Aim: Large carnivores play a crucial role in trophic cascades, affecting the population dynamics of both co-predators and prey within an ecosystem. Understanding the significance of these carnivores in trophic interactions is essential for developing effective conservation and management strategies. We examined the effects of occupancy dynamics and patterns of species interactions and co-existence within the carnivore guild in the Panna Tiger Reserve in India.

Methodology: We collected camera trap data (two seasons, 2019) in a presence-absence framework and applied multispecies occupancy models to assess the occupancy, co-occurrence, and interactions among species. We also examined activity overlap to understand the temporal segregation in the carnivore guild.

Results: The mean marginal occupancy was highest for leopards in winter (Ψ_winter 0.92±0.02, Ψ_summer 0.63±0.05) and hyenas in summer (Ψ_summer 0.93±0.03, Ψ_winter 0.78±0.03) and was lowest for tigers in both seasons (Ψ_winter 0.62±0.05, Ψ_summer 0.15±0.05). Co-occurrence probability among carnivores was higher in winter than in summer, and conditional occupancy was consistently higher when other species were present. Different environmental factors influenced marginal occupancy and co-occurrence patterns across seasons. Strong temporal overlaps were recorded between tiger–leopard (0.87–0.91) and tiger–hyena (0.78–0.79).

Conclusion: We detected a significant spatial segregation between tigers and leopards, as they prefer different habitat types in different seasons, along with high temporal overlap. Resource availability strongly governs the association of carnivores with their habitat selection. Hyenas demonstrated higher dependency on tigers than on leopards for resources. These findings indicate that co-existence with apex-predator species is feasible through strategic adaptation to fulfil resource requisition.

Data collection

Field methods

We conducted a camera trap survey from 2017 to 2021 during the summer and winter seasons. However, for this study, we used the two seasons of camera trap data for 2019 (winter, 475 stations; summer, 338 stations). We deployed camera traps during January–February for winter sampling (average 32 days) and May–June for summer sampling (average 38 days). We deployed a pair of automated motion-triggered digital camera traps (Cuddeback C1; www.cuddeback.com) in a 2 km² grid cell size framework (Jhala et al., 2018). The average distance between two adjacent camera trap stations was 1.05 km, ensuring intensive sampling. Camera traps were installed on either side (facing each other) of animal trails, forest roads, and the riverbed in mid-slope regions at 30–40 cm above the ground to optimize the capture of large carnivores (Chen et al., 2009; Johnson et al., 2009; Evans et al., 2019). All camera traps were active for 24 hours during the sampling period and programmed to take one photograph per trigger with an interval of 5 second. We checked the camera trap data every 15 days during the observation period. This research was carried out under permit number Technical/4301/, dated on 09 June 2015 and issued by the Principal Chief Conservator of Forest (Wildlife Division), state of Madhya Pradesh, India.

Analytical methods

Retrieved images from camera trap data were identified up to the species level. Data were sorted into species-specific folders for all the sites by season. Following species-level data segregation, we prepared a detection/non-detection matrix for three large carnivores: tiger, leopard, and hyena.

Predictor variables

We obtained data at each camera-trap sampling location by extracting values from remotely sensed raster datasets with a buffer size of 100 meters. We used continuous variables such as tree canopy density (in %), normalized differentiated vegetation index (NDVI), distance to a village (km), distance to a water source (km), land surface temperature (LST, °C), and slope (degree, Table S1). We prepared these thematic layers and resampled at 100-m resolution using the ‘raster’ package (Hijmans, 2024) in R (R Core Team, 2023). We used NDVI and slope to represent vegetation cover and topography, respectively. Distance to a village and distance to a water body were used as surrogates for human disturbance and fundamental resources for carnivores, respectively. LST was used as a substitute to explore the micro-climatic conditions at the camera trap station. Prior to analyses, we standardized all covariates using z-transformation. To check the multicollinearity among the covariates, we applied the cor function in the ‘corrplot’ package (Wei and Simko, 2017) in R 4.3.2 (R Core Team, 2023) to perform Pearson’s pairwise correlation test with a threshold value of |r| ≥ 0.70 (Figure S1, S2; Dornmann et al., 2013).

Multispecies occupancy modeling

We implemented a multispecies occupancy model (Rota et al., 2016) of three large carnivores of PTR using the ‘unmarked’ package (Fiske and Chandler, 2011) in R ver. 3.5.1 (R Core Team, 2023). For multispecies occupancy models, a priori assumptions are not required to determine the dominant or subordination of one species over another (Rota et al., 2016). We investigated how environmental and anthropogenic variables affect the marginal occupancy (without accounting for interactions with other species), co-occurrence (overlap in marginal occupancy between species), and conditional occupancy (effects of one species’ presence on another species’ occupancy and detection) of tiger, leopard, and hyena in PTR. We compiled the data from two seasons (winter and summer of 2019) into 5-day sampling occasions. We selected the best marginal occupancy model of each season based on the lowest AIC value, using the R package ‘MuMIn’ (Bartoń, 2020), to explain the best covariates for each individual species. Based on the top candidate model, we constructed pairwise co-occurrence models between tiger–leopard, tiger–hyena, and leopard–hyena. Furthermore, we ran the multispecies occupancy model for the three carnivore species. Similarly, we used the effort as the detection probability. The best models from all three different multispecies occupancy categories were treated with a penalized likelihood function to obtain a more precise occupancy probability (Clipp et al., 2011).

Temporal activity

To determine the diel activity patterns of carnivores, we considered the independent capture events at 30-min intervals (O’Brien et al., 2003) obtained from the camera trap data. We used the non-parametric kernel-density functions (Ridout and Linkie, 2009) of the ‘overlap’ package (Meredith and Ridout, 2014) in R 3.5.1 (R Core Team, 2023). This methodology yields a coefficient of overlap within the range of 0 to 1, with complete temporal separation between two species at 0 and absolute overlap at 1. A threshold of Δ > 0.8 was considered indicative of strong overlap, while a range of 0.5 < Δ < 0.8 represented moderate overlap, in accordance with the criteria established by Lynam et al. (2013). Since our sample sizes exceeded 75, we employed the Δ4 estimator (Dhat4). To enhance robustness, we computed 10,000 bootstraps for each species and generated 95% confidence intervals (CIs) for temporal overlap estimates (Meredith and Ridout, 2014).

Reference:

• Bartoń K (2024). MuMIn: Multi-Model Inference. R package version 1.48.4, https://CRAN.R-project.org/package=MuMIn.
• Chen, M. T., Tewes, M. E., Pei, K. J., & Grassman Jr, L. I. (2009). Activity patterns and habitat use of sympatric small carnivores in southern Taiwan.
• Clipp, H. L., Evans, A. L., Kessinger, B. E., Kellner, K., & Rota, C. T. (2021). A penalized likelihood for multispecies occupancy models improves predictions of species interactions. Ecology, 102(12), e03520.
• Dornmann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., et al. (2013). Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography 36, 027–046.
• Evans, B. E., Mosby, C. E., & Mortelliti, A. (2019). Assessing arrays of multiple trail cameras to detect North American mammals. PloS one, 14(6), e0217543.
• Fiske, I., & Chandler, R. (2011). Unmarked: an R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of statistical software, 43, 1-23.
• Hijmans R (2024). raster: Geographic Data Analysis and Modeling. R package version 3.6-27, https://rspatial.org/raster.
• Jhala, Y. V., Qureshi, Q., & Nayak, A. K. (Eds.). (2018). Status of Tigers Co-predators & Prey in India. National Tiger Conservation Authority, Government of India, New Delhi & Wildlife Institute of India, Dehradun. TR No./2019/05.
• Johnson, A., Vongkhamheng, C., & Saithongdam, T. (2009). The diversity, status and conservation of small carnivores in a montane tropical forest in northern Laos. Oryx, 43(4), 626-633.
• Lynam, A. J., Jenks, K. E., Tantipisanuh, N., Chutipong, W., Ngoprasert, D., Gale, G. A., ... & Leimgruber, P. (2013). Terrestrial activity patterns of wild cats from camera trapping. Raffles Bulletin of Zoology, 61(1)
• Meredith, M., & Ridout, M. (2014). Overlap: estimates of coefficient of overlapping for animal activity patterns. R package version 0.2, 4.
• O'Brien, T. G., Kinnaird, M. F., & Wibisono, H. T. (2003). Crouching tigers, hidden prey: Sumatran tiger and prey populations in a tropical forest landscape. In Animal Conservation Forum (Vol. 6, No. 2, pp. 131-139). Cambridge University Press
• Rota, C. T., Ferreira, M. A., Kays, R. W., Forrester, T. D., Kalies, E. L., McShea, W. J., ... & Millspaugh, J. J. (2016). A multispecies occupancy model for two or more interacting species. Methods in Ecology and Evolution, 7(10), 1164-1173.
• R Core Team (2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.
• Ridout, M. S., & Linkie, M. (2009). Estimating overlap of daily activity patterns from camera trap data. Journal of Agricultural, Biological, and Environmental Statistics, 14, 322-337.
• Wei, T., & Simko, V. (2017). R Package “Corrplot”: Visualization of A Correlation Matrix (Version 0.84). Vienna: R Foundation for Statistical Computing.