University of Cambridge > Talks.cam > CEDSG-AI4ER > Probabilistic Machine Learning for East US Storm Surges

Probabilistic Machine Learning for East US Storm Surges

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Jonathan Rosser.

Predicting storm surge flood frequency is challenging: storm surges are short lived, infrequent, and require high model resolution. Inspired by recent work [1], I propose two metrics for an element of the coast line: it’s convexity and bathymetric gradient. The responsiveness of the sea level to a wind state can be found as a function of these through kriging. This simplification can allow better sampling with smaller periods of data. To enhance interpretability I use warped Gaussian processes [2]. The model learnt generalises between different years of a 1/12 degree model [3]. The results are still tentative and I would appreciate robust feedback.

[1] https://doi.org/10.1175/JCLI-D-19-0551.1 [2] https://papers.nips.cc/paper/2481-warped-gaussian-processes.pdf [3] (ORCA12, hourly output, 2004/5, CORE2 reanalysis product forcing, EN4 initialisation)

I have been collaborating with Dan Jones, Laure Zanna, Pierre Mathiot, and Rory Bingham (respectively BAS , NYU, Met Office, & Bristol).

This talk is part of the CEDSG-AI4ER series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2020 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity