An agent-based model of elephant crop consumption walks using combinatorial optimization

Wildlife crop consumption is a worldwide problem. This paper builds on the theoretical framework of profit and utility maximization from economics as established in the theory of optimal foraging, bringing this perspective to the issue of wildlife crop consumption by testing whether elephants forage for crops in an optimal way. Using combinatorial optimization in an agent-based model, in which elephants’ objective is to find a valid walk that maximizes their energy balance. We used empirical data from GPS collars on African savanna elephants to train and test the model. When we focused solely on which terrain blocks the elephants of GNP visit and spend time in, our ABM got 56 percent of these blocks correct. Our ABM performed roughly 25 percent better than two alternative models, including the random walk model. In both subsamples of data that we looked at, the ABM performed better in terms of fitting the data on real walks than the alternative models. The ABM's performance improved, and the alternative models’ performance worsened, when we only looked at data on real walks that involve crop consumption. This suggests that there is more randomness involved when elephants are engaged in foraging activity that do not include crop consumption. At the same time, elephant walks involving crop consumption seem to more closely follow optimizing principles. Findings from this ABM approach support ecological understanding of elephant crop foraging, highlighting the optimal movements involved in crop foraging events as well as the importance of trespassing costs and landscape configuration. It may give conservationists and policy-makers a starting point to use in formulating policies to minimize the harms and costs that result from elephant crop consumption.