, diets, and semi-arboreal habits 19 , suggesting that they could be potential competitors, although they belong to different families. However, yellow-throated martens are diurnal and common palm civets are nocturnal, therefore, their temporal activity overlap was low (Table 3), indicating their temporal niche segregation, mitigating negative interactions by avoidance of direct encounters. Contrary to the felids and martens, three civet species of the same family exhibit the most extensive activity overlaps among the observed species (Table 3, Fig. 3). The three civet species have similar diet and body size 19 , and they occur in quite similar spatial and temporal spaces. Although there is no evidence of temporal niche partitioning among the three civet species, there appear to be minor differences in spatial activity patterns among them. Banded civets and common palm civets prefer interior forests, open-canopy habitats such as roadside, respectively, while Malay civets are found in both forest types 20 . These subtle ecological differences would be signi cant to maintain their coexistence in complex forest structures in Borneo.
Based on our results, yellow-throated martens and mongoose spp. are strictly diurnal, but the other species have nocturnal activity patterns in varying degrees ( Table 2, Fig. 2). All three civets, leopard cats, Sunda stink badgers, and banded linsangs are nocturnal, and most of them exhibit high overlaps (0.7 < ∆, Table 3) in their temporal activity patterns, except for banded linsangs. Overall, activity overlaps between banded linsangs and the other nocturnal carnivores are not high (∆ 1 < 0.7, Table 3) compared to the others. During the nighttime, differences in activity peaks may relate to the low activity overlaps of banded linsangs. Banded linsangs show clear bimodal peaks during the night and twilight periods (Fig. 2), and they are most active in the last half of the night (Table 2). Whereas, the other ve nocturnal species are active throughout the night, especially in the rst half of the night (Table 2). Thus, even among species with the same temporal activity patterns, some species differentiate activity peaks. However, temporal niche overlap among the other ve nocturnal species is still quite extensive. A possible reason for their coexistence is dietary niche partitioning.
In Borneo, only felids and linsangs are supposed to be hyper-carnivores 19,21 , while the other species are highly omnivorous: feeding on mammals, birds, invertebrates, plant matters 19 . Although information regarding the diets of most Bornean carnivore species is still scarce, such broad dietary breadth may compensate for the high temporal niche overlaps among the nocturnal carnivores.
In this study, we nd that six out of the 10 studied species, including mongoose spp., are nocturnal, and temporal activity overlaps are high among these species. Some species possessing either one or both morphological and ecological similarities exhibit clear temporal niche segregation, but some species, that is, civets, do not. Most of the studied carnivore species are small to medium (< 10 kg 19 ) except for the sun bears. In a guild of ve African sympatric smallmedium carnivores (< 10 kg), they are separated into two temporal groups: three nocturnal and two diurnal species 22 . In Madagascar carnivores, comprising a single-family Eupleridae have three nocturnal, one diurnal, and one cathemeral species 23 . In a Neotropical small-medium felid guild, there are two are nocturnal, one diurnal, and one cathemeral species 6 . Thus, it is suggested that the number of nocturnal small-medium carnivore species is extensive across the continent most likely, due to phylogenetic constraints 24 , but that of the Bornean community overwhelms the others. We nd no cathemeral small-medium species from the studied species. Due to data de ciency, it remains unclear whether the occurrence of these sympatric carnivore species during the same periods generates negative effects such as intraguild killing and interference competitions. Given that temporal niche segregation is one of the most effective mechanisms that diminishes competition 4 , the studied carnivore species may not compete intensively, or have relatively small ecological differences that have not yet been investigated.
Currently, camera-trapping is one of the most basic but effective tools for community ecology and conservation planning in mammals 3 . The temporal activity pattern is one of the main data obtained from camera trapping. Indeed, our data on temporal activity patterns of common palm civets and Malay civets successfully show results similar to radio-tracking in DVCA and TWR 15,25 , where both are predominantly nocturnal, but also show crepuscular behavior. For sun bears, the results are contradictory in an intensive study using both individual radio-tracking and camera trapping conducted in the DVFC for two years 26 : diurnal by radio-tracking, crepuscular and nocturnal by camera-trapping. However, our relatively robust dataset showed that sun bears are cathemeral. Considering that their activity patterns vary at the individual level 26 , the overall activity patterns of several individuals in a certain area may become cathemeral as indicated by this study. The lack of long-term empirical data in any taxa would hinder our understanding of its temporal activity pattern, which could consequently divert conservationists from effective protection measures. Therefore, we propose that at least three years of long-term study is necessary to understand an animals' temporal activity patterns by camera trapping.
All the three study sites are protected areas, but evidence of poaching have been reported, including sun bears in some of these areas 27 . Our results do not show statistical differences in temporal activity patterns of sun bears and the three civet species among the study sites, but this may change depending on the threat status given that some animals change temporal activity patterns because of hunting disturbances 28 . Non-lethal tourism activities may also affect animal activity. Tourism activity was conducted at all study sites during the study period. The potential bene ts gained from ecotourism may frequently counteract the risks of exposure to changes in animal activity patterns 29 . In LKWS, community-based ecotourism is common and can bring signi cant bene ts such as alternative income that incentivizes local communities and policy makers to protect the species in areas of interest. Spotlighting activities along the Menanggul River were often conducted by several motor boats during early morning, late afternoon, and night in LKWS.
However, no nocturnal tourism activities were conducted around the camera stations in DVCA and TWR. Common palm civets show at least two clear peaks of temporal activity levels in the latter two sites, whereas those in LKWS are unclear and delayed (Fig. 1). Given that common palm civets prefer open-canopy areas including riverine forests 30 , they might be directly affected by tourism activity, especially during nighttime in LKWS. Thus, there may be a need for evaluating the effect of tourism activity on animal behavior in future studies, even though it is non-lethal ecotourism.
Lastly, many studies are using camera-trapping data, including remote areas with relatively poor accessibility. Thus, it is the time to accumulate the information on rare species to determine their basic ecology, including temporal activity patterns and habitat selection, and to reassess the propriety of current conservation management strategies.

Study sites
We conducted this study in three protected areas in Sabah, Malaysian Borneo: Danum Valley Conservation Area (DVCA), the Lower Kinabatangan Wildlife Sanctuary (LKWS), and Tabin Wildlife Reserve (TWR). The minimum and maximum daily temperature and annual precipitation among the three study sites did not differ signi cantly (annual temperature:

Temporal Activity Analysis
We de ned non-independent photo capture events as consecutive photos of the same or different individuals of the same species taken within a 30-minute interval and removed these photos from the analysis. We plotted the activity patterns of each species using a von Mises kernel 36,37 using the package activity 38 in R version 4.0.2 39 . We estimated the activity level of animals with more than ten independent photo-capture events as indicated in the previous studies 23 . For our analysis, we pooled the images from all study sites if the photo number of a species was less than 10 in any study locations. If that was not the case, we used the package activity to compare species activity levels across the three research sites using a Wald test with Bonferroni correction for multiple pairwise comparisons 38 . When there were no signi cant differences among the sites, we pooled the photo numbers to estimate activity levels.
We divided a day into three periods: nighttime (19:00-04:59h local time (GMT + 8)); daytime (07:00-16:59h); and twilight (05:00-06:59h and 17:00-18:59h). During the study period, twilight hours essentially corresponded to 1 hour between sunset and sunrise, at 5:54 − 6:25 and 17:50 − 18:25 in DVCA, 5:51 − 6:23 and 17:47 − 18:25 in LKWS, and 5:50 − 6:21 and 17:46 − 18:22 in TWR (data from https://www.timeanddate.com). After converting the time data of each photo-capture event into radians, we tted a circular kernel density distribution estimated by 10,000 bootstrap resampling to radian time data, and we estimated the percentage of active time in each period. We then categorized the activity patterns of photocaptured carnivore species into four categories: nocturnal (active at night); crepuscular (active during twilight periods); diurnal (active during daytime); and cathemeral (active in all periods). We de ned the activity pattern of the species as showing a statistically higher proportion of photo-captures at nighttime, daytime, and twilight periods than at other periods, such as nocturnal, diurnal, and crepuscular, respectively. When photo-capture proportions showed no differences among the three periods, we de ned the activity pattern as cathemeral. For species with substantial sample size (50 <), we compared the number of independent photo-capture event among the three periods by species using generalized liner mixed models (GLMMs) to determine the activity patterns of each species using the lme4 40 and multcomp 41 packages. We set the period as a xed effect, the study site and the camera position as random effects, and the camera working days as an offset term. For other species, we tested if the animals were selectively active; in other words, they were photocaptured disproportionally during any of the three periods using the package adehabitatHS in R 42 . It was impossible to identify individuals from the photo data when only part of the body was recorded; therefore, we used a design resource selection function, selecting at the population level 43 .
After these procedures, we evaluated temporal activity overlaps among the species by the coe cients of overlap (Δ) for each species, ranging from 0 (no overlap) to 1 (complete overlap) using the overlap package in R 44 . We used Δ 1 to estimate the nonparametric overlap coe cient of species with < 75 sample sizes, while we used the Δ 4 estimator for other species with > 75 photos 36 . Next, we categorized the temporal activity overlap level as: low, moderate, and high, based on the values of coe cients of overlap (Δ) generated by the pairwise comparisons. Low, moderate, and high overlaps indicated that Δ values were ≤ 50, 50 < ∆ ≤ 75, and ∆ > 75, respectively 7 . Finally, we calculated the 95% con dence intervals of the overlap coe cient using a smoothed bootstrap with 10,000 resamples.

Data availability
The data used in this study was included in Supplementary Information.  Temporal activity patterns of the ten carnivore species estimated by kernel density estimates. Dotted vertical lines indicate approximate times of sunset and sunrise, and short vertical lines under the kernel density curves indicate photocaptured times of each species.

Figure 3
Temporal activity overlaps among the carnivore species belonging to the same family. Solid lines indicate species before "vs", and dotted lines indicate those after "vs". Grey shared areas indicate the coe cient of overlaps (Δ1 and Δ4) of the two density estimates. Dotted vertical lines indicate approximate times of sunset and sunrise, and short vertical lines under the kernel density curves indicate photo-captured times of each species

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Supplementalydata.xlsx