Finding the potential height of storm surges in a changing climate using Bayesian optimization
- 1Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Cambridge, United Kingdom (sdat2@cam.ac.uk)
- 2British Antarctic Survey, NERC, UKRI, Cambridge, United Kingdom
- 3Department of Mathematics, Emory University, Atlanta, United States of America
- 4School of Ocean and Earth Science, University of Southampton, Southampton, United Kingdom
Climate change is expected to increase the potential intensity and size of tropical cyclones, with implications for storm surges and other damaging effects associated with them. However, quantifying the maximum possible storm surge and the sensitivity of storm surges to the properties of a tropical cyclone remains computationally expensive and complex.
In this study, we use machine learning to find the upper bounds of storm surges, considering the coastline near New Orleans as a case study. To do this, we make use of the well-established potential intensity (Emanuel, 1986) and the recently introduced potential size (Wang et al., 2022) upper bounds for tropical cyclones. These encapsulate the physical constraints tropical cyclones will encounter in a changing climate. We use the max-value entropy search acquisition function from Bayesian optimization (Wang et al., 2017) to efficiently find the largest storm surge at each point along the coast given those constraints. The individual storm surge estimates are produced by forcing a barotropic ocean circulation model ADCIRC (Luettich, 1991) with a set of idealized tropical cyclones, the characteristics of which are determined by the Bayesian optimization procedure. To extrapolate these findings into the future, we replicate our experiment under a high emission CMIP scenario (SSP-585) for the year 2100, using potential intensity and potential size as constraints, evaluating potential differences and implications brought on by changing climate conditions.
Our study provides another way of understanding how climate change can influence storm surges. It aims to forge a pathway to more precise, computationally efficient storm surge predictions in the context of climate change, addressing a pressing issue for coastal regions globally. Our novel approach could easily be transferred to other coastlines around the world, influenced by tropical cyclones.
References:
Emanuel, K.A., 1986. An air-sea interaction theory for tropical cyclones. Part I: Steady-state maintenance. Journal of Atmospheric Sciences, 43(6), pp.585-605.
Wang, D., Lin, Y. and Chavas, D.R., 2022. Tropical cyclone potential size. Journal of the Atmospheric Sciences, 79(11), pp.3001-3025.
Wang, Z. and Jegelka, S., 2017, July. Max-value entropy search for efficient Bayesian optimization. In International Conference on Machine Learning (pp. 3627-3635). PMLR.
Luettich, R.A., R.H. Birkhahn and J.J. Westerink, 1991, Application of ADCIRC-2DDI to Masonboro Inlet, North Carolina: A brief numerical modeling study, Contractors Report to the US Army Engineer Waterways Experiment Station, August, 1991
How to cite: Thomas, S., Jones, D., Mayo, T., Taylor, J., Moss, H., Munday, D., and Haigh, I. D.: Finding the potential height of storm surges in a changing climate using Bayesian optimization, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-19806, https://doi.org/10.5194/egusphere-egu24-19806, 2024.
