Wildlife ranching, although not considered a conventional conservation system, provides a
sustainable model for wildlife utilization and could be a source of valuable genetic material.
However, increased fragmentation and intensive management may threaten the evolutionary
potential and conservation value of species. Disease-free Cape buffalo (Syncerus caffer
caffer) in southern Africa exist in populations with a variety of histories and management
practices. We compared the genetic diversity of buffalo in national parks to private ranches
and found that, except for Addo Elephant National Park, genetic diversity was high and statistically
equivalent. We found that relatedness and inbreeding levels were not substantially
different between ranched populations and those in national parks, indicating that breeding
practices likely did not yet influence genetic diversity of buffalo on private ranches in this
study. High genetic differentiation between South African protected areas highlighted their
fragmented nature. Structure analysis revealed private ranches comprised three gene
pools, with origins from Addo Elephant National Park, Kruger National Park and a third,
unsampled gene pool. Based on these results, we recommend the Addo population be supplemented
with disease-free Graspan and Mokala buffalo (of Kruger origin). We highlight
the need for more research to characterize the genetic diversity and composition of ranched
wildlife species, in conjunction with wildlife ranchers and conservation authorities, in order to
evaluate the implications for management and conservation of these species across different
S1 Appendix. Supplementary methods.
S1 Fig. Estimated effective population size (Ne) of the buffalo population from each
locality. Vertical lines indicate 95% confidence intervals. Numbers inserted for GNP and
GNP-MNP indicate the value of the upper bound of the 95% CI. The dashed line indicates the
lower 95% CI of GNP. Values are also shown in S3 Table. PVT: Private ranches combined.
S2 Fig. Statistical support for K. The first column of graphs [L(K)] show the mean log likelihood
of each value of K with its associated standard deviation, while the second column (DeltaK)
shows the most likely value of K as determined by the Evanno method. Rows indicate the
full data set (FDS) and the relatives removed (RR) data set. The graphs were generated using
StructureHarvester and further organized in Inkscape v0.92 (https://inkscape.org/).
S3 Fig. Individual assignment plots of the STRUCTURE analyses at K = 2 and K = 3. A–full
data set, B–relatives removed. The plots were generated using the online version of Clumpak
and further organized in Inkscape v0.92 (https://inkscape.org/).
S4 Fig. Discriminant analysis of principal components (DAPC) of the full data set at K = 3.
AENP Cluster: Addo Elephant National Park cluster, GNP-MNP Cluster: Graspan and Mokala
National Park cluster, “Other” Cluster: Third, unknown origin cluster.
S1 Table. Summary statistics of microsatellite loci used in this study. Calculated in Cervus
S2 Table. Mean and variance of the relatedness estimators available in COANCESTRY.
TrioML (values in bold) had the lowest variance for each sampling locality and produces positive
relatedness estimates between zero and one (as does DyadML).
S3 Table. Population summary statistics for each sampling locality.
S4 Table. Relatedness and individual inbreeding statistics.
S5 Table. Mean relatedness within sexes.
S6 Table. Pairwise DJOST and FST values with 95% confidence intervals.
S7 Table. Hardy-Weinberg Equilibrium (HWE) probability tests of each sampling locality,
with the full data set and relatives removed. Data sets from sampling localities conformed to
HWE after relatives were removed.
S8 Table. Original and Bonferroni-corrected linkage disequilibrium p-values of all pairs of
loci in all sampling localities. Both the full data set (FDS) and relatives removed (RR) data set