Although the choice of the level of spatial and temporal aggregation in sectoral and multi-sectoral models is viewed as critically important determinants of model credibility and impact, these choices are typically made on an ad hoc basis with limited and incomplete testing. This study proposes a more systematic approach to making these choices for normative models based on fundamental optimization and information theory concepts. It develops principles for trading off the accuracy of representation versus parsimony and a ‘modeling roadmap’ to help the modeling community apply the ideas. Simple examples of the application of this methodology and possible extensions of it to non-normative models are included.
This approach and possible extensions to it should help the modeling community improve the credibility and impact of multi-sector dynamics (MSD) modeling systems.
Choosing the granularity of the temporal or spatial resolution of such models is an important modeling decision, often having a first-order impact on model results. This type of decision is frequently made by modeler judgment, particularly when the predictive power of alternative choices cannot be tested. In this study we show how the implicit tradeoffs modelers make in these formulation decisions, in particular in the tradeoff between the accuracy of representation enabled by the available data and model parsimony, may be addressed with established information-theoretic ideas. The study provides guidance for modelers making these tradeoffs or, in certain cases, enables explicit tests for assessing appropriate levels of resolution. We will mainly focus on optimization-based normative models in the discussion here, and draw our examples from the energy and climate domain.