Molly Wieringa, National Center for Atmospheric Research (NCAR)

Sea ice as data assimilation sandbox: leveraging physical features and model structure to improve ensemble filtering of geophysical systems

Data assimilation methods are widely wielded in geophysical applications, many of which involve complex dynamical models and a relative paucity of real-world observations. Sea ice, which is an important component of global climate models, highlights many common challenges of assimilating Earth system observations. In addition to being under-observed, sea ice is a non-Gaussian system, particularly when discretized on a model grid. Popular sea-ice models further discretize sea ice into categorized thickness distributions, altering the relationship between sea-ice quantities that can be observed and those that are unobserved but underpin model dynamics and thermodynamics. Each of these discretizations presents challenges for Gaussian ensemble filtering methods that can lead to analysis inaccuracies. We will review recent advances in non-Gaussian ensemble filtering for geophysical applications developed in response to these sea-ice challenges, and present ongoing efforts to ensure consistency when updating both observed and unobserved sea ice variables.

Passcode for this talk is math-geo