Embracing large calibration spaces to unify model structure across scale: a migration example

The MIDAS (Migration, Intensification, and Diversification as Adaptive Strategies) framework was designed to apply across a range of scales from local rural-urban commuting through to international migration, drawing on prospect theory and push-pull-mooring theory to create a decision framework that places the many different factors that shape livelihoods decisions on a common footing. Two mechanisms that make this possible are i) a specification for all possible elements of a livelihood (income, social services, access to nature, asset use value, etc.) as streams of utility with seasonal structure, as well as monetary and time costs; and ii) a comparatively large calibration space in order that different contexts may be represented through calibration of unknown (or unknowable) parameters. This approach outsources land-use change and natural systems processes to other sub-models, asking only that they speak to MIDAS via the utility layer specification. One of the principal strengths of this approach is that it avoids the problem of the ‘too-simple’ decision model that must be redesigned for every new application and context. As well, I argue that the use of a large calibration space does not disrupt the modeler’s ability to control an experimental space of a subset of key variables and parameters, nor the capacity for a model to make strong inferences. I support these two claims via a specific application of MIDAS to examine coastal migration in Bangladesh under the driver of sea level rise.