Fig 1.
The Ruaha-Rungwa conservation landscape (A), within the wider context of Tanzania’s protected area system (inset). A: Sampling grid investigating co-occurrence at the home range scale; B: Sampling grid investigating co-occurrence at the short-term use within the home range scale. Sampling sites consist of 225 km2 grid cells at the home range scale, and of an area of ~1 km2 around 2 km survey transect segments at the short-term use scale. Both the gazetted and effective boundaries for Ruaha NP are depicted. Lunda-Mkwambi GCA comprises both Lunda-Mkwambi North and South, while Rungwa South OA comprises both Rungwa South OA and Rungwa Mzombe OA. Only villages in proximity of the PA complex are shown.
Fig 2.
Mean probability of wild dog site use (with associated standard error) given the presence (BA) or absence (Ba) of lion, from the top-ranked model, at the two spatial scales investigated in Ruaha-Rungwa.
Fig 3.
Species interaction factors (SIFs; with respective standard errors) representing the level of co-occurrence (φ; A) and co-detection (δ; B) between lion, leopard, and wild dog in Ruaha-Rungwa, from the top-ranked model, at the two spatial scales investigated. An SIF value exceeding 1 (dotted line) indicates that the species co-occur more often than expected; a value of less than 1 indicates that the species co-occur less often than expected; and a value of 1 indicates that the species are co-occurring randomly.
Table 1.
Summary of co-occurrence model rankings used to evaluate the role of interspecific interactions on occurrence of lion, leopard, and wild dog in Ruaha-Rungwa, at two spatial scales.
Models in which site use of the subordinate species depends on the presence (ΨBA) or the absence (ΨBa) of the dominant species (conditional models, denoted as ψBa ≠ ψBA) were compared against models in which site use of the subordinate species is independent of the presence of the dominant species (unconditional models, denoted as ψBa = ψBA). For each analysis, only the best-supported conditional and unconditional models are presented; see S2 Appendix for full model rankings.
Table 2.
Site use (ψ), detection probability (p and r), and species interaction factors (SIF–phi and delta) parameter estimates and associated standard errors from the top-ranked model investigating co-occurrence between lion, leopard, and wild dog in Ruaha-Rungwa, at the two spatial scales investigated.
Table 3.
Summary of co-occurrence model rankings used to evaluate the role of interspecific interactions on detection of three sympatric large carnivores (lion, leopard, and wild dog) in Ruaha-Rungwa, at two spatial scales (see main text for additional details).
Only models with strong support (i.e. ΔAICc<2) are presented; for full model rankings see S2 Appendix.