Climate Variability workshop addresses Uncertainty
Climatologists and modelers, statisticians and mathematicians, computational and decision support scientists gathered in Leiden, NL at the Lorentz center on 1 December - a showing of over twenty researchers from MCRN and its SAVI partners. The meeting was organized by Chris Jones of MCRN, Amit Apte of ICTS-TIFR, Henk Dijkstra of LINC, and David Stainforth of CliMathNet with the goal of exploring the mathematical questions of uncertainty specific to climate. Participants from the Netherlands, the United Kingdom, the United States and India discussed, developed, and defined new mathematical questions to understand the sources of uncertainty in climate prediction, and the impact of the uncertainty in climate prediction. Over half of the meeting was spent in one on one and group discussions to promote collaboration and interdisciplinary thinking.
We considered aspects of uncertainty from as fine a scale as the statistical bias of numerical discretizations to the coupling of many sources of uncertainty in the large GCM's which are the tool boxes for scientists and decision makers. Quantification and communication of uncertainty for decision makers and planning was emphasized throughout the meeting; we considered both how to extend our capabilities for certainty, and to fit methods and models for purpose, appreciating the limits of our certainty of results.
Strategies developed for reducing or quantifying uncertainty included hierarchies of model complexity and model purpose, metrics and evaluations for internal consistency, model emulation and deeper understanding of parameters - one of the long term goals emerging from the discussions is to systematically reduce sources of uncertainty within models and give transparency to their construction and tuning. While current generation models lack physical processes needed to resolve all climate feedbacks, many of these issues can be resolved by improvements in computational power. A much different question for mathematicians, statisticians and computational scientists is how to systematically refine our modelling techniques to confine the sources of uncertainty in models; a major component of this study is understanding the purpose of techniques, their limits, and their contribution to uncertainty so that we may accurately quantify our uncertainty.
Aspects of uncertainty for decision makers were also explored, including accessibility and quality of data, the use of conceptual models, integrated assessment models, numerical weather prediction and GCMs, and the use of scientific and climatology information for planning policy. Decision support scientists gave their perspective on the decision making process, providing ways mathematicians can address issues of uncertainty directly in the decision process, phrasing questions fit for the use of decision makers, as well as to develop original mathematical questions which can constrain and quantify uncertainty for decision makers.
Many thanks go to the Lorentz Center and its staff for their support of this highly informative and productive meeting.
Thank you to Colin Grudzien for the summary and picture!
For more recent work in Data Assimilation, see the Joint Data Assimilation Seminar with ICTS-TIFR Focus Group at https://mcrn.hubzero.org/groups/jdas.