Exploring multi-level model dynamics: performance and accuracy (WIP)

The choice of resolution for a simulation model at a given scale is a trade-off between the level of accuracy offered by the model and the computational cost of its execution. The understanding of this trade-off requires insight in how model resolution and system scale influence accuracy and computational cost. This paper examines performance and accuracy measurements obtained from models of a simple scenario simulated at different spatial resolutions and different scales. The base model under consideration consists of a battalion formed by four tanks moving towards a fixed point on a two dimensional lattice with a certain height profile. Changes in computational cost and prediction accuracy are studied for different levels of spatial aggregation and model variations studying individual tanks and aggregated battalions. The findings are explained based on model specification choices and the adopted aggregation mechanisms. From this analysis, general propositions are derived which improve our understanding of multi-resolution modeling.