Understanding Climate Change from Data

By Marijke Unger
10/03/2014 - 12:00am

Earlier this summer, NCAR hosted the Fourth Workshop on Understanding Climate Change from Data, an event organized by CISL’s IMAGe in conjunction with the University of Minnesota, and a result of an NSF-funded program known as Expeditions in Computing, which explores data-driven approaches to understanding climate change.

Among the keynote speakers were NCAR’s director Jim Hurrell, Senior Scientist and National Medal of Science recipient Warren Washington, Clara Deser, Head of the Climate Analysis Section of the Climate and Global Dynamics Division, and Distinguished Senior Scientist Kevin Trenberth. Rounding out the talks, the workshop also included panel discussions and a poster session.

 CISL’s Doug Nychka discusses findings with a workshop participant Erin Towler during the workshop’s poster session.

A major focus of the workshop was to explore computational data science tools that can extract the achievable predictive insights from climate data and capture the complex dependence structures among climate variables. Because climate change and its consequences are among the most significant challenges of our time, it is critical to develop improved assessments of global and regional changes, extreme events, and stresses on the environment. Specifically, questions relating to food security, water resources, biodiversity, and other socio-economic issues have important practical impact on policy and decision-making.

Climate and earth sciences have recently experienced a rapid transformation from a data-poor to a data-rich environment. In particular, climate related observations from remote sensors on satellites and weather radars, or from in situ sensors and sensor networks, as well as outputs of climate or Earth system models from large-scale computational platforms, provide terabytes of temporal, spatial and spatio-temporal data. In addition, the rapid growth of geographical information systems leads to the availability of multi-source data. These massive and information rich datasets offer a huge potential for advancing the science of climate change and impacts.

Imme Ebert-Uphoff, from Colorado State University, presents during the workshop.

This workshop brought together researchers who are advancing computational and data analysis methods necessary for addressing the key challenges in climate change science.  Three topics that use more computationally oriented techniques to tease out important features in climate data that are not obvious using traditional statistical methods are highlighted here:

  • James Faghmous, a recent PhD student from the University of Minnesota, discussed data mining in the context of ocean eddies as observed from space. Mesoscale ocean eddies are dynamic phenomena that move across the global oceans and play a critical role in the transport of heat, momentum, and nutrients. They are also instrumental in balancing the planet's energy budget  – helping to transport heat from the equator to towards the poles. Faghmous described the results from the OpenEddy project, a theory-guided data mining effort to catalog daily ocean eddy activity on a global scale using satellite altimeter data, as well as preliminary results of the impact of ocean eddies on tropical cyclones on global a scale.

  • Colorado State University’s Imme Ebert-Uphoff presented her preliminary findings on causal discovery, i.e. learning about causal relationships in data using machine learning algorithms. By applying causal discovery to climate science, generating new hypotheses about causal relationships between different climate variables, as well as generating "graphs of information flow" around the globe, Ebert-Uphoff seeks to help climate scientists better understand certain dynamical processes of our planet's climate.

  • Dimitris Giannakis, from New York University, explored methods for spatiotemporal decomposition of large-scale data from climate models and observations. This work has applications in the detection and forecasting of climate patterns on seasonal to interannual timescales.

The workshop presentations and panel discussions were geared toward finding new and transformative approaches to understand the potential impact of climate change, using data driven approaches that have been highly successful in other scientific disciplines. Methodologies developed as part of this project will be used to gain actionable insights and to inform policymakers. The workshop also served for collaborative “matchmaking” where climate scientists and data scientists could identify projects that would exploit their common research interests.