Elephants avoid costly mountaineering

Understanding the behavioural decisions underlying animal movements is a major challenge. Here we report evidence for the importance of the abiotic terrain feature 'gradient' in guiding the movements of African savannah elephants (Loxodonta africana). Global Positioning System (GPS) tracking data overlaid onto digital elevation and surface gradient models show that elephants tend to avoid steep slopes. Energy calculations suggest that even minor hills are considerable energy barriers for heavy animals. Elephants are keystone animals in Africa and Asia [1], and effective conservation planning strategies must integrate a thorough knowledge of the range use and spatial requirements of these magnificent animals. Only with such knowledge can we ensure that elephants will be able to survive despite increasingly aggressive human encroachment into their traditional territory [2]. Moreover, there is much to be learned scientifically from understanding the ecological requirements — as well as limitations — of the last remaining representatives of a once cosmopolitan and ecologically critical megafauna. Early studies of elephant movements deployed radio tracking from the air and provided rather infrequent 'fixes' which painted an incomplete picture of spatial utilisation [3]. Modern GPS collars using a satellite and/or cell-phone link allow us to collect movement data with high temporal and spatial resolution [4], reflecting true range use by also mapping areas not visited. Long-term elephant tracking studies are beginning to show are working to increase simulation accuracy while reducing the computational burden. A central challenge concerns simulation of interacting processes that occur on different time scales; the fast scale imposes a short simulation time step, but that makes the simulation too slow to observe the slow scale. This is being met with new algorithms and hybrid simulations that treat space and stochasticity only as required. Also, given the size and the possible nonlinearity and non-determinism represented by biological models, tools for analysis of models, such as those that provide parametric sensitivity analysis, and for comparing models to data for parameterization and (in)validation are both profoundly needed and in a relatively primitive state. It is easy to describe the ideal simulation tool: it should be able to simulate reactions and diffusion as accurately as needed, account for all relevant mechanical processes, help with model parameterization, validate and discriminate between models using data, and be easy to use. Many modeling tools are aiming towards this goal but it remains elusive, in part because of the extraordinary speed with which improved analysis methods …