Improving the computational efficiency of an agent-based spatiotemporal model of livestock disease spread and control

Abstract

Agent-based models (ABMs) are well suited to representing the spatiotemporal spread and control of disease in a population. The explicit modelling of individuals in a large population, however, can be computationally intensive, especially when models are stochastic and/or spatially-explicit. Large-scale ABMs often require a highly parallel platform such as a high-performance computing cluster, which tends to confine their utility to university, defence and scientific research environments. This poses a challenge for those interested in modelling the spread of disease on a large scale with access only to modest hardware platforms. The Australian Animal DISease (AADIS) model is a spatiotemporal ABM of livestock disease spread and control. The AADIS ABM is able to complete complex national-scale simulations of disease spread and control on a personal computer. Computational efficiency is achieved through a hybrid model architecture that embeds equation-based models inside herd agents, an asynchronous software architecture, and a grid-based spatial indexing scheme.