Abstract
Accurately predicting total sea-level including tides and storm surges is key to protecting and managing our coastal environment. However, dynamically forecasting sea level extremes is computationally expensive. Here a novel alternative based on ensembles of artificial neural networks independently trained at over 600 tide gauges around the world, is used to predict the total sea-level based on tidal harmonics and atmospheric conditions at each site. The results show globally-consistent high skill of the neural networks (NNs) to capture the sea variability at gauges around the globe. While the main atmosphere-driven dynamics can be captured with multivariate linear regressions, atmospheric-driven intensification, tide-surge and tide-tide non-linearities in complex coastal environments are only predicted with the NNs. In addition, the non-linear NN approach provides a simple and consistent framework to assess the uncertainty through a probabilistic forecast. These new and cheap methods are relatively easy to setup and could be a valuable tool combined with more expensive dynamical model in order to improve local resilience.
Highlights
Predicting accurately the sea water level variability from short to large time scales is of great importance for coastal communities
Global skills of the Neural Networks (NNs) The Continuous Ranked Probability Score (CRPS) is computed for the observed non-tidal residual to provide a baseline metric for the signal not captured by the astronomical harmonic analysis
For periods longer than 1 day, the energy is slightly under-estimated by both the regression and the NN. This highlights skills into predicting sea water anomaly and extreme events using a simple NN forced by a small range of atmospheric and wave data
Summary
Predicting accurately the sea water level variability from short to large time scales is of great importance for coastal communities. Deterministic numerical models have proven to be powerful tools for predicting sea variability. In particular they are effective for simulating storm surge propagation and impacts, and facilitate understanding of the complex physical processes associated with the storms [5–11]. They are relatively expensive and complex to set up and run operationally, with associated additional computation costs if ensemble forecasts are required for analysis of risk or variability
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
