Engage mathematicians and computer scientists through education

Climate Informatics Workshop principals
During the Climate Informatics Workshop poster session, Imme Ebert-Uphoff (center) talks with Xaiowei Jia (right) and Vipin Kumar (left). Jia and Kumar are computer scientists from the University of Minnesota who use remotely sensed image data to map vegetation changes, surface water, and other features of human activity that are useful for climate studies.
Philippe Naveau
Philippe Naveau (Laboratoire des Sciences du Climat et l'Environnement, Centre Nationnal de la Recherche Scientifique) teaches part of a course at NCAR on statistical methods for extremes.

Numerous workshops, seminars, and summer schools have brought a steady stream of mathematical and computer scientists to NCAR – often for the first time – and these have led to collaborations, influence on graduate student research, and a broader awareness of research that can be applied to the geosciences.

An important part of CISL's educational portfolio is the Theme-of-the-Year (TOY), a year-long focus on an aspect of applied mathematics for the geosciences designed to advance research and education between the mathematics and the geosciences communities. Typically the TOY sponsors a series of workshops or schools along with a visitor program that coordinates with NCAR science groups and partners with other mathematics institutes. CISL has supported 10 separate TOYs with the most recent being Extremes in Climate Sciences: a Statistical, Dynamical, and Machine Learning Inquiry.

The theme for FY2016 focuses on extremes in climate sciences, and is being organized by visiting scientist Philippe Naveau, a leading expert in the field. Extreme value theory has long been an active area of mathematical statistics, but it is only recently that approaches have been devised to handle the practical problems for large climate data sets. One goal of this TOY is encourage a process of technology transfer between the data science and the work at NCAR. Besides a focused workshop on the statistics of extremes held in May 2016, this TOY also ran an informal seminar series during the year and also partnered in two other international meetings: International Detection and Attribution Group, NCAR, February 2016, and Uncertainty Modeling in the Analysis of Weather, Climate, and Hydrological Extremes, Banff International Research Station, June 2016.

Emerging from the TOY were new collaborations and the potential for more deliberate interaction between researchers at Centre National de Recherche Scientifique in France and NCAR. From a scientific point of view, what has emerged from the TOY is a plan for data analysis that can seamlessly move from characterizing the center of a distribution to that of the tails. In this way extremes can be incorporated into a single complete analysis of the climatological distribution.

The International Workshops on Climate Informatics are held annually to build collaborations between climate scientists and researchers from statistics, machine learning, and data mining. This year marked the sixth annual event (NCAR has hosted and cosponsored all except for the first workshop). The purpose of this workshop series is to build interdisciplinary partnerships between these researchers.

The format for this year focused on several plenary talks on statistical methods for large spatial data, the role of risk, and insurance in climate change, paleoclimate, and observational data for the polar regions. Interspersed were micro-presentations of selected posters and much discussion. The 2016 workshop hosted more than 80 participants from more than 50 institutions and corporations.

The day preceding the workshop featured an all-day hackathon where the researchers collaborated on solving a big-data problem. Although such events are common in computer science forums, they are still new to NCAR and much of the geosciences. For this hackathon, the students were tasked to predict the monthly sea ice extent based on historical atmospheric variables that were derived from one of NCAR’s models. The students collectively worked toward their predictions by using the platform RAMP (Rapid Analytics and Model Prototyping), which allows modelers to share pieces of their code with others. Balazs Kégl, a hackathon organizer, said the hackathon provided valuable experience for the students because “the [hackathon] pipeline is what you do in reality.”

IMAGe’s Theme of the Year activities are supported by NSF Core funding. Climate Informatics is supported by NSF Core funds with contributions from Nvidia corporation and the following NSF awards to external institutions: 1345052 (George Washington University, Monteleoni) and 1106862 (University of Washington, Guttorp).

Climate Informatics hike
A short walk by participants during the Climate Informatics workshop. Among the many activities for informal interaction, discussion, and research networking is a popular walk from the Mesa Lab during the second afternoon.