Machine Learning‐Driven Skillful Decadal Predictions of German Bight Storm Surges

Geophysical Research Letters, Journal Year: 2025, Volume and Issue: 52(4)

Published: Feb. 17, 2025

Abstract The German Bight coastline is regularly affected by storm surges driven extratropical cyclones. Decadal‐scale predictions of local would foster coastal protection and decision making in areas. We examine the prediction skill Max‐Planck‐Institute Earth System Model (MPI‐ESM) decadal system for three different surge metrics at Cuxhaven (Germany), Esbjerg (Denmark), Delfzijl (The Netherlands). To avoid dynamical downscaling from coarse model output to heights, we use machine learning train a neural network on observed heights reanalyzed fields mean sea‐level pressure (MSLP). apply this MSLP our generate heights. falls short generating skillful high water event durations individual lead years general, but windows more arise deterministic longer multi‐year times.

Language: Английский

---

Skillful multiyear to decadal predictions of sea level in the North Atlantic Ocean and U.S. East Coast

Communications Earth & Environment, Journal Year: 2023, Volume and Issue: 4(1)

Published: Nov. 17, 2023

Abstract Long-term sea-level rise and multiyear to decadal sea level variations pose substantial risks for flooding erosion in coastal communities. Here we use observations climate model predictions show that along the U.S. East Coast are skillfully predictable 3 10 years advance. The most component of is a basin scale upward trend, decade advance primarily response increasing greenhouse gases. Significant additional predictability comes from multidecadal Atlantic Meridional Overturning Circulation (AMOC). While perfect simulations AMOC-related 5-7 years, biases initialization uncertainties reduce realized predictive skill 3-5 depending on location. Overall, gas warming AMOC lead prediction Coast. Such could have significant societal benefit planning adaptation.

Language: Английский

---

Seasonal Predictability of Bottom Temperatures Along the North American West Coast

Journal of Geophysical Research Oceans, Journal Year: 2024, Volume and Issue: 129(9)

Published: Sept. 1, 2024

Abstract Bottom Temperature anomalies (BTA) along the North American West Coast strongly influence benthic and demersal marine species. However, to date seasonal BTA forecast efforts have been limited sources of predictability largely undiagnosed. Here, an empirical model called a Linear Inverse Model (LIM), constructed from high‐resolution ocean reanalysis, is developed predict BTAs diagnose predictive skill. The LIM considerably more skillful than damped persistence, particularly in winter, with anomaly correlation (AC) skill values 0.6 at 6‐month lead. Analysis LIM's dynamics shows that elevated linked developing El Niño‐Southern Oscillation (ENSO) events, driving predicted responses whose peaks occur longer leads increasing latitude. Weaker ENSO‐related signals northern coastal region still yield high because noise there also weaker. Likewise, signal‐to‐noise ratio highest for bathymetry depths ∼50–150 m, maximizing there. Together, these components lead “forecasts opportunity” when anticipates especially prediction For top 20% events identified by as forecasts opportunity, hindcasts AC averaging 0.7, while remaining 80% mean only 0.4, suggesting can leverage produce forecasts.

Language: Английский

---

Machine learning-driven skillful decadal predictions of the German Bight storm surge climate

Authorea (Authorea), Journal Year: 2024, Volume and Issue: unknown

Published: July 31, 2024

The German Bight coastline is regularly affected by storm surges driven extratropical cyclones. Decadal-scale predictions of the local surge climate would foster coastal protection and decision making in areas. We therefore examine prediction skill Max-Planck-Institute Earth System Model (MPI-ESM) decadal system for three metrics at Cuxhaven, Germany: annual upper percentiles heights durations number surges. To avoid dynamical downscaling from coarse model output to heights, we use machine learning train a neural network on observed reanalyzed fields mean sea-level pressure (MSLP). apply this MSLP our generate Cuxhaven. find that falls short generating skillful individual lead years, but can predict certain multi-year averages skillfully.

Language: Английский

---

Projecting 21st century global and regional sea-level changes

Elsevier eBooks, Journal Year: 2023, Volume and Issue: unknown, P. 259 - 271

Published: June 25, 2023

Language: Английский

---