Machine learning-based storm surge predictions in the German Bight
5-1.214 - Linke Seite – Großer, unterteilbarer Konferenzraum
GEOMAR - Standort Ostufer / GEOMAR - East Shore
Speaker: Dr. Daniel Krieger, Hamburg University
Hybrid - Zoom: https://geomar-de.zoom.us/j/84289388604?pwd=dGlpeTBUd1Nxem5Ec3dRYXh4NFpOUT09
Abstract
Decadal predictions of the storm surge climate in the German Bight are important for coastal protection and management. While storms, the main drivers of surges, have been shown to be skillfully predictable on a decadal scale with a dynamical large-ensemble prediction system, the spatial resolution is too coarse to generate predictions of local surge heights. We therefore use machine learning to statistically downscale the output of a decadal prediction system based on MPI-ESM to generate decadal predictions of surge metrics at three locations of the German Bight coastline. We show that these predictions can be skillful, especially for multi-year averages and simple surge metrics.