GR
G.P. Reverda
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
2 records found
1
Modelling Beach Profile Reshaping Due To Sea-Level Rise
Simulating Laboratory Experiments including Sea-Level Rise with a Physics-Resolving Model
Master thesis
(2026)
-
G.P. Reverda, José A. Á. Antolínez, A.J.H.M. Reniers, P.B. Bayle, Maurizio D'Anna, Francesca Ribas, Giovanni Coco
Mean sea-level rise (MSLR) fundamentally reshapes coastal profiles, threatening densely populated coastlines. While physics-resolving numerical models like XBeach offer a deterministic alternative to analytical estimators like the Bruun Rule, their application to long-term MSLR reshaping is hindered by high computational costs. This study systematically assesses the capacity of XBeach (Surfbeat mode) to predict cross-shore profile reshaping under MSLR by simulating reduced-scale laboratory flume experiments.
To overcome the computational bottlenecks of traditional brute-force calibration, a highly efficient uncertainty quantification framework driven by an active learning algorithm based on Treed Gaussian Processes was developed. The framework successfully isolated non-linear parameter interdependencies, converged on an optimal parameter set, and demonstrated that anchoring non-target parameters at their optimal values constrains the predictive standard deviation within narrow, physically interpretable bounds ($1.0$ to $2.0 \text{ m}^2$). This tightly bounded variance confirms high surrogate confidence across the explored parameter space.
Blind validation tests demonstrate that the calibrated model predicts the final post-MSLR shoreline position with high accuracy (absolute error = 0.068 m) and reproduces the adjustment time to dynamic equilibrium within approximately 4\% of the laboratory observations. However, the model consistently fails to resolve complex, localised sub-aqueous multi-bar morphologies. Overall, the active learning framework effectively identifies the structural limitations of XBeach while demonstrating that the model reliably captures the macro-scale feedbacks governing long-term shoreline evolution, supporting its application to shoreline management and climate adaptation planning. ...
To overcome the computational bottlenecks of traditional brute-force calibration, a highly efficient uncertainty quantification framework driven by an active learning algorithm based on Treed Gaussian Processes was developed. The framework successfully isolated non-linear parameter interdependencies, converged on an optimal parameter set, and demonstrated that anchoring non-target parameters at their optimal values constrains the predictive standard deviation within narrow, physically interpretable bounds ($1.0$ to $2.0 \text{ m}^2$). This tightly bounded variance confirms high surrogate confidence across the explored parameter space.
Blind validation tests demonstrate that the calibrated model predicts the final post-MSLR shoreline position with high accuracy (absolute error = 0.068 m) and reproduces the adjustment time to dynamic equilibrium within approximately 4\% of the laboratory observations. However, the model consistently fails to resolve complex, localised sub-aqueous multi-bar morphologies. Overall, the active learning framework effectively identifies the structural limitations of XBeach while demonstrating that the model reliably captures the macro-scale feedbacks governing long-term shoreline evolution, supporting its application to shoreline management and climate adaptation planning. ...
Mean sea-level rise (MSLR) fundamentally reshapes coastal profiles, threatening densely populated coastlines. While physics-resolving numerical models like XBeach offer a deterministic alternative to analytical estimators like the Bruun Rule, their application to long-term MSLR reshaping is hindered by high computational costs. This study systematically assesses the capacity of XBeach (Surfbeat mode) to predict cross-shore profile reshaping under MSLR by simulating reduced-scale laboratory flume experiments.
To overcome the computational bottlenecks of traditional brute-force calibration, a highly efficient uncertainty quantification framework driven by an active learning algorithm based on Treed Gaussian Processes was developed. The framework successfully isolated non-linear parameter interdependencies, converged on an optimal parameter set, and demonstrated that anchoring non-target parameters at their optimal values constrains the predictive standard deviation within narrow, physically interpretable bounds ($1.0$ to $2.0 \text{ m}^2$). This tightly bounded variance confirms high surrogate confidence across the explored parameter space.
Blind validation tests demonstrate that the calibrated model predicts the final post-MSLR shoreline position with high accuracy (absolute error = 0.068 m) and reproduces the adjustment time to dynamic equilibrium within approximately 4\% of the laboratory observations. However, the model consistently fails to resolve complex, localised sub-aqueous multi-bar morphologies. Overall, the active learning framework effectively identifies the structural limitations of XBeach while demonstrating that the model reliably captures the macro-scale feedbacks governing long-term shoreline evolution, supporting its application to shoreline management and climate adaptation planning.
To overcome the computational bottlenecks of traditional brute-force calibration, a highly efficient uncertainty quantification framework driven by an active learning algorithm based on Treed Gaussian Processes was developed. The framework successfully isolated non-linear parameter interdependencies, converged on an optimal parameter set, and demonstrated that anchoring non-target parameters at their optimal values constrains the predictive standard deviation within narrow, physically interpretable bounds ($1.0$ to $2.0 \text{ m}^2$). This tightly bounded variance confirms high surrogate confidence across the explored parameter space.
Blind validation tests demonstrate that the calibrated model predicts the final post-MSLR shoreline position with high accuracy (absolute error = 0.068 m) and reproduces the adjustment time to dynamic equilibrium within approximately 4\% of the laboratory observations. However, the model consistently fails to resolve complex, localised sub-aqueous multi-bar morphologies. Overall, the active learning framework effectively identifies the structural limitations of XBeach while demonstrating that the model reliably captures the macro-scale feedbacks governing long-term shoreline evolution, supporting its application to shoreline management and climate adaptation planning.
Student report
(2025)
-
J.L. den Dekker, V.M. van Dijk, G.P. Reverda, N. Holt, J.A. Arriaga Garcia, K.B.J. Dunne, R.J. van der Ent, Rudi Darson
This report documents the impact of floods in the Tapanahony region as a step to start developing the region in a sustainable manner. The Tapanahony region of the Okanisi people harbours descendants of enslaved people that freed themselves during the 1700s. Currently the development of the region is stagnating and the Okanisi experience problems when trying to build a living for themselves and their families. By conducting interviews in the 28 villages along the river an impression is obtained of the role that the river plays in this area. These results are then combined with a flood vulnerability model for which depth and velocity measurements of the river have been conducted. This study shows how the lives of the inhabitants are intertwined with the river. It also stresses the severity of the issues at play here. A lack of clean drinking water in the dry season in combination with heavy floods in the wet season make it difficult for people to progress and drives them to turn to goldmining as a source of income. This report also shows that there are solutions to the situation, as long as there is strong collaboration between developers, researchers and especially the villages themselves.
...
This report documents the impact of floods in the Tapanahony region as a step to start developing the region in a sustainable manner. The Tapanahony region of the Okanisi people harbours descendants of enslaved people that freed themselves during the 1700s. Currently the development of the region is stagnating and the Okanisi experience problems when trying to build a living for themselves and their families. By conducting interviews in the 28 villages along the river an impression is obtained of the role that the river plays in this area. These results are then combined with a flood vulnerability model for which depth and velocity measurements of the river have been conducted. This study shows how the lives of the inhabitants are intertwined with the river. It also stresses the severity of the issues at play here. A lack of clean drinking water in the dry season in combination with heavy floods in the wet season make it difficult for people to progress and drives them to turn to goldmining as a source of income. This report also shows that there are solutions to the situation, as long as there is strong collaboration between developers, researchers and especially the villages themselves.