Journal Article

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

Krieger, Daniel, Baehr, Johanna, Borchert, Leonard F., Weisse, Ralf

2025

Overview

The German Bight coastline is regularly affected by storm surges driven by extratropical cyclones. Decadal‐scale predictions of local surges would foster coastal protection and decision making in affected areas. We examine the prediction skill of the Max‐Planck‐Institute Earth System Model (MPI‐ESM) decadal prediction system for three different storm surge metrics at Cuxhaven (Germany), Esbjerg (Denmark), and Delfzijl (The Netherlands). To avoid dynamical downscaling from the coarse model output to local surge heights, we use machine learning and train a neural network on observed surge heights and reanalyzed fields of mean sea‐level pressure (MSLP). We apply this network to MSLP output of our prediction system to generate decadal predictions of surge heights. The prediction system falls short of generating skillful predictions for high water event durations and individual lead years in general, but windows for more skillful predictions arise for deterministic predictions at longer multi‐year lead times. Plain Language Summary In the German Bight, storms regularly push water masses toward the coast, causing so‐called storm surges at the coastline. Forecasting whether these storm surges will occur more or less often over the next couple of years would be very helpful for the long‐term planning of coastal protection. However, current climate models, which are used to predict the global climate over the next years, are not able to produce forecasts on such a small regional level. To overcome this problem, we use machine learning to build a translator which can translate the large‐scale behavior of the atmosphere over Europe to the regional water level at Cuxhaven, Esbjerg, and Delfzijl, three locations on the German Bight coastline. We use this translator to quickly turn forecasts from a global climate model into forecasts of storm surges at the German Bight coastline. We then analyze how good these storm surge forecasts are for different years in the future. We find that the forecasts work better when we try to predict the mean climate over multiple years than when we make a forecast for one specific year. Key Points We train a neural network to translate large‐scale mean sea‐level pressure to surge heights at three locations in the German Bight We apply the neural network to mean sea‐level pressure output of a decadal prediction system The generated decadal predictions of surges show considerable prediction skill for deterministic and some for probabilistic predictions

• Publisher: John Wiley & Sons, Inc,Wiley
• Subject: Climate / Climate models / Climate prediction / Coastal protection / Coastal storms / Coastal zone management / Coasts / Cyclones / Decision making / Environmental protection / Extratropical cyclones / Global climate / Global climate models / Learning algorithms / Machine learning / Neural networks / Storm surge forecasting / Storm surge prediction / Storm surges / Storms / Tidal waves / Translators / Water levels / Water masses