Extreme storm surge classification for risk assessment of coastal infrastructure

Research output: Chapter in Book/Report/Conference proceedingConference contribution book

Abstract

Alerting responsible authorities of the occurrence of the most extreme surge hazards is integral to the safe operation of coastal transport infrastructures. As such, it is critical to design a system to identify these events to responsible authorities giving them enough time to take meaningful action to mitigate the potential impacts of these hazards. Machine learning and artificial intelligence are powerful tools for analysing data and constructing predictive models. However, for real application data is not always available in the desired quantity or quality to train these tools to an operational standard. In this paper, procedures for preprocessing data and improving the accuracy of neural network designs and prediction are presented. The proposed approach is applied for the classification of extreme storm surge heights at Millport in Southwest Scotland. Several classifier architectures are considered and tested and the best among them tested against real unseen data to evaluate its serviceability as part of the forecasting system. A multilayer perceptron network with 125 nodes in the first hidden layer and 50 nodes in the second hidden layer performed best with recall and precision values of 0.93 and 0.93.

Original languageEnglish
Title of host publicationProceedings of the 32nd European Safety and Reliability Conference, ESREL 2022 - Understanding and Managing Risk and Reliability for a Sustainable Future
EditorsMaria Chiara Leva, Edoardo Patelli, Luca Podofillini, Simon Wilson
Place of PublicationSingapore
Pages2945-2952
Number of pages8
DOIs
Publication statusPublished - 28 Aug 2022
Event32nd European Safety and Reliability Conference, ESREL 2022 - Dublin, Ireland
Duration: 28 Aug 20221 Sept 2022

Publication series

NameProceedings of the 32nd European Safety and Reliability Conference, ESREL 2022 - Understanding and Managing Risk and Reliability for a Sustainable Future

Conference

Conference32nd European Safety and Reliability Conference, ESREL 2022
Country/TerritoryIreland
CityDublin
Period28/08/221/09/22

Keywords

  • classification
  • forecasting
  • machine learning
  • preprocessing
  • surge hazard

Fingerprint

Dive into the research topics of 'Extreme storm surge classification for risk assessment of coastal infrastructure'. Together they form a unique fingerprint.

Cite this