Indiana University Indiana University IU

Event: IUNI Workshop: Using Climate Networks to Predict Heat Waves

Can we use climate networks to predict the next big heat wave?

Friday, Apr 1st 2022 at 1:00 PM - 4:00 PM
Zoom / Hazelbaker Hall (Wells Library, IUB)

Instructors: Dr. Ben Kravitz (Assistant Professor, Earth and Atmospheric Sciences, IUB), Dr. Travis O'Brien (Assistant Professor, Earth and Atmospheric Sciences, IUB), Dr. Didier Vega-Oliveiros (postdoc, CNetS, IUB), Dr. Filipi Nascimento Silva (research scientist, IUNI)

Earth’s climate is one of the most complicated, interrelated systems one can study. Because signals can propagate spatially or temporally throughout this system, climate networks are becoming increasingly popular methods for understanding the Earth, especially for teleconnections – responses in the climate system to things that happen far away. A workshop that leverages IU’s expertise in network science and climate science could catalyze advances in using networks as predictive tools for extreme weather and climate events. We propose a workshop that brings together faculty from the geosciences and network sciences to discuss the potential for climate networks to better understand and predict extreme events. As an initial topic, we will focus on heat waves: the temporal clustering of heatwaves was unusual this summer, and this behavior is poorly understood.

Network representations of the climate system have been used for diagnosing global climate effects and teleconnections [Boers 2019, Falasca 2019], wild-fire events [Ferreira 2020], anomalies in annual hurricanes events [Elsner 2009], seismic events [Ferreira 2018], El Niño precursors [Sonone 2021], continental moisture recycling process [Zemp 2014], and others. Network sciences bring a powerful toolbox for representing and characterizing the complex relationships between climate phenomena. For example, networks can provide a topological characterization of the complex global effects of the El Niño Southern Oscillation and other climate systems [Falasca 2019, Donner 2017]. 

Existing applications of network science to climate involve treating different geographical locations as nodes and using correlations between the locations as weights on the edges of a graph. Pearson-based correlations are the standard functions employed for assessing the similarity between time-series [Bialonski 2011, Donner 2017, Meng 2018, Falasca 2019]. Such approaches have thus far shown limited utility in revealing new understanding of the climate system and have essentially no predictive power. This is arguably because these existing methods only harness the power of network science in limited ways. More robust measures, like causality functions [Silva 2021], show promise but may be limited by similar issues. By bringing together climate scientists and network scientists together for a deeper exploration of the potential of climate networks, an IUNI workshop could yield substantial advances in the use of networks for predictive understanding of extreme weather and climate events.

It is not clear why extremes happen in clusters in some years but not others, and it is an open question whether that might change in a warmer climate. Toward the goal of advancing predictive understanding of climate extremes with network science, topics will center on the fundamentals of how networks might be applied to gain predictive climate understanding. Specific points of discussion could include

  • variable selection: use of temperature, precipitation, or potential vorticity
  • data availability: spatiotemporal data resolution and advantages/disadvantages of different synthetic datasets
  • network structure: how to calculate edges and weights, such as using correlation; whether network representations other than assigning nodes to locations could be appropriate (examples include points in a wave spectrum with edges representing scale interactions or nodes representing different known aggregate dynamical processes that interact)
  • network properties: for example, dynamically evolving networks whereby different connections can reveal propagation of signals through the system
  • aggregation: how to extract signals from noisy data

Prerequisites: familiarity with either climate science or network science. Students and postdoctoral research associates are encouraged to attend.

This workshop will be held on-line, with a limited number of seats available in-person. Both attendance options require registration. Click here to register.

Return to the workshop series home page.