The influence of pups on aggressive interactions between smooth‐coated otters and water monitor lizards in Singapore

Abstract Smooth‐coated otters (Lutrogale perspicillata) and Malayan water monitor lizards (Varanus salvator) occupy similar habitats and interact regularly in Singapore's waterways. These interactions have a range of potential outcomes and are sometimes lethal. Few formal behavioral studies exist for either species. We analyzed interactions between otters and monitor lizards by gleaning data from publicly available videos from citizen scientists to examine what factors influence aggressive and defensive behaviors and what influences vigilance in otters. Behavioral sequence analysis revealed no obvious monitor lizard behavior that predicted otter aggression toward monitors. We found that the presence and number of otter pups are positively associated with otter aggression. Otters also tended to be more vigilant in groups with more pups and more vigilant on land than water. Monitor lizards almost always displayed aggressive and defensive behaviors, regardless of whether otters were aggressive toward the lizards. These observations suggest that otters vary their aggression and vigilance levels depending on their group composition and the physical environment of their interactions with monitor lizards.

otter population . Otters in Singapore have been extremely successful in their recolonization of local waterways (Khoo & Lee, 2020), and there are now at least 11 family groups across Singapore (Khoo & Sivasothi, 2018b) as well as lone otters not attached to any family group.
Otters encounter other species, sometimes with negative consequences (Belanger et al., 2011;see Chung, 2017;Ng, 2020). Feral dogs can be aggressive to otters and vice versa (Clements, 2019), and social media includes reports of otters harassing estuarine crocodiles (Crocodylus porosus; Toh, 2018), a behavior similar to that of giant otters (Pteronua brasiliensis), which sometime attack caimans (Ribas et al., 2012). Smooth-coated otters also frequently encounter Malayan water monitor lizards (Varanus salvator). These large lizards (up to 20 kg) inhabit a range of tropical environments from the Molucca Islands to Sri Lanka (Twining & Koch, 2018). Water monitors are opportunistic, and like smooth-coated otters, they prey on fish and other animals, but unlike otters also frequently scavenge (Twining & Koch, 2018). Monitors persisted in Singapore, even as otters were driven out by a deterioration of suitable habitat. Today, they are common in Singapore's waterways, whether concrete canals or natural riverbanks.
Interactions between otters and monitors can take several forms. Water monitors can be commensalists that scavenge fish remains otters leave (personal observations), kleptoparasites that steal fish from otters (e.g., Tan, 2019), competitors for food resources, or predators that attack young otters (e.g., Lee, 2019; personal observations). Sometimes otters and monitors interact aggressively ( Figure 1), which can lead to otters attacking monitors and to the injury or even death for the monitor lizard (e.g., Mitchell, 2021). Ottermonitor conflict has been noted before (Goldthorpe et al., 2010), albeit from a single observation in Peninsular Malaysia.
Few formal studies have been conducted on either smoothcoated otter or monitor lizard behaviors. The demographics of smooth coated otters in Singapore and elsewhere are well studied (e.g., Khoo & Sivasothi, 2018a, 2018b, but there is scant literature on their behavioral traits, especially those that may be unique to urban otters. Urban environments change behaviors in other animals (e.g., Breck et al., 2019;Slabbekoorn & den Boer-Visser, 2006), and urban otters also adjust to their environments. They construct unique holts in man-made structures (Khoo & Sivasothi, 2018b) and use manmade features such as ladders in their movements (Lay, 2021). The robust population of otters in Singapore is notable because in other areas, smooth-coated otters are associated with natural landscapes (Kamjing et al., 2017). Likewise, outside of a few ecological and descriptive studies (e.g., Twining & Koch, 2018;Uyeda, 2009), very little is known about water monitor behaviors (but see DeLisle, 2007), especially in urban environments.
These sometimes-violent encounters between otters and monitor lizards raise questions of their causes. In other species, otter groups can be formidable. In one study, giant South American otters (Pteronura brasiliensis) mobbed jaguars until the jaguars left the area, but a lone otter did not engage the larger cat by itself (Leuchtenberger et al., 2016). Smooth-coated otters live in family groups like those of giant otters, and group dynamics may play a role in their interactions with water monitors. Otter pups are particularly vulnerable; the presence and number of pups may influence otters' responses. Likewise, whether monitor lizards' are aggressive or defensive may depend on both otter behaviors and the otter group composition. Otters are vigilant to potential threats, adopting a characteristic upright posture and other behaviors in and out of water, but the relationship between group size, the number of pups present, and how vigilant otters are to monitors or other threats has not been addressed. Here we examine encounters between otters and monitor lizards to assess factors that lead to aggressive interactions and specifically to otters attacking monitor lizards.
Collecting behavioral data can be extraordinarily time-consuming, especially data related to uncommon events such as otter-monitor interactions. However, the ease with which consumer-grade technology such as smart phones and relatively inexpensive cameras can record photos and videos has led to an explosion of raw data on social media and other corners of the internet. Crowd-sourced data have been used for citizen science studies related to occupancy, ecology, and conservation for some time (see Cooper, 2016 for review). Citizen science can also be effective in animal behavior studies, and some studies glean data from public sources (e.g., Boydston et al., 2018;Bungum et al., 2022;Krueger et al., 2019;Loong et al., 2021). Gleaning data from public repositories, such as social media or YouTube, can be particularly useful when the behavior or species is rare or difficult to find (Nelson & Fijn, 2013). Here we glean data from publicly available sources of video to address otter-monitor interactions. Because social media posts amount to ad lib sampling (Altmann, 1974), there is a potential for bias toward obvious or conspicuous behaviors. We try to avoid that potential by narrowing the scope of our study to otter-monitor interactions.
F I G U R E 1 Two smooth-coated otters approach a monitor lizard in Singapore. Notice that the monitor lizard's neck frill is extended in an aggressive manner. The monitor whipped its tail at the otters shortly after this photo was taken. Photo courtesy of Alicia Ellen Brierley.

| MATERIAL S AND ME THODS
Almost all data in this study came from online sources due in part to the restrictions imposed by COVID-19 pandemic. Selected videos were obtained from YouTube (www.YouTu be.com) using search terms such as "Singapore otter monitor" and "Singapore otter lizard" (Appendix 1). One YouTube channel had a large collection of videos, "RandomSG" (Wong, 2019; www.youtu be.com/chann el/UCLz7 pIXxz aFzz_MN02k NzsQ), and the videos posted on this channel were not edited. Otters in Singapore's waterways are generally active early and late in the day (roughly from 06:30 to 10:00 and from 16:00 to 19:30), and there was a nearly complete record of twicedaily videos spanning 2019-2021, ceasing only during the mandated lockdowns due to COVID-19. The regular recordings were free of obvious bias and allowed us to track pup age over time. We identified 160 videos with potential interactions between otters and monitors; due to quality, duplications, and scoring protocols, we scored 63 for the study, all taken between October 2018 and July 2019, of the Pandan otter family group, which resided near the Ulu Pandan River in southwestern Singapore. (In Singapore, people name otter family groups after the first place they were observed; in 2019, after the events in this study, the Pandan family group was forced out of the river by the Jurong Lake Gardens family group.) The Pandan family group consisted of nine adults and up to nine pups at the time.
Videos provided were taken using a handheld, fixed-lens DSLR camera, generally positioned at the bank opposite the otters' activities, 10-30 m away.
These 63 videos were chosen from a time of 70 days before the first emergence of the pups in May 2019 until 40 days postemergence. Pups remain in the holt for roughly 6 weeks after birth (Khoo & Sivasothi, 2018a). At this age, pups are not yet fully weaned and so are still heavily reliant on group members' help to survive (Hwang & Larivière, 2005). This chronologically complete set of videos allowed us to compare behavior of adults before pups were born, after pups were born but before they emerged from their holt, and after the pups emerged and as they aged.
First, we examined the association between otter and monitor aggression in all 63 videos without distinguishing between potential aggression based on a minimum distance criteria. If monitor lizards and otters came within one otter body length of each other (about 1 m), that encounter had the potential for direct aggressive. Videos with the potential for aggressive encounters were then divided into "bouts," where each bout lasted up to 1 min before and after: (1) an aggressive behavior between otters and monitors; (2) the first instance of otters and monitors coming within one otter body length of each other; or (3) until the subjects left the frame and scoring ceased within the one-minute window. Videos without the potential for aggression could not be separated into bouts because otters and monitors were never close enough to physically interact, although monitors were free to display a defensive posture. (It is difficult to disentangle monitor aggression and defense, and here we group monitor aggressive and defensive behaviors.) We treated bouts within videos as independent events. The bout ended if the lizards and otters turned away and disengaged from each other; if they performed a "benign" rather than aggressive or defensive behavior -typically something like "sit" or "groom" -or if a focal otter interacted with another otter. By these criteria, the average bout length was 53.24 + 35.78 s (median 50.49 s). Otters are extremely active and often engage in several behaviors per minute. In the one incident in which there were back-to-back bouts between the same monitor and otter, the behavior that ended one bout began the next.
By these criteria, we observed 234 bouts within the 63 videos. Each bout included 1-18 otters and f 1-3 lizards. All 234 bouts reached a conclusion before the videos cut off, and there were no recorded serious injuries in any of the scored observations. There were also no edits or other post-production cuts within videos, i.e., the videos accurately reflected interactions. (Many otter watchers in Singapore edit their videos before posting them.) We analyzed behavioral sequences in the 234 bouts to describe typical behavioral exchanges between otters and monitor lizards and to examine possible suites of behaviors that might spur attacks. In our behavioral sequence analysis, we included behaviors from both otters and monitors to address cause and effect between species.
We used four pilot videos, excluded from the final analysis, to generate an ethogram of otter and monitor behaviors (Table 1; Figure 2).
Smooth-coated otters usually lack individual markings, which precluded our tracking individuals from one video to the next. Within each video, we assigned each otter and each monitor lizard an identifier ("subject") based on the animal's initial position at the start of the interaction. If either the otter or lizard exited the frame, we ceased all scoring, i.e., both otter and monitor needed to be in view for any scoring to occur. If both subjects returned to the frame, they retained their subject designation only if we could be certain they were the same individuals. If there were any doubt or lack of clarity, subjects were reassigned, and we analyzed subsequent behaviors as a new bout. We exported behavioral strings from BORIS to the BORIS tool Behatrix (http://www.boris.unito.it/pages/ behatrix), in which we conducted a 10,000 permutation test to limit the behavioral transitions to those that occurred more frequently than chance (p < .05, transition occurring at least 1% of the time).
For each video and bout, we also recorded: presence of pups, number of pups if present, number of adults, number of monitors, age of pups, whether the events happened on land or water, whether there was an aggressive interaction, location of interaction, time of day, relative size of monitor to otter, minimum distance between otter and monitor, and general activity of otters (Table 2). We then constructed generalized linear mixed models (GLMMs), where we examined each bout as unique (n = 234), unless it occurred both on land and in water, for a total of 270 analyzable interactions. We considered general movement between land and water benign behaviors, meaning the bout would end and a new one begins, should this occur. We considered the presence or absence of otter aggression, per otter, per bout. We examined otter vigilance with linear mixed models (LMMs), where we treated the initiation of otter vigilance as a point event, and because bouts varied in length, we treated vigilance as a continuous rate, per otter, per bout-minute. For consistency, we analyzed vigilance using the same subset of potentially aggressive encounters, even though vigilance could potentially occur when otters and monitors were further than one body length apart.
We generated GLMMs and LMMs with functions glmer and lmer, depending on whether the dependent variable was binomial (otter aggression) or continuous (vigilance), using the R package (R Core Team, 2021; https://www.R-proje ct.org/) lme4 (Bates et al., 2020). Fixed and random effects are in Table 2. We visualized flow diagram scripts from BORIS using the R package "DiagrammeR" (Iannone, 2020), and we used the Tidyverse package (Wickham & RStudio, 2019) for data analysis and created figures with the ggplot2 and sjPlot packages. Because this is a descriptive study, we present multiple linear models and then discuss their common features (Burnham et al., 2011).

| RE SULTS
Monitor lizards displayed aggressive behaviors in more videos than otters did. Monitors displayed aggression in 79% (50/63) of videos, but otters were generally not overtly aggressive, displaying aggressive behaviors in only 27% (17/63). We found a significant association between otter and monitor aggression among videos (Fishers exact p = .013). In all videos where otters displayed aggression, monitors also displayed aggression (17/17); but otters were only aggressive in 34% of the videos where monitors were aggressive (17/50).
Monitors displayed aggression in 72% of videos where otters did not (33/46), and otters never displayed aggression when monitors did not (0/13). Including only videos where otters and monitors were within a body length, monitors almost always displayed aggressive behaviors (33/34; 97%) while otters did only half the time (17/34).
To gain more insight into the causes of aggression, we examined the behavioral sequences within bouts. Permutation tests TA B L E 1 Ethogram describing behaviors scored in otter-monitor interactions. The original ethogram is based on Tan (2017) and builds on ethograms for other otter species (e.g., Azevedo et al., 2015;Green et al., 2015;Reed-Smith et al., 2014). Behaviors categorized into: Aggression, locomotion, and general sound/sign.  We examined whether the composition of otter groups within bouts influenced whether otters were aggressive. We found otters in groups with pups were more likely to display aggression toward monitor lizards than otters in groups without pups ( Figure 5; Fisher's Exact test p = .0018). We further examined the effects of group composition on the occurrence of otter aggression by fitting GLMMs to the number of pups, adults, pup age, and the environment (land or water; Table 2). In all models, the number of pups in a group was a significant predictor of otter aggression (Table 3; Figure 6); groups with more pups were more likely to be aggressive.
In the best model (Model 1; AIC = 59.1), the number of adults, the interaction between number of pups and adults, and pup age were also significant predictors of otter aggression ( Figure 6). However, in   Figure 7b; Appendix 2), and therefore, groups with more adults required more pups before becoming aggressive. Model 4 includes the interaction between the number of pups and adults as a significant predictor of otter aggression (Figure 7c). Although Model 4 (AIC 77.2) has less support than Models 1, 2, and 3, the effect of adults is qualitatively similar to that in Model 3: more adults in a group shifted the inflection point for otter aggression (Figure 7c). Pseudo-R 2 values (Nakagawa & Schielzeth, 2013) for these models ranged from 0.031 to 0.064 (Table 3), but notice that low pseudo-R 2 values are typical of logistic models (Hosmer et al., 2013).

| Vigilance
We examined what influenced otter vigilance by fitting the same fixed and random effects (Table 2) to LMMs explaining the rate of vigilance (Table 4). The first three vigilance models performed similarly (AIC = 650.0-653.3; Table 4); and some combination of pups, environment, and pup age were significant predictors in all four models. The number of pups in a group was a significant predictor in three out of the four models ( Figure 8). There was a difference between vigilance rates on land and water ( Figure 9a); adult otters displayed higher vigilance levels on land. Pup age was a significant predictor of otter vigilance in Model 4 (Figures 8 and 9b). All four models have pseudo-R 2 values above .3, indicating reasonable explanatory power, especially for an unmanipulated observational study (Table 4).

| DISCUSS ION
In interactions between otters and monitor lizards, we found monitor lizards were aggressive or defensive far more often than are otters. Monitors adopted defensive postures, consisting primarily of a frilled neck and curled tail, even when otters were too far away to cause any physical harm, and monitors were especially aggressive or defensive when otters were within range of an attack (33/34 observations). The frequency of defensive monitor lizards suggests that otters are not responding to monitors, usually, but that monitors respond to the otters' presence and approach. In general, monitors react to otters as a threat.
Our behavioral sequence analysis supports this conclusion. Many of the monitors' defensive and aggressive behaviors are preceded by adults = 1. 05 * pups F I G U R E 4 Behavioral sequences of otter-monitor interactions in water (a) and on land (b). Ovals represent otter behaviors; rectangles represent monitor behaviors; colored symbols indicate aggression. See Table 1 for ethogram. Numbers describe transitions between behaviors; only transitions that occur more frequently than chance are shown. Letters indicate behaviors that precede aggression or notable behavioral sequences (see text).

F I G U R E 5
Occurrence of bouts with (y) and without (n) otter aggression in otter groups with (Y) and without (N) pups. Aggression n y an otter behavior, notably otters biting or touching monitors' tails ( Figures 3 and 4), and in water, these behaviors are often preceded by diving or floating nearby (Figure 4). Monitors nearly always frill their necks and curl their tails when an otter is near them on land, even when the otter does not directly approach the lizard. Behavioral sequence analysis revealed no single cue from lizards that caused otters to react aggressively; nor was a cue obvious from our personal observations. Some behaviors led to predictable outcomes, e.g., if a monitor whipped its tail at an otter, then the otter retreated, but we found no behaviors that predicted why the otters first attacked the lizard.

| Otter aggression
Overall, we found that otters were only aggressive in encounters where monitors were also aggressive. However, that included situations where the otters instigated an attack, and the monitors responded, i.e., otters are not responding to monitor aggression per se. This observation supports the notion that monitors are reacting TA B L E 3 Generalized linear mixed models 1-5 predicting otter aggression toward water monitors. Odds ratios and 95% confidence intervals (CI), and p-values for each predictor; marginal and conditional R 2 , AIC, and BIC for each model. See Table 2 for description of variables. to the presence of the otters, rather than otters reacting to the behaviors of the lizards. Our sequential analysis did not give a clear indication of a "trigger" for otter aggression, other than a monitor approaching the otter head on, which was a rare occurrence (3.8% of transitions). Although otter aggression toward monitor lizards can be lethal, in general, otters are not especially aggressive, displaying aggression half the time an attack was possible.

Predictors
In all the models we examined, the number of otter pups in a group contributed to otter aggression toward monitor lizards. Otters are less likely to be aggressive in the absence of pups, and otter aggression is more likely as the number of pups increases, especially if there are more than five pups. When more pups present, there is a greater chance that at least one of them will encounter a threat.
Otter pups are notoriously curious and active; they show little hesitancy to approach or even touch a monitor lizard. We generally observed at least one adult otter in the immediate vicinity of pups, but there were times when it seemed as if the adults could not oversee all the pups at once. One pup might wander near a monitor, and an adult would then attack the lizard; this was the case in one of the fatal incidents we observed (below).
One observation that occurred after our study period was es- Adult numbers play a role as well. In two of the five models predicting otter aggression, the number of adults had a significant effect, and in three models the interaction between the number of adults and pups had a significant effect (Table 3). The combined effects of adults and pups were qualitatively similar across models ( Figure 6, Appendix 2): if few adults are present, having only a few pups in a group can lead to otter aggression; with more adults in a group, more pups need to be present to lead to aggression. Small groups of adults can be aggressive. Even lone otters sometimes attack monitors and even larger species, such as estuarine crocodiles (Goldthorpe et al., 2010). One explanation for the relationship between adults and aggression is that larger numbers of adults are a natural deterrent to monitor attacks. Consequently, in groups with more adults, otters may be less prone to react to monitors as threats.

F I G U R E 6
Coefficients for GLMM models 1-5 explaining occurrence of otter aggression. Another possibility is that, with larger numbers of adults, there are fewer wayward pups. Larger groups may have more babysitters, effectively. We observed groups with wide ranges of adult and pup numbers, including several adults with few pups, so the pattern is unlikely to be an artifact of group size, per se.
Group composition itself could play a role, and which adults are present may determine whether otters are aggressive or not. For example, if certain individuals such as the breeding pair initiate attacks on monitor lizards, then additional adults may not increase TA B L E 4 Linear mixed models 1-4 predicting rate of otter vigilance in the presence of water monitors. Coefficient estimates and 95% confidence intervals (CI), and p-values for each predictor; marginal and conditional R 2 , AIC, and BIC for each model. See Table 2 for description of variables. F I G U R E 7 Models 2, 3, and 4 (panels a, b, c, respectively) predicting otter aggression on the number of pups in a group (Xaxis; see Figure 6). In panels a and c, Y-axis represents probability of otter aggression; in panel b Y-axis represents the number of adults in a group, with a line of inflection points, adults = 1.05*pups, to the right of which otters are more likely to be aggressive (see text).  Pup age may also play a role. We found that pup age was positively related to otter aggression in the best model ( Figure 6) but negatively related to otter vigilance (Figure 8). Presumably at some point pups become independent enough that adults do not need to watch over them, but it is hard to reconcile these contrasting patterns in aggression vs vigilance. Pup age, like other factors in this observational study, warrants further investigation.

| Vigilance
In none of our models did the number of adults in a group have a significant effect on vigilance rates. This differs from many other studies of vigilance, where group size often decreases vigilance rates (Beauchamp, 2001(Beauchamp, , 2008Bertram, 1980;Elgar, 1989;Fernández-Juricic, 2012;Quenette, 1990;Roberts, 1996). However, we found that vigilance rates generally increase as the number of pups in a group increase, which supports the hypothesis that adult otters increase their vigilance rates to compensate for unequal pup contribution. The effect was significant in three out of our four models. Pups F I G U R E 8 Coefficient estimates for LMMs, models 1-4, explaining the rate of vigilance, per otter, per minute. are curious and frequently venture near lizards; adults likely need to keep a better lookout. Vigilance is typically interspersed with other stationary riverbank behaviors, such as grooming and playing, but we found it did not by itself lead to aggression (Figures 3 and 4).
Pup age was a significant factor toward increased vigilance levels in one model (model 4, Figures 3 and 4), but was either insignificant or marginal for the other three models. The duration of this study may not have been long enough to measure a change in adult behavior. At some point pups grow up, and presumably adults stop compensating for them then.
Vigilance rates were also higher when the otters were on land than in water. There could be several reasons for increased vigilance rates on land. One is that visibility is less obstructed on land.
Otters often go onto banks to groom, eat, or visit spraint sites, and the group remains more stationary than in water, allowing otters to better "keep watch" over each other. But predators can spot otters, too, and otters on land may be more vulnerable to predators than swimming otters. The increased risk may compel otters to be more vigilant on land. On land, they are likely more vulnerable not just to large monitors, but also to packs of dogs and even crocodiles (Clements, 2019), plus, of course, humans. Swimming otters' agility seems to explain why lizards are reluctant to be in the water when otters are around; there are few animals in Singapore's waterways that otters cannot outswim. The difference between land and water was probably not an artifact of our criteria for vigilance; we defined vigilance such that it could be observed on land and water.
Our study was not aimed at water monitor lizards, per se, but so little research exists on wild monitor behavior that it provided some insights. Otter watchers and other scientists had hunches that monitor lizards sometimes avoid smooth-coated otters, and our analyses bear this out: monitors respond defensively to otters.
We observed only one incident of what was presumably a failed attempt by a monitor lizard to prey on an otter pup. Whether monitors regularly prey on otter pups is something we cannot ascertain from this study. Adult otters may respond aggressively to monitors, not specifically because they are a threat to pups, but because the monitors are large and live in the same areas. Anecdotally, groups of smooth-coated otters engage aggressively with larger species, such as dogs (e.g., Chua, 2018), crocodiles (e.g., Choudhary, 2019), and occasionally even tigers (Narasimhamurthy, 2021), and family groups of giant otters engage with caimans (Ribas et al., 2012) and jaguars (Leuchtenberger et al., 2016). Monitor lizards might simply do their best to avoid groups of angry otters.
Our findings derive from a study of a single family group of otters living on a highly modified river in Singapore, the Ulu Pandan River, where otters and monitors are both common. The extent to which our conclusions are generalizable to other families within Singapore, and to locations outside of Singapore, is not clear. Collecting animal behavior data is very time-and labor-intensive by nature, and here we were able to glean information from videos collected by local otter watchers. By crowdsourcing video recording, citizen science has the potential to be an extremely powerful tool in animal behavior studies due to the "many eyes" effect: if more people are recording animal behaviors, we can collect a more complete record of what animals do. In this manner, this analysis of data gleaned from the internet can be viewed as a form of "next-gen" natural history (Tosa et al., 2021). But this approach has limitations. The current study gleaned information from videos collected in an ad lib manner, albeit with impressive regularity. (The only gap in otherwise daily video recordings of wildlife was about 3 weeks during the COVID-19 lockdown). Biases in ad lib data collection can skew data toward rare, conspicuous behaviors. However, our videos were taken at about the same time and place every day, which should reduce bias toward any particular set of behaviors. Further, we limited which videos we included in the analysis to those with otters and monitors. We make no claims about the overall frequency of otter-monitor interactions (other than that they are surprisingly frequent) and limit our analysis to what happens during otter-monitor interactions. We hope the regular, frequent, and relatively unbiased collection of videos, combined with our filtering the videos to a particular narrow topic, reduced unintentional biases.

| CON CLUS IONS
While we did not discover a specific behavioral "trigger" that otters use as a cue to attack monitor lizards, we did find several factors The presence of young pups increases the chance that otters act aggressively to monitors and increases the rate of vigilance within the otter group. The increased vigilance rates could then lead to otters being more aware of monitors' presence, with a greater chance of aggression resulting. To what extent the growing population of large otter family groups in Singapore's very urban environment contributes to this is something we cannot ascertain from this study.

ACK N OWLED G M ENTS
We would like to thank the otter-watching communities of Singapore Professional Experience (CIPE), and by NParks permits NP/RP20-073 and NP/RP20-075.

CO N FLI C T O F I NTE R E S T
The authors have no competing financial or non-financial interests that are directly or indirectly related to this study.