Connectivity between clusters was evaluated using circuit the-
ory (McRae et al., 2008) and the program Circuitscape (McRae and
Shah, 2009). Home ranges were used as the source and end nodes
of the circuit. Home ranges were combined into a single node if
they were within 20 km of each other, unless they were separated
by a fence. We used the one-to-many criterion, whereby one
source was connected to all end nodes in an iterative fashion.
The habitat use index created from the RSF model was used as
the resistance surface. Areas with a high probability of use were as-
sumed to allow a greater current flow (or less resistance) than
those with a low probability of use. Owing to the processing limi-
tations of the Circuitscape program, the study area was segmented
into five sections and the habitat use index was generalised to
500 m2 pixels. We began by using only the habitat use index as
the resistance surface, but then further limited movement by add-
ing absolute barriers. Based on known elephant habitat use, we
hypothesised that distance to water, slope, and human population