Fig 1.
a) Excavation work on the slope where the wildebeest bones were concentrated in 2012. b) One of the wildebeest carcasses recovered in 2012.
Fig 2.
Variable factor map (a) and distribution of samples (b) in the two-component PCA solution for the analysis of skeletal part profiles and their use in discriminating felid and hyenid bone assemblages. C, Hierarchical cluster analysis using the same samples as PCA. Key for PCA: sk: skull, md: mandible, cv: cervical vertebrae, tv: thoracic vertebrae, lv: lumbar vertebrae, cdv: caudal vertebrae, rb: rib, sc: scapula, cox: innominate, h: humerus, r: radius, u: ulna, mc: metacarpal, f: femur, t: tibia, mt: metatarsal, c_t: carpal/tarsal, ph: phalanges. Key for CA: for leopard dens (numbers 4,5 and 6; data from ref 8, 51), spotted hyaena dens (numbers 1,2 and 3; data from ref 60, 48, 71) and for the OCS (number 7).
Fig 3.
Random forest analysis for the skeletal part representation.
Left, error rate for hyenas (Green), felids (Red) and combined oob error rate (black). Stabilization of error was achieved after 1500 decision tress. Right, gini index showing the most important variables for the most accurate classification. The most influential variables are shown on top. (data from ref 8, 51, for leopard dens and data from ref 60, 48, 71, for spotted hyaena dens). Key: cv: cervical vertebrae, tv: thoracic vertebrae, sk: skull, lv: lumbar vertebrae, ph: phalanges, cdv: caudal vertebrae, f: femur, rb: rib, c_t: carpal/tarsal, u: ulna, r: radio, h: humerus, cox: innominate, t: tibia, md: mandible, mt: metatarsal, mc: metacarpal, sc: scapula.
Fig 4.
Plot showing survivorship values (%MAU) at OCS compared to density (%) values for the wildebeest skeleton (ref 70).
Key: cv: cervical vertebrae, tv: thoracic vertebrae, ot: others, lv: lumbar vertebrae, ph: phalanges, cdv: caudal vertebrae, fe: femur, rb: rib, c/t: carpal/tarsal, ul: ulna, ra: radio, hu: humerus, in: innominate, ti: tibia, md: mandible, mt: metatarsal, mc: metacarpal, sc: scapula.
Fig 5.
Variable factor map (right) and distribution of samples (left) in the two-component PCA solution analyzing complete long bones frequencies from Amboseli and Masai Mara spotted hyaena dens (data from ref 48) and OCS.
Fig 6.
Limb shaft circumference type for Syokimau spotted hyaena den (data from ref 60) and OCS.
See definition of types in text.
Fig 7.
Variable factor map (right) and distribution of samples (left) in the two-component PCA solution analyzing furrowing patterns for lions (data from ref 79 and unpublished data from Gidna), spotted hyaenas (data from ref 80) and OCS.
The OCS assemblage shows similar amount of furrowing to the lion samples (all with negative values in the first component), but more than the lions because distal ends of humeri and radii are more furrowed (as in hyenas), hence the separation on the second component
Fig 8.
Taphotypes for lions (pink), spotted hyaena (red) (data from ref 81) and the OCS (green), according to CA.
Notice the separation of the confidence interval between the felid and the hyenid samples and the overlap of both with the OCS sample, which is most closely associated with the confidence ellipse of the felid sample.
Fig 9.
Pits size for the spotted hyena, lion, OCS, cheetah and leopard.
Data for the spotted hyena and lion from ref 78. Data for the cheetah and leopard from [30].