Spatiotemporal interactions facilitate sympatry in a diverse mammalian community

Abstract

Understanding mechanisms underlying coexistence among potential competitors, and between predators and prey, is a persistent challenge in community ecology. Using 6 years (2013–2018) of camera‐trapping data and species interaction models, we investigated the spatiotemporal patterns of inter‐ and intra‐guild interspecific interactions in a diverse terrestrial mammalian community in Pakke Wildlife Sanctuary and Tiger Reserve (PTR), Northeast India. We found no evidence of spatial interaction among apex predators (tiger Panthera tigris, leopard Panthera pardus, and dhole Cuon alpinus). However, dholes temporally separated themselves from tigers and leopards. Among small carnivores, marbled cat (Pardofelis marmorata) and leopard cat (Prionailurus bengalensis) exhibited temporal separation, whereas leopard cat overlapped spatially and temporally with other small carnivores. Herbivores exhibited neither spatial nor temporal separation with each other. All apex predators exhibited diel activity and space‐use patterns to overlap with their preferred prey. Our results suggest that the assembly of the diverse mammalian community of PTR is a complex process, and coexistence among potential competitors, and predators and prey is likely facilitated by several mechanisms including spatial and temporal segregation, and potentially dietary separation.

INTRODUCTION

Human population growth and associated agricultural expansions have resulted in the destruction and fragmentation of wildlife habitat to the extent that many wildlife species now exist in small habitat fragments, often restricted to protected areas (Bleich, 2016; Geldmann et al., 2014; Srivathsa et al., 2014). This can influence species interaction patterns, which in turn may affect the structure and function of the ecological community that they are a part of; behavioral mechanisms such as resource partitioning and shifts in activity patterns may play important roles in promoting species coexistence and enhancing biodiversity (Davies et al., 2007; Dorazio & Karanth, 2017; Palomares & Caro, 1999; Steinmetz et al., 2021). Thus, understanding spatiotemporal interactions that facilitate sympatry among species is not only ecologically relevant, but it can also be important from a conservation perspective.

Mechanisms hypothesized to facilitate sympatry among mammalian species with similar resource needs include diet partitioning, microhabitat partitioning, or differences in activity cycle (Cole, 1949; Harihar et al., 2011; Karanth et al., 2017; Karanth & Sunquist, 1995; Kronfeld-Schor & Dayan, 2003; Lucherini et al., 2009). Spatiotemporal segregation allows species to coexist by reducing the intensity of interactions (e.g., Sladecek et al., 2017; Sosa-Lopez & Mouillot, 2007; Vanak et al., 2013). When resources are aggregated, and spatial segregation of the microhabitat is not possible, sympatry among carnivores results in various forms of interference competition, including intimidation of the subordinate species, kleptoparasitism, and sometimes complete spatial or temporal exclusion of the subordinate species (Cusack et al., 2017; Fedriani et al., 1999; Karanth et al., 2017; Prugh & Sivy, 2020). The landscape of fear created by the potential of a lethal interaction with the dominant species may cause subordinate species to adjust their strategies and avoid certain areas frequented by the dominant species or times of the day when the dominant species is most active (Bianchi et al., 2016; Gompper et al., 2016; Vanak & Gompper, 2010). There is thus a push-and-pull mechanism at play whereby predators concentrate their foraging activities in areas of high prey density or during times when prey are most active (Lima, 2002), while prey tend to avoid sites frequented by the predators, and times when the risk of predation is high (Hopcraft et al., 2010).

Our goal was to investigate the spatiotemporal pattern of species interactions that may facilitate the coexistence of potential competitors, predators, and prey species in a diverse mammalian community in Pakke Wildlife Sanctuary and Tiger Reserve (PTR) (92°36′ E, 26°54′ N), India. PTR is reported to be home to 41 (95% CI = 39–42) species of mammals (Appendix S1: Table S1), including seven species of carnivores, 10 species of herbivores and 11 species of omnivores (Chaudhary et al., 2022; Mukherjee et al., 2019; Velho et al., 2016). Situated in the Eastern Himalayas, PTR is critically important for conservation because it is a biodiversity hotspot, and harbors several threatened and endangered species, such as the tiger (Panthera tigris), dhole (Cuon alpinus), clouded leopard (Neofelis nebulosa), Asiatic golden cat (Catopuma temminickii), and Asian elephant (Elephas maximus) (Chaudhary et al., 2022). Although some species such as the tiger, leopard (Panthera pardus), and dhole have received limited scientific attention in PTR (Selvan et al., 2013), the ecology of most species of mammals, or mechanisms underlying coexistence of this diverse mammalian community, remain poorly understood, and we sought to fill that knowledge gap.

We used six years (2013–2018) of camera-trapping data and two-species occupancy models to investigate spatial patterns of interspecific interactions among mammals of PTR (MacKenzie et al., 2018; Richmond et al., 2010). Interspecific interactions are often asymmetric and directional, where dominant species influence the behavior of subordinate species (Blanchet et al., 2020). Thus, we used our biological knowledge and information from the literature to form dominant-subordinate species pairs within each guild (large and meso-carnivores, small carnivores, herbivores, and omnivores) (Table 2; Appendix S1: Tables S1 and S2). For the three apex predators (tigers, leopards, and dholes), we investigated their spatial and temporal interactions with each other, and with their preferred prey (Appendix S1: Tables S1 and S2). Predators can create a landscape of fear, forcing prey to avoid areas frequented by the predators and time of the day when predators are most active. Prey can also alter the behavior of predators because they prefer to forage in areas and times when encounters with predators are least likely (Lima, 2002). When spatial segregation or dietary separation are not possible, temporal segregation may allow competing species (or predators and prey) to coexist (Davies et al., 2007; Sunarto et al., 2015). Thus, we examined the diel activity patterns, and overlap in activity patterns to assess evidence for (or against) potential competitors and between predators and prey in PTR. Specifically, we (1) examined the pattern of spatial interaction between each dominant-subordinate species pairs, and between predator–prey pairs using two-species occupancy models and (2) quantified the diel activity patterns of each species and investigated temporal overlap in activity pattern for each species pair using the kernel density estimate, and coefficient of overlap in diel activity pattern. We predicted that (1) the occupancy and detection of subordinate species will be lower in the presence of dominant species; (2) the occupancy of prey species will be lower in the presence of predators, and probability of detection of prey species would be lower when predator is detected; and (3) subordinate species (or prey species) will be temporally less active during times of day when the dominant species (or predators) is most active, leading to little or no overlap in diel activity patterns.

METHODS

Study area

We carried out the study in Pakke Wildlife Sanctuary and Tiger Reserve (PTR), an 862-km² protected area in Arunachal Pradesh (AP) situated within the Eastern Himalayan Region (EHR) in India (Myers, 2003). The dominant landcover type in PTR is Assam valley tropical evergreen forest, which is further divided into nine habitat types: subtropical broad-leaved primary and secondary forest (~40.0%), Himalayan moist temperate forest (~2.0%), tropical moist-deciduous and semi-evergreen forest (~38.0%), secondary bamboo forest (~4.0%), degraded land (~4%), wetlands and riverine (~2%), and grassland (~1%) (Roy et al., 2015) (Figure 1). The terrain is undulating with an elevation ranging from 150 to 2300 m above sea level. The climate is subtropical with temperatures ranging from 12 to 36°C. The average annual rainfall is 2500 mm, primarily received during the monsoon season (June–September). PTR is home to about 41 species of mammals, 282 species of birds, and harbors a highly diverse plant community (Kumar, 2014; Tag et al., 2012). It is protected under the Indian Wildlife Protection Act of 1972 (). Logging and hunting are not allowed within the reserve, but regulated tourism and ranger patrols are allowed; some human habitations are located immediately outside the reserve boundary. PTR is a well-managed protected area with strong leadership that enjoys positive sociopolitical engagement from local tribes (Ghosh-Harihar et al., 2019). Surrounded by community-managed state forests with varying degrees of anthropogenic disturbance, PTR is a part of the Kameng Protected Area Complex, the largest (3500 km²) contiguous forested area in the EHR (Velho et al., 2016). Other protected areas within this complex include Sonai Rupai Wildlife Sanctuary (WLS) in Assam, the Eaglenest, and Sessa Orchid WLS in AP, and the reserve forests associated with them.

[IMAGE OMITTED. SEE PDF]

Camera-trap surveys

Camera traps have received widespread use in ecological research and have been used to estimate species richness, abundance, and other population parameters and to quantify occupancy, habitat use, and species interaction patterns (Karanth et al., 2017; Sunarto et al., 2015; Tobler et al., 2008). Between 2013 and 2018, we conducted annual camera-trap surveys in the winter (November–March) in PTR (Figure 1). We overlaid a 2 km $$$ \times $$$ 2 km grid on the study area and placed 1–3 camera traps (Panthera V4, Cuddeback Attack 1149 or similar; Cuddleback, Green Bay, WI, USA) per grid cell about 25 cm above the ground on elephant trails or human trails. Cameras were triggered by animal movement and remained active 24 h a day for 30–60 days (Table 1). No lure or bait was used. Each grid cell was considered as a survey site, and community and species-specific inferences were made at the grid-cell level.

TABLE 1 Description of camera-trap surveys conducted between 2013 and 2018 in Pakke Wildlife Sanctuary and Tiger Reserve (PTR), India.

Year	No. grid cells	No. survey periods	No. species of mammals	No. carnivores detected	No. herbivores detected	No. omnivores detected
2013	70	8	24	7	7	10
2014	78	12	27	7	9	11
2015	76	12	24	7	8	9
2016	68	12	24	7	9	8
2017	80	6	25	6	9	10
2018	92	10	24	6	8	10

Camera-trap images were managed using Camelot software version 1.5.9 (Hendry & Mann, 2018) and species were identified manually. The detection of a species within 30 min was recorded as a single detection event. For spatial interactions, we pooled detections over five full days (i.e., 5 $$$ \times $$$ 24-h periods) into one camera-trap survey; thus, there were 6–12 secondary sampling periods (surveys) per year. The number of cameras per grid cell ranged from 1 to 3, and the number of grid cells surveyed in any given year ranged from 68 to 92 (Table 1). Based on the existing literature, anecdotal evidence, and our knowledge of PTR and its wildlife, we created a list of 42 species of mammals that are likely to occur in PTR (Appendix S1: Table S1). We excluded flying, aquatic, and small burrowing mammals from this study because they would be unlikely to be detected by cameras under our study design. Mongoose species were difficult to differentiate, and thus, we categorized them as Herpestes sp. irrespective of the species. A similar approach was used for macaque species, and they were categorized as Macaque sp. There were no discernible changes in forest structure, anthropogenic pressures, and management protocols during the study period, and there was no annual variation in estimated species richness of the study area (Chaudhary et al., 2022). Therefore, we assumed no annual variation in spatial and temporal interspecific interactions and pooled data from each grid cell across years resulting in data from 108 unique grid cells. (Table 1).

Data analysis

Spatial interaction pattern

We divided mammals of PTR into four loosely defined categories or “guilds” based on their dietary habits and body size: large and medium-sized carnivores (hereafter, large carnivores; body mass >10 kg), small carnivores (body mass <10 kg), herbivores, and omnivores. To understand spatial and temporal interactions among small mammals that have overlapping prey requirements and are of similar body size, we included all the mustelids, viverrids, and herpestids as small carnivores. Even though Asiatic black bears (Ursus thibetanus) are omnivores, we considered them as carnivores as they are known to usurp and scavenge kills made by tigers.

Based on existing information about the body size, prey requirements and behavior of individual species, we created dominant (species A)–subordinate (species B) species pairs within each guild (large carnivores, small carnivores, herbivores, and omnivores) (Table 2; Appendix S1: Tables S1 and S2). Furthermore, for the three apex predators (tigers, leopards, and dholes), we investigated their spatial and temporal interactions with each other, and with their preferred prey (Appendix S1: Tables S1 and S2). To understand the effect of prey and predators on each other and the direction of their interaction, we analyzed the data both ways; with predator as species A and prey as species B, and prey as species A and predator as species B. For each species pair, we constructed two-species, single-season occupancy models. Note that, by the term “subordinate species,” we refer to the species whose occupancy (or detection) probability is conditional on the occupancy (or detection) probability of the dominant species (MacKenzie et al., 2004; Richmond et al., 2010). The two-species occupancy models are an extension of single-season occupancy models and are particularly useful for quantifying spatial interaction patterns (MacKenzie et al., 2004, 2018). Since species interactions are often asymmetric (Blanchet et al., 2020), we used the conditional two-species occupancy parameterization presented by Richmond et al. (2010), which allows occurrence and detection to be modeled as a function of covariates while also permitting designation of one of the species as dominant species. In this parametrization, species A represents the dominant species, and species B represents subordinate species (Appendix S1: Table S2). This model estimates the following occupancy parameters: probability of occupancy of species A ($$$ {\uppsi}_{\mathrm{A}} $$$), probability of occupancy of species B given A is absent ($$$ {\uppsi}_{\mathrm{Ba}} $$$), and probability of occupancy of species B given A is present ($$$ {\uppsi}_{\mathrm{BA}} $$$). Detection parameters estimated by this model include the following: probability of detection of species A given species B is absent ($$$ p\mathrm{A} $$$), probability of detection of species B given species A is absent ($$$ p\mathrm{B} $$$), probability of detection of species A given both species A and B are present ($$$ r\mathrm{A} $$$), probability of detection of species B given the presence of both species and detection of species A ($$$ r\mathrm{BA} $$$), and probability of detection of species B given the presence of species A and non-detection of species A ($$$ r\mathrm{Ba} $$$). We estimated species interaction factors (SIFs), the likelihood of species A and B being present (ϕ) and detected (δ) at the same grid cell as (Richmond et al., 2010): $$$ \upphi =\frac{\uppsi_{\mathrm{A}}{\uppsi}_{\mathrm{BA}}}{\uppsi_{\mathrm{A}}\left({\uppsi}_{\mathrm{A}}{\uppsi}_{\mathrm{BA}}+\left(1-{\uppsi}_{\mathrm{A}}\right){\uppsi}_{\mathrm{Ba}}\right)}. $$$ $$$ \updelta =\frac{r\mathrm{A}r\mathrm{BA}}{r\mathrm{A}\left(r\mathrm{A}r\mathrm{BA}+\left(1-r\mathrm{A}\right)r\mathrm{Ba}\right)}. $$$

TABLE 2 Species interaction factor for occupancy (ϕ) and detection (δ) along with their lower and upper 95% CIs (LCL, UCL in parentheses) estimated based on the top performing two-species occupancy models (Appendix S1: Figure S1) that tested for the interaction between hypothesized dominant (or predator) (sp A) and subordinate (or prey) (sp B) species pairs.

Sp A	Sp B	Sites (sp A)	Sites (sp B)	Sites (both)	Model	ϕ	δ
Large carnivores
Tiger	Leopard	9	22	72	ΨBA = ΨBa; pA = rA, pB = rBA = rBa	NA	NA
Tiger	Dhole	20	12	61	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Tiger	Clouded leopard	42	5	39	ΨBA = ΨBa; pA = rA, pB = rBA = rBa	NA	NA
Tiger	Asiatic black bear	48	4	33	ΨBA $$$ \ne $$$ ΨBa; pA = rA, pB = rBA = rBa	1.08 (0.99–1.18)	NA
Leopard	Dhole	28	7	66	ΨBA = ΨBa; pA = rA, pB = rBA = rBa	NA	NA
Dhole	Leopard	7	28	66	ΨBA = ΨBa; pA = rA, pB = rBA = rBa	NA	NA
Leopard	Clouded leopard	55	1	39	ΨBA = ΨBa; pA = rA, pB = rBA = rBa	NA	NA
Dhole	Clouded leopard	38	9	35	ΨBA = ΨBa; pA = rA, pB = rBA = rBa	NA	NA
Small carnivores
Marbled cat	Mongoose	17	30	13	ΨBA = ΨBa; pA = rA, pB = rBA = rBa	NA	NA
Marbled cat	Large Indian civet	3	67	27	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	0.99 (0.95–1.03)	2.2 (1.4–3.1)
Marbled cat	Small Indian civet	21	42	9	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB = rBA = rBa	0.61 (0.33–0.90)	NA
Marbled cat	Common civet	16	43	14	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA = rBa	0.69 (0.42–0.96)	0.70 (0.63–0.77)
Marbled cat	Yellow-throated marten	10	44	20	ΨBA $$$ \ne $$$ ΨBa; pA$$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	0.98 (0.96–1.06)	2.3 (0.17–4.8)
Leopard cat	Marbled cat	57	2	28	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	0.95 (0.90–1)	1.2 (0.31–2.1)
Leopard cat	Mongoose	47	5	38	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	0.98 (0.94–1.00)	2.2 (1.22–3.19)
Leopard cat	Large Indian civet	5	14	80	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	1.01 (0.97–1.06)	2.07 (1.77–2.37)
Leopard cat	Small Indian civet	39	5	46	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	1.02 (0.93–1.08)	2.64 (1.94–3.35)
Leopard cat	Common civet	37	9	48	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	0.95 (0.91–0.99)	2.92 (2.08–3.7)
Leopard cat	Yellow-throated marten	31	10	54	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Large Indian civet	Small Indian civet	43	0	51	ΨBA $$$ \ne $$$ ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	0.93 (0.84–0.97)	NA
Large Indian civet	Common civet	40	3	54	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Yellow-throated marten	Large Indian civet	31	10	54	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	1.02 (0.94–1.10)	1.6 (1.09–2.1)
Herbivores and omnivores
Asian elephant	Gaur	32	1	73	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	1.01 (0.98–1.04)	1.63 (1.58–1.68)
Asian elephant	Sambar deer	2	1	103	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	1.01 (0.98–1.04)	1.65 (1.52–1.79)
Sambar deer	Barking deer	7	3	97	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	0.99 (0.99–1)	1.59 (1.53–1.65)
Capped langur	Macaque	8	19	0	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Predator–prey
Tiger	Wild boar	7	20	74	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Tiger	Sambar deer	2	25	79	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Tiger	Gaur	18	11	63	ΨBA $$$ \ne $$$ ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	1.05 (0.98–1.12)	NA
Leopard	Wild boar	7	7	87	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Leopard	Barking deer	4	10	90	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	1 (0.96–1.03)	NA
Leopard	Sambar deer	2	12	92	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Dhole	Barking deer	3	30	70	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA = rBa	1.03 (0.99–1.07)	1.24 (0.97–1.50)
Dhole	Wild boar	4	25	69	ΨBA$$$ \ne $$$ΨBa; pA$$$ \ne $$$rA, pA$$$ \ne $$$rA, pB$$$ \ne $$$rBA$$$ \ne $$$rBa	0.99 (0.98–1.00)	1.59 (1.08–2.10)
Prey–predator
Wild boar	Tiger	20	7	74	ΨBA = ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	NA	NA
Sambar deer	Tiger	25	2	79	ΨBA $$$ \ne $$$ ΨBa; pA $$$ = $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	1.002 (0.97–1.03)	1.40 (1.30–1.60)
Gaur	Tiger	11	18	63	ΨBA $$$ \ne $$$ ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	1.05 (0.98–1.12)	NA
Wild boar	Leopard	7	7	87	ΨBA $$$ \ne $$$ ΨBa; pA $$$ = $$$ rA, pB = rBA = rBa	1.02 (0.96–1.03)	NA
Barking deer	Leopard	10	4	90	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	1.006 (0.96–1.06)	1.51 (1.30–1.70)
Sambar deer	Leopard	12	2	92	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA = rBa	0.99 (0.99–1.005)	NA
Barking deer	Dhole	30	3	70	ΨBA$$$ \ne $$$ΨBa; pA$$$ \ne $$$rA, pB$$$ \ne $$$rBA$$$ \ne $$$rBa	0.99 (0.99–1.07)	1.29 (1.01–1.56)
Wild boar	Dhole	25	4	69	ΨBA $$$ \ne $$$ ΨBa; pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa	0.99 (0.98–1.00)	1.59 (1.08–2.10)

Independence of occupancy of species B from species A is indicated by $$$ \upphi =1;\upphi <1 $$$ indicates species B is less likely to co-occur with species A and $$$ \upphi $$$ > 1 indicates species B is more likely to co-occur with species A (Richmond et al., 2010). Similarly, $$$ \updelta =1 $$$ indicates independent detection of species B from species A, $$$ \updelta $$$ < 1 indicates species B is detected less when species A is detected and $$$ \updelta $$$ > 1 indicates overlap or more co-detection of species B with species A than expected at random (Richmond et al., 2010). We constructed models where (1) occupancy and detection probabilities of species B were not conditional on that of species A ($$$ {\uppsi}_{\mathrm{BA}}={\uppsi}_{\mathrm{Ba}} $$$) and (pA = rA, pB = rBA = rBa); (2) occupancy of species B is conditional on the occupancy (but not detection) of species A ($$$ {\uppsi}_{\mathrm{BA}}\ne {\uppsi}_{\mathrm{Ba}} $$$) and detection of species A is conditional on occupancy of species B (pA $$$ \ne $$$ rA, pB = rBA = rBa); (3) occupancy and detection of species B are conditional on the occupancy of species A ($$$ {\uppsi}_{\mathrm{BA}}\ne {\uppsi}_{\mathrm{Ba}} $$$ and pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA = rBa); and (4) occupancy of species B is conditional on the occupancy of species A ($$$ {\uppsi}_{\mathrm{BA}}\ne {\uppsi}_{\mathrm{Ba}} $$$) and detection of species A is conditional on occupancy of species B and detection of species B is conditional on the occupancy and detection of species A (pA $$$ \ne $$$ rA, pB $$$ \ne $$$ rBA $$$ \ne $$$ rBa). We selected the best-supported model as the one with the lowest Akaike information criterion (AIC) value (Burnham et al., 2011). If the top model assumed that the occupancy (or detection) of B was not conditional on occupancy (or detection) of species A, the species interaction factor was not estimated. Because size of the grid cells used in our study was much smaller than home ranges of most of the mammalian species, we interpret the occupancy parameters as space use rather than true occupancy. We implemented all models in program Presence 13.4 (Hines, 2006) through RPresence (MacKenzie & Hines, 2018) package for R 4.1.3 (R Core Team, 2022).

Assessing activity patterns and overlap in temporal activity patterns

We classified diel activity patterns into four categories: diurnal (one hour after sunrise and one hour before sunset), nocturnal (one hour after sunset and one hour before sunrise), crepuscular (one hour before and after sunrise and sunset), and cathemeral (irregular throughout the day) (Prat-Guitart et al., 2020). At PTR, the sun rises between 5:00 and 6:00 am and sets between 4:30 and 5:30 pm during December–March when the surveys were conducted. We considered only the time of encounter, ignoring the calendar date, thereby collapsing all encounters into a single 24-h period. We estimated the core 50% activity period for each species using the circular.modal.region function from the Circular package (Lund et al., 2017). We estimated the smoothing parameter using the getbandwidth function of Overlap package (Ridout & Linkie, 2009).

To quantify the overlap between the species pair, we used Overlap package (Meredith & Ridout, 2020) in R 4.1.3 (R Core Team, 2022). This method assesses the activity pattern of each species using kernel density and estimates the coefficient of overlap (∆) in activity pattern (Ridout & Linkie, 2009). We estimated the coefficient of overlap using the recommended nonparametric estimators $$$ \left({\hat{\Delta }}_4\right) $$$ for large sample size (≥75 samples). For each overlap coefficient, we obtained 95% CIs based on 10,000 bootstrap samples (Linkie & Ridout, 2011).

RESULTS

During this study period, we detected 28 species of mammals, but the number of species detected varied across years, ranging from 24 in 2013, 2015, 2016 and 2018 to 27 in 2014 (Table 1). Asiatic golden cat, binturong (Arctictis binturong), goral (Naemorhedus goral), serow (Capricornis thar), slow loris (Nycticebus bengalensis), Masked civet (Paguma larvata), Indian hare (Lepus nigricollis), Hystrix sp., and Lutrogale sp. had very few detections; these species were excluded from further analysis. Detailed analyses were conducted on 19 species (Appendix S1: Tables S1 and S2), which included 8 directional species pairs in the large carnivore guild, 14 species pairs in the small carnivore guild, 4 species pairs in herbivore and omnivore guilds, and 8 species pairs in the predator–prey category (Table 2; Appendix S1: Table S2).

Spatial co-occurrence

Large carnivores

Among the eight large carnivore species pairs, the model that received the highest support assumed that the occupancy and detection probabilities of subordinate species were not dependent on that of dominant species for all species pairs except tigers—Asiatic black bear (Table 2). There was some evidence that Asiatic black bears spatially overlapped with tigers (Table 2; Appendix S1: Figure S1), but the 95% CIs of regression coefficients for $$$ {\uppsi}_{\mathrm{BA}} $$$ and $$$ {\uppsi}_{\mathrm{Ba}} $$$ overlapped zero, and 95% CI for ϕ overlapped 1 (Table 2; Appendix S1: Table S3).

Small carnivores

Among the 14 small carnivore species pairs, the best-supported model assumed that occupancy probability of the subordinate species was conditional on the occupancy of dominant species for all but two-species pairs (Table 2). However, 95% CI of the difference in beta coefficient of $$$ {\uppsi}_{\mathrm{BA}} $$$ and $$$ {\uppsi}_{\mathrm{Ba}} $$$ overlapped zero (Table 2) and 95% CI for ϕ overlapped 1 for all the species pairs, except for three species pairs: marbled cat and small Indian civet (ϕ = 0.61, 95% CI: 0.33–0.90), marbled cat and common palm civet (ϕ = 0.69, 95% CI: 0.42–0.96), and leopard cat and common palm civet (ϕ = 0.95, 95% CI: 0.90–0.99) (Table 2; Appendix S1: Figure S1).

The model that assumed detection of subordinate species to be conditional on the presence and detection of the dominant species received higher support for eight small carnivore species pairs (Table 2). Conditional on the presence and detection of marbled cat, the probability of detection of large Indian civet (δ = 2.20, 95% CI: 1.40–3.10) was higher than when marbled cat was absent or was potentially present but not detected. Similarly, conditional on the presence and detection of leopard cat, probability of detection of large Indian civet (δ = 2.07, 95% CI: 1.77–2.37), small Indian civet (δ = 2.64, 95% CI: 1.94–3.35), and common palm civet (δ = 2.92, CI: 2.08–3.70) was higher than when leopard cat was absent or present but not detected. In contrast, conditional on the presence and detection of marbled cat, the probability of detection of small Indian civet was lower (δ = 0.70, 95% CI: 0.63–0.77) when marbled cat was present and detected. The 95% CI for δ overlapped 1 for all other species pairs (Table 2).

Herbivores and omnivores

Among the four herbivore and omnivore species pairs, the model that assumed that the occupancy of subordinate species was conditional on the occupancy of dominant species received higher support for all but one species pair (Table 2). However, the influence of dominant species on occupancy of subordinate species was not substantial as the 95% CI of the difference in beta coefficient of $$$ {\uppsi}_{\mathrm{BA}} $$$ overlapped 0 (Appendix S1: Table S3) and 95% CI for ϕ overlapped 1 for all three pairs (Table 2; Appendix S1: Figure S1). The model that assumed that the detection of subordinate species was conditional on the presence and detection of the dominant species received higher support for the same three pairs (Table 2). Conditional on the presence and detection of elephant, probability of detection of sambar deer (δ = 1.63, 95% CI: 1.58–1.68) and gaur (δ = 1.65, 95% CI: 1.52–1.79) was higher than when elephants were absent, or present but not detected (Appendix S1: Figure S1). Similarly, probability of detection of barking deer was higher when sambar deer were present and detected as compared with when sambar deer were absent or potentially present but not detected (δ = 1.59, 95% CI: 1.53–1.65) (Table 2).

Predator–prey

Among the eight predator–prey species pairs, in scenarios where predators were considered as species A or the dominant species, the model that assumed the occupancy of prey to be conditional on the occupancy of predators received higher support for three species pairs (Table 2). However, the spatial aggregation of predators with prey was insubstantial as the 95% CI for the difference in beta coefficient of $$$ {\uppsi}_{\mathrm{BA}} $$$ and $$$ {\uppsi}_{\mathrm{Ba}} $$$ overlapped zero  and 95% CI for ϕ overlapped 1 for other all three pairs (Table 2, Appendix S1: Table S3). The model that assumed that the detection of subordinate species was conditional on the presence and detection of the dominant species received higher support for two-species pairs (Table 2). Conditional on the presence and detection of the dhole, probability of detection of wild boar (δ = 1.59, 95% CI: 1.08–2.10) was higher than when dholes were absent or were potentially present but not detected (Table 2). The CI for δ overlapped 1 for the dhole and barking deer (Table 2; Appendix S1: Table S3, Figure S1).

Prey–predator

Under the scenario where prey was considered as species A (dominant species) to investigate if the occupancy and detection of predator was dependent on the presence of prey species, the model that assumed the occupancy of predators to be conditional on the occupancy of prey received higher support for all species pairs except one (Table 2). However, the interaction was insubstantial as the 95% CI for the difference in beta coefficient of $$$ {\uppsi}_{\mathrm{BA}} $$$ and $$$ {\uppsi}_{\mathrm{Ba}} $$$ overlapped zero (Appendix S1: Table S3) and 95% CI for ϕ overlapped 1 for all the species pairs (Table 2; Appendix S1: Table S3, Figure S1).

The model that assumed that the presence and detection of the predator was conditional on detection of prey received higher support for four species pairs (Table 2). Conditional on the presence and detection of sambar deer, probability of detection of the tiger was higher than when the sambar deer were absent or present but not detected (δ = 1.4, 95% CI: 1.30–1.60) (Table 2). Similarly, conditional on the presence and detection of the barking deer, probability of detection of the leopard (δ = 1.51, 95% CI: 1.30–1.70) and dhole (δ = 1.29, 95% CI: 1.01–1.56) was higher than when the barking deer were absent or present but not detected. Conditional on the presence and detection of the wild boar, probability of presence of dholes was higher than when the wild boar was absent or present but not detected (δ = 1.59, 95% CI: 1.08–2.10) (Table 2).

Diel activity patterns and overlap in diel activity patterns

Large- carnivores

Clouded leopards and tigers exhibited primarily nocturnal activity patterns, with 50% of their core activity occurring between 6:00 pm and 2:00 am. Dholes were primarily diurnal, with 50% of their core activity occurring between sunrise and sunset. Asiatic black bears were predominantly nocturnal, and leopards showed peak activity during the day as well as a few hours after sunset (Appendix S1: Figure S2). When testing for overlap in diel activity patterns, we found that the tiger exhibited the highest overlap in activity pattern with the Asiatic black bear ($$$ {\hat{\Delta }}_4 $$$ = 0.79, 95% CI: 0.71–0.89) and lowest with the dhole ($$$ {\hat{\Delta }}_4 $$$ = 0.51, 95% CI: 0.43–0.56). The leopard ($$$ {\hat{\Delta }}_4 $$$ = 0.59, 95% 95% CI: 0.47–0.68) and the dhole ($$$ {\hat{\Delta }}_4 $$$ = 0.30, 95% CI: 0.18–0.38) exhibited lowest temporal overlap with the clouded leopard (Figure 2).

[IMAGE OMITTED. SEE PDF]

Small carnivores

Marbled cat, mongoose, and yellow-throated marten were the only diurnal small carnivores, as 50% of their core activity occurred between 6:00 am and 4:00 pm. Other small carnivores were primarily nocturnal (Appendix S1: Figure S2). Similar sized marbled cats and leopard cats showed low temporal overlap ($$$ {\hat{\Delta }}_4 $$$ = 0.40, 95% CI: 0.26–0.46) (Figure 3). Marbled cats exhibited low temporal overlap with the small Indian civet ($$$ {\hat{\Delta }}_4 $$$ = 0.35, 95% CI: 0.21–0.40), but high temporal overlap with the yellow-throated marten ($$$ {\hat{\Delta }}_4 $$$ = 0.78, 95% CI: 0.69–0.88) (Figure 3). Leopard cat showed high temporal overlap with the large Indian civet ($$$ {\hat{\Delta }}_4 $$$ = 0.92, 95% CI: 0.90–0.96). The leopard cat showed low temporal overlap with the yellow-throated marten ($$$ {\hat{\Delta }}_4 $$$ = 0.25, 95% CI: 0.14–0.26). The large Indian civet showed high temporal overlap with other two subordinate civet species (Figure 3).

[IMAGE OMITTED. SEE PDF]

Herbivores and omnivores

Among herbivores, the elephant was most active during evenings and nights, whereas the gaur was primarily crepuscular. The sambar deer was primarily nocturnal, whereas the barking deer was cathemeral. Hystrix sp. was mainly nocturnal (Appendix S1: Figure S2). Among omnivores, wild boar was strictly diurnal, as were the two primate species. All dominant herbivores showed high degrees of temporal overlap with subordinate species, as was also the case with the two primate species ($$$ {\hat{\Delta }}_4 $$$ = 0.79, 95% CI: 0.65–0.92) (Figure 4).

[IMAGE OMITTED. SEE PDF]

Predator–prey

The wild boar showed the least temporal overlap with the tiger ($$$ {\hat{\Delta }}_4 $$$ = 0.57, 95% CI: 0.50–0.61); the gaur and sambar deer exhibited high temporal overlap with the tiger (Figure 5). The leopard showed high temporal overlap with both the wild boars and sambar deer. The dhole showed high temporal overlap with the wild boar ($$$ {\hat{\Delta }}_4 $$$ = 0.73, 95% CI: 0.65–0.77) and relatively low temporal overlap with the barking deer (Figure 5).

[IMAGE OMITTED. SEE PDF]

DISCUSSION

Community ecology theory informs us that sympatric species that are morphologically and ecologically similar compete and that superior competitors can potentially competitively exclude the inferior ones (Cusack et al., 2017; Davies et al., 2007; Robinson et al., 2014; Vanak et al., 2013). An important question, then, is, how do morphologically and ecologically similar species coexist, and what factors and processes determine community structure? It is generally thought that biotic interactions, especially predation and intraguild competition, play important roles in structure, function, and diversity of ecological communities (Davies et al., 2007; Ripple & Beschta, 2012; Ritchie & Johnson, 2009). For example, competitive interactions among sympatric species can lead to resource partitioning (Schoener, 1974) and character displacement (Grant, 1994; Schluter & McPhail, 1992; Schoener, 1974), mechanisms that can potentially enhance species coexistence and promote diversity. PTR supports a rich diversity of mammals, including species that are ecologically and morphologically similar, and we sought to discern mechanisms facilitating coexistence among sympatric mammals in our study site.

Obligate large carnivores and ambush predators, tigers and leopards, are sympatric through a large extent of their respective ranges. These two species are also sympatric with dholes in many protected and forested areas in India and Southeast Asia. All three large carnivore species have some dietary overlap in PTR (Selvan et al., 2013), which likely results in both resource and interference competition, whereby the larger and dominant tigers potentially affect the subordinate species (leopard and dhole) adversely, as seen in other landscapes as well (Harihar et al., 2011; Lovari et al., 2015). Even though subordinate competitors are expected to seek “competition refuges” by selecting marginal habitat and separating from dominant competitors spatially (Durant, 1998), we found no evidence that dhole and leopard spatially segregated with tigers, or with each other. However, dhole exhibited temporal separation with tigers and leopards by altering their diel activity patterns. We found that dholes in PTR were primarily diurnal, whereas tigers were primarily nocturnal, and leopards were active during early to late morning and evening into the night, which minimizes the possibility of encounters of dholes with tigers and leopards. On the other hand, leopard showed neither spatial nor temporal avoidance of tigers. Large carnivores exhibit considerable behavioral adaptations to variations in resource availability (Karanth et al., 2017; Odden et al., 2010). Even though PTR is a well-managed protected area, the density of tigers, leopards, and dholes is much lower than other protected areas in central and western India (Selvan et al., 2014). PTR also has a high diversity of ungulate prey, including gaur, sambar, barking deer, serow, goral, hog deer (Axis porcinus), porcupine, wild boar, macaques, and capped langur, allowing large carnivores to differentiate preferred prey. Leopards in PTR prey on medium-sized and smaller sized prey including rodents, whereas tigers preferentially feed on large ungulates, potentially alleviating the intensity of intra-guild competition among top predators of PTR (Selvan et al., 2013).

Although the spatiotemporal patterns of interaction among large carnivores is well understood (Chaudhary et al., 2020; Karanth et al., 2017; Mukherjee et al., 2019; Ramesh et al., 2012), we know little about interactions among midsized and small predators. PTR is home to one of the most diverse carnivore guilds where the clouded leopard coexists with three apex predators (Chaudhary et al., 2022). Clouded leopards are potential competitors with leopards in this community, yet we found no evidence of spatial segregation of clouded leopards with leopards or other apex predators. Dietary separation is the most likely mechanism; the three apex predators are ground-dwelling and hunt midsized or large prey, whereas the clouded leopard is primarily arboreal and feed on small prey (Sunquist & Sunquist, 2017). Furthermore, dholes are primarily diurnal and show little temporal overlap with clouded leopards. Leopards often rest and consume their prey on trees, but clouded leopard demonstrate a short window of activity during the night and temporally separate themselves from leopards (Figure 2). We expected the Asiatic black bear to avoid tigers in space and time, as tigers are known to kill bears elsewhere (Seryodkin et al., 2018). Contrary to our expectation, we found some evidence of spatial aggregation and strong evidence of diel activity overlaps between the Asiatic black bear and tiger.

Among small carnivores, the marbled cat and leopard cat are similar in size and have similar food habits and therefore are potential competitors when sympatric (Sunquist & Sunquist, 2017). Leopard cats are nocturnal (Mukherjee et al., 2019; Singh & Macdonald, 2017; Sunarto et al., 2015), and marbled cats have been reported to be both nocturnal (Grassman et al., 2005; Johnson et al., 2009) and diurnal (Sunarto et al., 2015). In PTR, marbled cats were primarily diurnal, and leopard cats were primarily nocturnal, and both small cats showed high degree of temporal separation (Figure 3). However, there was no evidence of spatial segregation between these species (Table 2), suggesting that temporal partitioning is likely the primary behavioral adaptation to facilitate sympatry between the two species. Among other small carnivores, there were only four pairs that showed evidence of spatial avoidance, and only one pair (marbled cat and small Indian civet) showed evidence of temporal avoidance (Figure 3). This suggests that spatial and temporal partitioning between small carnivores does not necessarily mirror each other, and different species pairs use either spatial or temporal partitioning, and likely dietary partitioning to coexist, which we did not investigate.

PTR supports a diverse herbivore community, including the Asian elephant, at least three species of bovids, two species of cervids, and three species of primates (Appendix S1: Table S1). Typically, larger herbivores require large quantity of food, but quality of food matters relatively less as compared with small herbivores that require food with higher nutritional quality and lower grazing time (Owen-Smith, 2015). Among the herbivores of PTR, elephants and gaur showed no evidence of spatial segregation; gaur had higher detection when elephants were present and detected, suggesting similar foraging preferences. Barking deer also had higher detection when sambar deer were present and detected, suggestive of aggregated browsing sites and salt licks. Sambar deer are largely nocturnal, whereas barking deer are active throughout the day. Both species of primates showed no spatial avoidance but exhibited high temporal overlap with each other. Most primates, like humans, are active during the day and roost during the night, which makes temporal overlap intuitive (Fruth et al., 2018).

Predator–prey interactions are complex, with the nature and intensity of the interactions being determined by factors such as competition among predators, prey species diversity and activity patterns, and habitat structure and productivity. Predators move in search of high-density prey and prey move to locate resources and avoid predators (Lima, 2002). These conflicting responses between predators and prey are thought to determine the spatiotemporal pattern of interactions between predators and prey (Farris et al., 2015; Lima, 2002; Persons & Rypstra, 2001; Sih, 1987). Thus, we expected the spatiotemporal patterns of interaction to be bidirectional. We found that, in PTR during our study period, spatial interactions between predators and prey were primarily influenced by prey, as predators altered their space use and diel activity patterns to match those of their preferred prey. Occupancy of most predators was higher when their preferred prey was present (Appendix S1: Tables S2, S3, S4), and predators were detected with higher probabilities when their preferred prey was also present and detected (Table 2; Appendix S1: Table S3, S4). Consistently with our expectations and findings of previous studies, all three apex predators showed high degree of overlap in diel activity patterns to match with those of their preferred prey: dholes with wild boars; leopards with sambars; and tigers with gaurs and sambars (Selvan et al., 2013).

Although spatiotemporal patterns of interspecific interactions among large carnivores have received research attention in some areas (Cusack et al., 2017; Karanth et al., 2017), little information exists about interspecific interactions among midsized and small carnivores and herbivores and that between apex predators and their prey. Using noninvasive camera-trap data, we were able to investigate spatiotemporal interactions both within and among guilds of a diverse, yet little studied, mammalian community in Northeast India. Community assembly is a complex process and relying on observational data, while assuming that the community has reached equilibrium, is contentious (Blanchet et al., 2020). However, when experimental research is not possible due to logistical and conservation concerns, observational studies based on noninvasive camera traps offer a unique opportunity to understand behavioral ecology of rare and data-deficient species such as the clouded leopard and marbled cat. An alternative explanation for the observed pattern is that each carnivore species in PTR may have evolved to be a particularly efficient predator in a specific landcover type. However, a vast majority (63%) of camera traps were in onelandcover type (subtropical broad-leaved evergreen forests) and 85% in just two structurally similar habitat types (63% in subtropical broad-leaved evergreen forests, and 22% in tropical moist-deciduous forests; N = 108; Figure 1), suggesting that the interaction patterns observed in our study cannot be fully explained by habitat-specific specializations.

Our study is one of very few studies that have examined both inter and intra-guild species interaction patterns for a large part of a diverse mammalian community inhabiting a protected area. We show that subordinate species of mammals in PTR organize themselves in space and time to coexist with the dominant species and their predators. Multiple carnivore species coexist in the same protected area mainly via spatial and temporal segregation, and potentially due to dietary partitioning—uninvestigated in this study but an open avenue for future investigations. Intact carnivore guilds are vital for ecosystem structure and function (Ripple et al., 2014), and our study offers novel insights into the spatiotemporal interactions that may facilitate sympatry in a diverse carnivore community in a subtropical ecosystem. Similar studies on other diverse mammalian communities would be needed to test the generality of our findings.

AUTHOR CONTRIBUTIONS

Vratika Chaudhary contributed to conceptualization, methodology, formal analysis, data curation, writing—original draft, and writing—review and editing. Varun R. Goswami contributed to conceptualization, methodology, data curation, formal analysis, writing—review and editing, and supervision. Chandan Ri contributed to conceptualization, data curation, writing—review and editing. James E. Hines contributed to methodology, data curation, formal analysis, writing—review and editing, and supervision. Madan K. Oli contributed to conceptualization, methodology, data curation, formal analysis, writing—original draft, writing—review and editing, and supervision.

ACKNOWLEDGMENTS

This work was supported in part by the University of Florida Biodiversity Institute, University of Florida Informatics Institute, University of Florida Tropical Conservation and Development Program, WildLandscapes International, Rufford Foundation, and Pakke Tiger Reserve, India. We are immensely grateful to Mr. T. Tapi, Mr. K. Rambia, Pakke Wildlife Sanctuary and Tiger Reserve Forest guards and Special Tiger Protection Force Rangers for their support and contributions to data collection. We are grateful to Dr. J. D. Nichols, Dr. A. Mortelliti, and two anonymous reviewers for many helpful comments. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Data and R code needed to perform species interaction occupancy analyses (Chaudhary, 2024) are available from Figshare: .

References

Bianchi, R. D., N. Olifiers, M. E. Gompper, and G. Mourao. 2016. “Niche Partitioning among Mesocarnivores in a Brazilian Wetland.” PLoS One 11(9): [eLocator: e0162893]. [DOI: https://dx.doi.org/10.1371/journal.pone.0162893].