Predicting coastal erosion trends using non-stationary statistics and process-based models

Storms and water levels are subject to seasonal variations but may also have decadal or longer trends that need to be included when estimating risks in the coastal zone. We propose a non-stationary multivariate generalised extreme value model for wave height, wave period, storm duration and water levels that is constructed using Archimedean copulas. The statistical model was applied to a South African case study to test the impacts of decadal trends on beach erosion. Erosion was estimated using three process-based models — SBEACH, XBEACH, and the Time Convolution model. The XBEACH model provided the best calibration results and was used to simulate potential future long-term trends in beach erosion. Based on the simulated erosion results of 5 beach profiles for storms with 25, 50 and 100 year return periods, it is estimated that the erosion rate could increase by 0.20%/year/storm and should therefore be a significant factor in long-term planning.

► Trends in sea storms and water levels can cause increased coastal erosion.
► Multivariate statistics are coupled with process-based models for erosion prediction.
► The potential future impacts are quantified for a case study in South Africa.
► Increased erosion is found to be potentially significant for long-term planning.