Modeling with Multiple Models Made Easy
New code allows scientists to generate and analyze multiple models that vary in how processes are represented.
Researchers developed a new modeling software package that allows many alternative models to be posed, run, and analyzed as an ensemble, saving scientists time and providing a path to decrease uncertainty in modeling analyses.
There are many ways to represent real-world processes in computer models. But it is common that only a single representation is used in any given model, leading to results that are model specific. This new code now allows the modeling community to move away from the single-representation method to using many alternative models in a single study for a richer analysis that more broadly encompasses the current state of knowledge about ecosystem processes.
Alternative ways that real-world processes can be represented in computer models is a huge source of uncertainty in model output. Yet, tools and modeling systems to examine these alternatives are not available. Researchers at Oak Ridge National Laboratory and a team of national and international collaborators have developed software that can combine alternative ways to represent many real-world processes into a complete set of all possible combinations of the alternatives. This will give a full range of possible model results and goes beyond the single-instance approach to running models. The software also includes novel tools for analysis of model sensitivity to alternative process models.
Anthony P. Walker
Oak Ridge National Laboratory
Oak Ridge National Laboratory Terrestrial Ecosystem Science Scientific Focus Area and Next-Generation Ecosystem Experiments (NGEE)–Tropics project by the Office of Biological and Environmental Research within the U.S. Department of Energy (DOE) Office of Science.
Walker, A. P. et al. “The multi-assumption architecture and testbed (MAAT v1.0): R code for generating ensembles with dynamic model structure and analysis of epistemic uncertainty from multiple sources.” Geoscientific Model Development 11(8), 3159–3185 (2018). [DOI:10.5194/gmd-11-3159-2018]