Agent-based integrated land use/transport models: a study on scale factors and transport model simulation intervals

Abstract Modern agent-based models often suffer from long runtimes (computation times). While quite a few studies deal with runtime issues in transport models, there is less research with respect of agent-based, integrated land-use transport (ILUT) models. This paper examines how the transport model interval in an ILUT model suite affects runtime and results. In addition, the possibility of scaling the synthetic population, which is a common approach in agent-based transport models, is tested on the land-use side. Results suggest that transport models do not necessarily need to run every one or two years and larger intervals like five years lead to similar results. However, if intervals are too long, there error of estimated travel times in the years between transport model updates become larger. Scaling the land-use population does not result in large reductions of runtimes if the transport model population is kept uniform. Thus, scaling the land-use model population can be used to actually model every agent individually in the transport model. On the other hand, if the transport model population is scaled down with respect of the land use model, runtimes can be improved. In both cases, the results of the land use model appear to be stable in when using smaller population sub-samples, as far as the resolution of the analysis is not too small.