A multi-model ensemble approach to coastal storm erosion prediction. (April 2022)

• Record Type: Journal Article
• Authors: Simmons, Joshua A.; Splinter, Kristen D.
• Is Part Of: Environmental modelling & software. Volume 150(2022)
• Publication Date: 2022-04
• Subjects: Neural network -- SBeach -- XBeach -- Dune erosion -- Shoreline retreat; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2022.105356
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: The accurate prediction of storm-driven coastal erosion along sandy coastlines is fundamental to addressing coastal hazards now and into the future. Here, four storm erosion models (an empirical model, the numerical models SBEACH and XBeach, and a machine learning model) were individually trained and tested on a 39-year storm erosion dataset to examine skill and error distributions. Four weighted average model ensemble approaches were also tested. The machine learning method showed the overall best skill for an individual model, followed by SBEACH, the empirical model, and XBeach. A weighted ensemble combined the models in such a way as to improve prediction (over any single model) for the largest events while maintaining comparable skill to the machine learning model during smaller events as well. These results indicate that a weighted multi-model ensemble approach can provide overall improved accuracy and reliability over a wide range of storm conditions compared to individual models. Highlights: A multi-model weighted ensemble approach outperforms 4 individual coastal erosion models over a 40-year storm data set. A machine-learning (neural network) model is the best performing individual model over the full range of storms observed. The empirical model is the best performing model on the more extreme events. The neural network and SBeach model errors are distributed around 0, whereas XBeach and the empirical model tend to over-predict erosion. Adaptive model averaging shows promise for future work where more storm data is available.