Information theory provides a powerful conceptual framework for learning, model building, and prediction in the Earth sciences. It extends probability theory in certain important ways that make it particularly applicable to questions related to value and uncertainty associated with models, data, and forecasts:
- How much information does a model or hypothesis contain about the environmental systems we are interested in?
- How much information does a data set contain, and how much is redundant with respect to a specific science or forecasting question?
- What is the best way to combine information contained in local data with information encoded in general hypotheses, laws, or theories?
- How do information theory approaches complement and go beyond existing methods (e.g., Bayesian) for uncertainty analysis?
Last May, a workshop held at the Environmental Hydraulic Institute of Cantabria (IHCantabria) aimed to stimulate new developments in the implementation of information theory to solve practical problems in the Earth sciences. This meeting followed from a workshop held in Germany in 2016 and is linked to a series of upcoming AGU and European Geosciences Union meeting sessions (e.g., “Better Informed Than Uncertain: Applications of Information Theory in the Earth Sciences” to be held at AGU’s Fall Meeting 2018).
Interested participants attended a preworkshop tutorial day, which covered the basics of information theory and gave a broad overview of applications in the Earth sciences. After this tutorial day, participants engaged in moderated discussions, presented posters, and attended lectures on the following topics:
- the main concepts and methods of information theory and its relationship with other fields
- information flow in process networks of complex system dynamics
- the information physics of outlier events sometimes referred to as black swans
- the impact of deforestation on rainfall recycling
Discussion topics included information theory and uncertainty as basic concepts of science, approaches to assimilate information and knowledge of many types and sources into models, the potential of formulating generalized null models based on information theory, the use of information theory for diagnostic model evaluation, and approaches to overcome barriers to practical application of information theory to the Earth sciences.
In addition to the inspiring exchange of concepts and ideas, the workshop produced a joint debate paper and a journal research paper (both under preparation), the decision to organize a 2019 summer school on information theory at IHCantabria, and plans for a 2020 workshop in Canada.
More information on this year’s workshop is available on the workshop’s website and on a community website about information theory in the Earth sciences. To be added to the mailing list, please contact [email protected]
In addition to the authors, the organization team consisted of Hoshin Gupta, Ilias Pechlivanidis, Steven Weijs, Florian Wellman, Ben Ruddell, Praveen Kumar, Rohini Kumar, Gab Abramowitz, Saman Razavi, and Jesus Carrera. We gratefully acknowledge IHCantabria for facilitating the workshop.
—Cristina Prieto (email: [email protected]), Environmental Hydraulic Institute of Cantabria, Santander, Spain; also at Imperial College London; also at Bristol University, U.K.; Uwe Ehret, Karlsruhe Institute of Technology, Germany; and Grey Nearing, University of Alabama, Tuscaloosa