Bridging Scales and Paradigms in Natural Systems Modeling

Abstract

I present a new modeling formalism that enables multiple-scale, multiple-paradigm, and modular modeling. The formalism starts with a generalization of the semantics of scientific observations, where specialized observation classes compute their states by running models, using the states of the dependent observations as input, inheriting, intersecting and harmonizing their topologies of time and space. This formalism, called semantic meta-modeling, offers a uniform and cohesive approach that encompasses data management, storage, querying and many aspects of traditional modeling. I will show how simple, elegant model specifications can be rewritten into queries that can be run on a semantic database to produce semantically annotated model results. The algorithm automatically operates context translation, matching probabilistic with deterministic data and models, performing data-driven structural transformations of model structure as required by the context, and seamlessly mixing traditionally isolated paradigms such as agent-based with process-based or temporally- with spatially-explicit.