The November talk of the
West Midlands Local group took place at Coventry University. The speaker was Dr. Alireza Daneshkhah, who is an associate of the Coventry research Centre for Computational Science and Mathematical Modelling.
Dr. Daneshkhah gave an enlightening talk on modelling resilience of land mass against flood and erosion impacts.
The flood example concerned the river Wharfe at Tadcaster in Yorkshire, which is prone to burst its banks and flood the town if there is excessive rainfall; the lowest road crossing-point on the river Wharfe is at Tadcaster, which makes the town particularly vulnerable to flooding.
The erosion impact example was a bit further from home and concerned coastal erosion at the northern end of the Bay of Bengal, which impacts the coast of Bangladesh and the effectiveness of Mangrove forests to prevent erosion.
In both cases it is useful to model the effects of the climate, on flooding and erosion, so that counter measures can usefully evaluated. Dr. Daneshkhah illustrated the use of Bayesian Gaussian Processes to model these phenomena which proved to be computationally faster than Neural Networks (either the Convoluted Neural Network, or the Physics Informed Neural Network), by a factor of about 80:1. The idea is to replace the underlying physics-based models (based on the Navier Stokes equations and the Gauckler–Manning formula), with a surrogate model derived from the Gaussian process (GP) model.
Dr. Daneshkhah, finished by discussing the proposed countermeasure for each case. For Tadcaster, a flood defence barrier similar to the one already in place at Bewdley on the River Severn, and for Bangladesh the preservation of the mangrove forest, which acts as a baffle, the effectiveness for which was nicely illustrated with a computer animation based on the GP modelling. The flood defense at Tadcaster is planned for 2026, once its exact positioning has been determined. Further details can be found in Donnelly et al. (2022).
Technical details of the statistical modelling of both of these case studies can found in the references below.
This report prepared by Tim Davis (chair, West Midlands Local Group)
References
Donnelly, J., Abolfathi, S., Pearson, J., Chatrabgoun, O., & Daneshkhah, A. (2022). “Gaussian process emulation of spatio-temporal outputs of a 2D inland flood model”.
Water Research,
225, 119100.
Fanous, M., Daneshkhah, A. Eden, J. M, Palade, P., Remesan, R. (2022). “Hydromorphodynamic modelling of mangroves Using finite element discontinuous Galerkin method”. A working paper, submitted to Coastal Engineering.