| Title: |
Quantifying calibration data requirements for coastal erosion models: how many storms is enough? |
| Authors: |
Simmons, J; Harley, MD; Turner, IL; Splinter, KD |
| Source: |
Australasian Coasts and Ports Conference, Cairns, Australia, 2017-06-21 - 2017-06-23 |
| Publication Year: |
2017 |
| Collection: |
UNSW Sydney (The University of New South Wales): UNSWorks |
| Description: |
Numerical models underpin a wide range of coastal management applications and are increasingly being relied upon to provide precise and detailed predictions as to the timing, location and extent of morphological change at the coast. Faced with the task of selecting appropriate models for these applications, numerical modellers must carefully consider the overall performance of common beach erosion models, and in particular the calibration data that is required to achieve this. This paper explores the amount of calibration data required to achieve optimum performance for two of the most commonly used storm erosion models (the process-based XBeach and the semi-empirical SBEACH), and examines the variability in optimised parameter values within a single coastal embayment. Generalized Likelihood Uncertainty Estimation (GLUE) is used as a rigorous calibration technique facilitating this analysis, applied to a comprehensive beach survey dataset consisting of immediately pre- and post- storm measurements collected for four individual storm events recorded at the 3.6 km long Narrabeen-Collaroy Beach in southeast Australia. SBEACH is found to be the more skilful model when default parameter values are used, but XBeach shows comparatively better model performance when calibrated to local storm erosion measurements. Importantly, SBEACH performance did not improve when additional storm data was used for calibration, whereas XBeach performance is shown to improve markedly when calibrated to data from two storms. The benefit of including a third storm dataset for model calibration with XBeach however is marginal. Using the Narrabeen-Collaroy embayment as a case study, a distinct difference in the optimum parameter values is found at different locations along the beach, further emphasising that modellers must exercise caution when applying calibrated parameter values that were not specifically derived at the particular location of interest, even when these were obtained from extensive datasets at adjacent locations. |
| Document Type: |
conference object; report |
| Language: |
unknown |
| Relation: |
http://purl.org/au-research/grants/arc/DP150101339; https://hdl.handle.net/1959.4/unsworks_46861 |
| Availability: |
https://hdl.handle.net/1959.4/unsworks_46861 |
| Rights: |
metadata only access ; http://purl.org/coar/access_right/c_14cb ; CC-BY-NC-ND ; https://creativecommons.org/licenses/by-nc-nd/4.0/ |
| Accession Number: |
edsbas.392D50A8 |
| Database: |
BASE |