Human disturbances and the daytime activity of sympatric otters along equatorial Amazonian rivers

Study area

This study was conducted in the Araguari river basin, eastern equatorial Amazonia, in the central region of the Brazilian State of Amapá (Fig. 1). The basin includes rivers classified as “clear water” (Junk et al., 2011). The Araguari River has an extension of 560 km (Ziesler & Ardizzone, 1979) and rises in the Guiana Shield at the base of the Tumucumaque uplands, bifurcates and discharges into both the Amazon River and the Atlantic Ocean. The region’s climate is humid tropical (“Am” Tropical monsoon; Kottek et al., 2006), with a dry season from September to November (<150 mm of monthly rain) and the rainy season (>300 mm of monthly rain) from February to April.

The forests surrounding the study rivers are part of the eastern Amazon Guianan forests (Ter Steege et al., 2001) and consist predominantly of never-flooded closed canopy tropical rainforest vegetation dominated by members of the Fabaceae, Sapotaceae, Lecythidaceae and Lauraceae (Batista et al., 2015). Although deforestation has recently increased in Brazil, Amapá is the state with the lowest deforestation rate within the 5 Mkm² Brazilian Legal Amazon (INPE, 2021; data available from: http://terrabrasilis.dpi.inpe.br/app/dashboard/deforestation/biomes/legal_amazon/increments, accessed 22 November 2021). In our study area there has been some localized deforestation extending approximately 50 km upstream from the nearest town (Porto Grande), yet the majority of survey rivers are immediately bordered by continuous forest cover, i.e., < 1% forest loss from 2000–2020 within 10 km of the rivers (Hansen et al., 2013; https://glad.earthengine.app/view/global-forest-change#dl=1;old=off;bl=off;lon=-51.59987769825302;lat=1.0675835988661395;zoom=10, accessed 7 July 2022) with a border of canopy trees typically starting 1–4 m from the river’s edge and the riverbank rises abruptly along the studied rivers such that forests close to the margin (e.g., within 110–554 m) are never flooded (Caron et al., 2021).

Ethics statement

This study used data from non-invasive field observations and did not involve direct contact or interactions with animals. Fieldwork was conducted under research permit numbers IBAMA/SISBIO 26653, 49632, and 69342 to DN and FM, issued by the Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio).

Data collection

Field data was collected over 53 months during nine years (2011-2013 and 2015-2020) along 218 km of rivers. Standardized river based boat surveys were used (Groenendijk et al., 2005) with direct observations recording the geographic location and timing of otter activity in the rivers. The principal objective was to monitor temporal and spatial patterns, not evaluate behavior. Therefore, observations were limited to a 15 min maximum that prevented observation of additional details such as feeding/social behavior that require extended acclimatization of the study species. Diurnal boat surveys at a standard speed (mean ± SD = 9.6 ± 2.8 km/h, min-max = 1.4–15.9 km/h), by a minimum team of two people, were used to search for direct sightings of giant and neotropical otters along the river stretches (Groenendijk et al., 2005; Oliveira, Norris & Michalski, 2015). To minimize any possible observer related bias between years, a local resident with over 20 years of knowledge on otters, was present in our searches throughout the entire study period. Locations and number of boats and fishing nets during the study period were also obtained during our boat surveys with the aid of a handheld GPS.

Data analysis

We adopted complementary analytic approaches to understand how season and human disturbances influenced otters diurnal activity i.e., “if” and “when” otters were active along rivers. Firstly, generalized additive models (GAMs) were used to evaluate the relative importance of human influences, season and time of day on the presence of both otter species. Activity patterns were subsequentially compared between species using (i) kernel density estimates (Ridout & Linkie, 2009; Rowcliffe et al., 2014) and (ii) comparison of relative abundances using non-parametric tests with P-values calculated using Bonferroni corrections for multiple comparisons. A copy of the data used in the analysis is available in Supplementary Material S1.

We did not transform hours as surveys were only conducted during daylight hours and in contrast to known biases at high latitudes (Vazquez et al., 2019) there is no daylight saving time and minimal variation in day length at our low latitude study area, where the sun rises at approximately 06:30 am year round (day length is 11.95 h in December and 12.05 h in June at the central latitude of 0.93N). The focus on daytime (diurnal) activity also precluded the need to adopt techniques developed for comparisons across the circular 24 h diel cycle (Ridout & Linkie, 2009).

Otter sightings and human river use (boats and fishing nets) were summarized along rivers subdivided into 41 river reaches (continuous sub-segments) of approximately 5 km (mean = 5.3 km, SD = 0.05). Such an approach enabled us to evaluate species activity in relation to variables such as boats and nets that are spatially heterogeneous across the rivers. This approach of dividing river lengths into shorter reaches (sub-sections sensu Smith et al. (2020)) has been applied previously for giant otters (Oliveira, Norris & Michalski, 2015) and neotropical otters (Smith et al., 2020). The decision as to the length of reaches i.e., separation distances between observations is important to avoid issues of spatial autocorrelation. A previous study with giant otters in the same region (Oliveira, Norris & Michalski, 2015) obtained an optimal bandwidth (i.e., separation distance; Berman & Diggle, 1989; Hengl, 2009) of any value greater than 456 m. The final 5 km value was chosen based on a combination of practical and statistical considerations. A length of 5 km corresponds (approximately) to the minimum seasonal giant otter range size of 4.6 km reported along Amazon rivers (Evangelista & Rosas, 2011) and is therefore a relevant and informative scale for our understanding of the species responses. Although there are no comparable studies documenting movements or ranging of neotropical otters across Amazonia (Camp, 2021), a recent study also suggests a length of 5–10 km for spatially independent registers of this species (Smith et al., 2020). Additionally, a short-term (35 day) radio-telemetry study from southeastern Brazil showed a single male neotropical otter moved 1–2.6 km per day around coastal islands (Nakano-Oliveira et al., 2004). The number of independent replicate river reaches (n = 41) provided sufficient information to evaluate variation in the species activity and human disturbances. This number of river reaches enabled us to capture spatial variability yet was not at such a fine scale as to generate zero inflation problems (i.e., finer scales generate severely “inflated” proportion of reaches with zero otter encounters and zero records of houses/boats etc.).

Records from all years were summarized for each hour and river level per 5 km river reach to address each of the three questions. This was necessary as there were insufficient observations to model finer scale variation among years within reaches. To compare activity among seasons, otter sightings were summarized among river levels grouped into four classes: low (September to November), rising (December, January, February), high (March to May) and decreasing (June to August). Associations with otter activity and human influences were compared by grouping river reaches by the relative intensity of human river-based activity. Reaches were classified by boat frequency (number of boats per survey km). Those with none or relatively few boats (less than 0.006 boats per survey km the 25% quantile of density values), intermediate (25–75% quantile range of density values) and many (greater than 0.08 boats per survey km the 75% quantile of boat frequency values). Diurnal activity patterns were also compared between reaches with (n = 26) and without (n = 15) fishing nets during the nine years. Although it was not logistically possible to obtain equal survey effort across all the different factor levels, our extensive survey efforts provided a representative sample across the 41 reaches (Supplemental Material S2). While there was variation in effort among different reaches (mean ± SD days per reach = 65.9 ± 31.6), overall, there was greater survey effort (km per river reach; Fig. S2.2) in areas with more anthropogenic impacts. Therefore, any reduction in otter activity cannot be explained by differences in effort.

Generalized additive models (GAMs) were used to evaluate the relative importance of human influences, season and time of day between otter species. GAMs are an extension of generalized linear models but have important and fundamental differences in their estimation as they relax parametric assumptions (Wood, 2017). GAMs are a powerful and flexible modeling technique that provides a systematic description of the patterns in the data rather than focusing solely on the statistical significance of the differences between the response and explanatory variables (Pedersen et al., 2019; Van Rij et al., 2019; Wood, 2017). GAMs were chosen to develop models explaining patterns in the data as the responses of otter encounters (presence/absence) could be modelled against covariates using a combination of parametric, non-parametric (smoothed non-linear) and random terms (Pedersen et al., 2019; Wood, 2017). As such GAMs also enabled us to control spatial and temporal variation and autocorrelation and provided a more robust and comprehensive evaluation of spatial and temporal patterns than the subsequent kernel density and non-parametric tests in isolation. The modelling approach adopted followed guidance and recommendations presented by Pedersen et al. (2019), Van Rij et al. (2019) and Wood (2017).

GAMs modelled how the probability of encountering the otter species varied with season (river level), time of day (hour), boat frequency, and the presence of fishing nets (Table 1). Summaries for each hour and river level per 5 km river reach provided a total of 1296 presence-absence observations for the GAMs. To control for differences in survey effort GAMs included total km as an observation level offset term. Species were modelled separately, and the presence of giant otters was also included in the model of neotropical otter encounters. Binomial response (presence/absence) was modelled with the quasibinomial distribution family to account for overdispersion (i.e., a lack of independence so that extra-binomial variation occurs) and zero-inflation in the data; with models estimated using restricted maximum likelihood (REML; Anderson, Burnham & White, 1994; Pedersen et al., 2019). Further details of GAM model specifications are presented in Supplemental Material S3.

Information theoretic model selection (Burnham & Anderson, 2002) was used to establish the relative importance of the different variables. Models are ranked and scaled by an information criterion to allow an understanding of model uncertainty over the set of candidate models (Burnham & Anderson, 2002: pp. 281–284). We evaluated models based on their information content, as measured by QAIC Quasi-Akaike Information Criterion implemented in R package ‘MuMIn’ (Barton, 2022; Bolker, 2022). QAIC was adopted as global models for both species were overdispersed, i.e., the global model residual deviance was greater than residual degrees of freedom (Burnham & Anderson, 2002: pp. 66–70; Crawley, 2007: pp. 528).

A subset of models excluding highly correlated variables was used for information theoretic model selection. Models with both fishing nets and boats were excluded for giant otters. Models with a combination of fishing nets, boats and/or giant otter were excluded in the case of neotropical otters. With a strong a priori justification for inclusion, we retained all subset models and all variables; therefore, all predictors were on equal footing to calculate their relative importance as measured by the variables’ Akaike weights (Burnham & Anderson, 2002: pp. 75–77, 167–172), which is a scaled measure of the likelihood ratio that ranges between 0 (least important) and 1 (most important). This approach also enabled us to compare different model specifications including hierarchical specifications representing different assumptions about the degree of intergroup variability in otter species functional response (Pedersen et al., 2019). Variable importance was obtained from a model subset including models having a ΔQAIC-value ≤ 6 (Richards, 2008). Although ΔAIC values ≤2 are often adopted to select models, simulations show ΔAIC values ≤6 are more reliable when the true effect of a measured factor is relatively weak, data are highly overdispersed and/or data are few (Richards, 2008). These issues also limit the ability to reliably estimate effect size and direction particularly in the case of observational data such as ours. Considering the small sample sizes (75 giant otter and 65 neotropical otter detections) and large proportion of zeros, we limited our GAM analysis to comparing the relative importance of variables between otter species.

Activity patterns were then compared between species using (i) kernel density estimates (Ridout & Linkie, 2009; Rowcliffe et al., 2014) (ii) the distribution of relative abundances. Kernel density estimates were obtained using calculations bounded by daytime hours and weighted by the overall effort (km) at each hour for each river reach to account for differences in detectability (Rowcliffe et al., 2014). Overlap in activity was tested via a randomisation test for the probability that two sets observations come from the same distribution and activity levels compared using a Wald test for the statistical difference between activitiy level estimates (Ridout & Linkie, 2009; Rowcliffe et al., 2014).

The distribution of relative abundances was compared using sightings classified by time of day grouped in 3-hour intervals (early morning, 06:00-08:59; late morning, 09:00-11:59; early afternoon, 12:00-14:59; and late afternoon, 15:00-17:59). To avoid spatial correlations (Supplemental Material S2) in the comparison of relative abundances boat frequency and net presence were combined into a single “human-use” factor with four levels: few, intermediate no nets, intermediate with nets and many. Finally, to examine if patterns in neotropical otter activity changed due to the presence of giant otters, the activity of neotropical otters was also compared among the 25 reaches with neotropical otters and reaches with (n = 14) and without (n = 11) giant otter detections. To test for differences in mean values (Kruskal-Wallis rank sum test) and distribution (Two-sample Kolmogorov–Smirnov test) among groups we used P-values calculated using Bonferroni corrections.

Seasonal patterns in diurnal activity

Both otter species were detected during all daytime hours and all seasons (Figs. 3 and 4). Neotropical otters tended to be more active earlier in the day compared with giant otters (Fig. 3), but there was only weak statistical difference in activity levels (Wald test P = 0.010) and strong overlap in temporal activity (86.2%, P = 0.214, Fig. 3). Additionally, there was no significant difference in the overall distribution of daytime activity between species (Fig. 3A, Two-sample Kolmogorov–Smirnov test, P = 0.771). Although there was some seasonal variation in activity, with the relative abundance of both species lowest during high water levels (Fig. 4) there was no statistical difference in the overall relative abundances grouped into 3 h bins between the four seasons for either species (Fig. 4, Kruskal-Wallis rank sum test, P = 0.337 and 0.438 neotropical and giant otter, respectively). Seasonal differences in the distribution of relative abundances grouped into 3 h bins were found for giant otters between low and high river levels (Two-sample Kolmogorov–Smirnov test, P = 0.029) and low and rising levels for neotropical otters (Two-sample Kolmogorov–Smirnov test, P = 0.029).

There appeared to be seasonal patterns to the spatial and temporal distances between species when they were seen on the same day. Both species were observed on the same day on 12 occasions. with species closer in space and time during the low river level months (Fig. 5). These same day observations occurred most frequently during low river levels, with both species seen on 5 days in four years (2011, 2012, 2016, 2019). Species were not usually close when seen on the same day (median distance between species 12.5 km, range 0.8 to 30.9 km), with the closest spatial difference of 808 m having a temporal separation of 8.9 h (28 April 2013, high water, 6:58 three adult giant otters and 15:49 single neotropical otter). Species were also separated temporally when seen on the same day (median time between species 3.0 h, range 14 min to 8.9 h). The temporally nearest same day registers had a difference of 14 min and 1.1 km (November 2011, low water, 9:36 single neotropical otter and 9:50 two adult giant otters).

Diurnal activity in relation to human disturbances

The proportion of reaches with giant otters decreased threefold from 67% of the least disturbed (none/few boats no nets) to 18% of the most disturbed reaches with many boats and fishing nets. In contrast neotropical otter presence nearly doubled from 44% of the least disturbed reaches to 73% of the most disturbed reaches with fewer giant otter detections. Comparison of relative abundances also showed contrasting differences in activity between species. Giant otters were more active in reaches with less human activity, with a greater relative abundance at reaches with none or little human river use (Fig. 6, Kruskal-Wallis rank sum test P = 0.011). There was less variation in neotropical otter relative abundances and no significant differences among river reaches with different levels of human river use (Kruskal-Wallis rank sum test P = 0.580). As there were so few giant otter detections at the reaches with more boats it was not possible to compare the distribution of temporal activity among all river use classes. Giant otters tended to be more active later in the day at the river reaches with intermediate boat use and no nets, but there was no statistically significant difference in the distribution of activity between the few and intermediate use classes (Two-sample Kolmogorov–Smirnov test, P > 0.940 for all pair-wise comparisons).

There was also no statistically significant difference in relative abundance of neotropical otters in the stretches with different boat use (Fig. 6). Neotropical otters also showed no statistical difference in the timing of activity (Fig. 6, two-sample Kolmogorov–Smirnov test, P > 0.946 for all pairwise comparisons).

Neotropical otter activity

There was some variation between temporal activity of neotropical otters in river reaches with and without giant otters (Fig. 7). Yet overall temporal variation was generally similar between reaches with and without giant otters (Fig. 7). Neotropical otters tended to be more active in the late afternoon (15:00–17:59) in reaches with giant otters, with relative abundances double the median value (Fig. 7A). Relative abundances also increased from 0.13 to 0.20 in reaches without and with giant otters respectively. There was no significant difference in average values or the temporal distribution of relative abundances in 3 h bins between reaches with and without giant otters (Fig. 7A, Kruskal-Wallis rank sum test, P = 0.563, Kolmogorov–Smirnov test, P = 0. 0.596). These similarities were reflected in the strong overlap in kernel density estimates of neotropical otter activity in reaches with and without giant otters (Figs. 7B and 7C, mean ±95% CI [0.85, 0.73−0.98]; P = 0.428). There was also no significant difference in neotropical otter activity levels in reaches with and without giant otters (Figs. 7B and 7C, Wald test P = 0.214).

Seasonal patterns in diurnal activity

Our results show that diurnal activity of both otter species was not strongly affected by seasonal river level changes. While such findings require confirmation from other study sites, our results strongly suggest that any seasonal variation in river based activity is likely to be secondary to changes caused by human disturbances. There did however appear to be seasonal patterns to species coexistence, with species encountered closer in space and time during the low river level months (September-November). The small number of same day co-occurrences suggests that competition could be avoided by a combination of both spatial and temporal separation when the species occur in sympatry. Spatial and temporal separation is possible for giant and neotropical otters as both species use scent marking to maintain territories and reduce negative inter and intra-specific interactions (Duplaix, Evangelista & Rosas, 2015; Rheingantz, Santiago-Plata & Trinca, 2017). Although the small sample size limits the insight possible, increased proximity during low river levels was expected based on the relatively limited availability of aquatic feeding habitats compared with the high water level season in the study region and agrees with findings from Suriname (Duplaix, 1980), where both species were reported as likely using the same pools at different times to feed during the dry season.

Both species were active throughout the day across our study rivers. The general patterns in diurnal activity we found follow activity patterns reported for both species (De Almeida & Ramos Pereira, 2017; Duplaix, Evangelista & Rosas, 2015; Rheingantz, Santiago-Plata & Trinca, 2017). The midday resting, and morning and afternoon activity peaks we found for giant otters follows very closely pattens reported from previous studies (e.g., Brazilian Pantanal (Leuchtenberger et al., 2014) and Suriname (Duplaix, 1980)). The peaks of activity in the morning and reduced activity in the afternoon we found for the neotropical otter follows den use patterns reported from the Brazilian Pantanal with low human population densities (Rheingantz et al., 2016). Such consistency provides support for a lack of systematic bias in the results from our extensive boat surveys. River based activities in both species include hunting and movement to and from dens (Duplaix, Evangelista & Rosas, 2015; Rheingantz, Santiago-Plata & Trinca, 2017), it is therefore unsurprising that activity was observed throughout the day. Daytime river based activity has been widely documented for both species but as far as we are aware there are no similar broad scale standardized studies to enable direct comparisons with our results. Future studies including the addition of camera traps monitoring dens in our study area would be useful to facilitate comparisons with the species activity in other neotropical regions (Leuchtenberger et al., 2014; Rheingantz et al., 2016).

Diurnal activity in relation to human disturbances

We found that giant otters were more active in river reaches with less human activity, with a greater relative abundance at reaches with none or few boats and no fishing nets. There is no evidence to suggest that differences in species activity patterns were due to differences in daylight variation as in our low latitude study area (central latitude of 0.93 N) the sun rises at approximately 06:30 am year round (day length is 11.95 h in December and 12.05 h in June). Neotropical otters showed similar patterns of diurnal river based activity across reaches with different degrees of human disturbances (boat use and with or without fishing nets). These results are not surprising as neotropical otters can be found in areas with high levels of human influence, including areas where agricultural and livestock activities occur (Pardini & Trajano, 1999; Rheingantz, Santiago-Plata & Trinca, 2017; Rheingantz et al., 2011; Smith et al., 2020).

As nocturnal activity has been recorded in both species (Duplaix, Evangelista & Rosas, 2015; Rheingantz, Santiago-Plata & Trinca, 2017) additional data from complementary methods such as camera traps and/or telemetry would generate additional insight across the 24 h period. Activity (diurnal and nocturnal) generally occurs close to the dens of both species (Evangelista & Rosas, 2011; Leuchtenberger et al., 2014; Rheingantz et al., 2016), and a previous study found no giant otter dens in the areas most intensely used by humans in the study area (Oliveira, Norris & Michalski, 2015). It is therefore unlikely that the absence of diurnal observations was due to a temporal change in giant otter activity in the most disturbed river reaches. The most plausible explanations are that giant otters have either moved away from and/or been hunted out of areas more intensely used by humans (Raffo et al., 2022). While our data provides a broad scale assessment, it is likely that giant otters with complex social interactions will also demonstrate population and individual level changes to human disturbances. For example, studies with ecotourism suggest giant otters can adapt to well regulated low intensity human disturbances (Barocas et al., 2022). It is possible that changes in activity and other behaviors to avoid humans could be facilitated by social learning in giant otters (Barocas et al., 2022; Davenport, 2010; Schmelz et al., 2017). But, to date there have been very few long term behavioral studies of wild groups that are needed to confirm such a theory.

Neotropical otter activity

Our findings support those from previous studies that suggest no direct competition between giant and neotropical otters. As far as we are aware there have been no reports of direct interspecific competition between any otter species. Indeed, sympatry facilitated by habitat, trophic and temporal separation has been widely documented between otter species across the globe, with sympatry in up to three otter species found in Thailand (Ebensperger & Botto-Mahan, 1997; Krupa, Borker & Gopal, 2017; Kruuk et al., 1994; Lélias, Lemasson & Lodé, 2021; Somers & Purves, 1996). We found giant and neotropical otters were on average approximately 12 km and 3 h apart when detected on the same day, which indicates that competition is avoided both temporally and spatially. As suggested by previous studies our findings support the idea that neotropical otters can be more active later in the day in more disturbed areas (Rheingantz et al., 2016), but that human disturbances do not appear to generate changes at the same scale as the sympatric giant otter.

The reduced presence and stable relative abundances in reaches with fewer boats and nets suggests that there are other factors limiting diurnal river use by neotropical otters in these relatively undisturbed reaches. There are several non-mutually exclusive explanations for why neotropical otters did not show increased activity as expected in least disturbed reaches. We suggest the most plausible was the presence of giant otters. Unlike the giant otter, neotropical otters did not increase activity in reaches with few human disturbances. It was expected that neotropical otters would increase activity (i.e., “if” and “when” they were active) where there was less human disturbances (Rheingantz, Santiago-Plata & Trinca, 2017). The neotropical otter is more generalist with a much broader diet (including diverse small bodied vertebrate species) and less restrictive habitat requirements as they are found across diverse tropical and terrestrial aquatic habitats from Mexico to Argentina (De Almeida & Ramos Pereira, 2017; Rheingantz, Santiago-Plata & Trinca, 2017). As such this more generalist species is not expected to show such strong broad scale responses to environmental changes across 218 km of Amazonian rivers as the more sensitive giant otter. As giant otters were recorded in the reaches with fewer human disturbances, we can therefore safely assume that these reaches also had suitable habitat and prey availability for the more generalist neotropical otters. The most plausible explanation for the observed broad-scale differences in neotropical activity is that neotropical otters may use the main river channels less in the presence of giant otters and use more the smaller perennial inland streams (typically less than 2 m wide and 1 meter deep) that were not possible to survey with motorized boats.

In the case of our study, neotropical otter activity increased between 15–17 in river reaches used by giant otters. These findings support those from perhaps the earliest study showing sympatry that suggested neotropical otters could be more crepuscular and more nocturnal when sympatric with giant otters (Duplaix, 1980). Nocturnal activity by neotropical otters has been recorded along the study rivers (Michalski et al., 2021) and future studies could focus on comparing den activity recorded by camera traps in river reaches with and without giant otters to test if nocturnal activity of neotropical otters also changes.

Although human disturbances may increase temporal overlap among terrestrial carnivores (Sévêque et al., 2020; Sévêque et al., 2021), our findings show that human river use dramatically reduced giant otter activity and therefore reduced spatial overlap with neotropical otters. It is possible that the absence of the larger bodied giant otter could enable neotropical otters to remain active for longer. This could lead to increased feeding opportunities, which could result in population increases and expansion. At the same time any increased neotropical otter activity could also increase negative interactions along rivers used by humans. We are not aware of studies quantifying such dynamics in neotropical otters. Establishing how anthropogenic impacts that reduce the occurrence of giant otters may subsequently influence activity, trophic dynamics and populations of sympatric neotropical otters should be a priority for future studies.

A potential caveat to our findings is that our analysis specifically focuses on broad scale river based activity patterns. Both species also use terrestrial habitats to different degrees and the dynamics of river and terrestrial activity in response to disturbances remains unexplored. Additionally, the use of boat surveys means that it was more likely to detect the larger and less secretive giant otter. Neotropical otters are smaller and solitary, which makes them harder to detect from boats. Here we assume that such differences in detectability do not bias our broad scale comparisons among river reaches. However, more detailed studies are required to determine more localized responses at the scale of individuals and/or extended family groups in the case of giant otters.