Tipping-Points and Robustness of Sea Ice using Koopman Mode Decomposition

Alan B. Cao, Michael Planer, James Hogg, Maria Fonoberova, Lake Bookman, Tyler Macdonald, Jennifer Hemingway, Qinghua Ding, Igor Mezic

Keywords: climate model, tipping point, Koopman Operator, Arcitc Sea Ice, Dynamical Systems

Abstract:

We demonstrate the application of Koopman Operator Theory
(KOT) to model Arctic sea ice concentrations on decadal
timescales and to identify potential climate tipping-points.
Koopman-based models are computationally inexpensive to
train and evaluate compared to traditional climate models,
enabling robustness analyses of long-term climate trends and
sensitivity analyses of the trends to various assumptions and
uncertainties.We identify a potential tipping-point in the Barents
and Kara Sea through Koopman Mode Decomposition
(KMD) and verify that the Koopman-based models are robust
to the uncertainty in the data used to train the model.