Dutch shoreface dynamics at barrier islands; a BFAST sensitivity study

By Joris Bekedam van Eck

Abstract
Climate change induced sea level rise is raising pressure on low-laying coastal areas and (barrier) islands around the world, such as the Wadden islands in the Netherlands. The shoreface is where the marine environment first touches with the terrestrial environment of barrier islands. Understanding of this dynamic area is therefore of great importance in protecting terrestrial hinterlands. In the Netherlands the shoreface often studied by means of separate coastal profiles or coastlines to predict changes in the location of the shoreline. These studies can’t capture all spatiotemporal variability. Therefore, the aim of this study is to use and assess sensitivity of BFAST (Break for Additive Season and Trend) modelling in order to identify significant changes in shoreface dynamics. This model is normally used for vegetation monitoring, however in this study it will monitor shoreface dynamics such as accretion and erosion.

The JARKUS (yearly shoreface profiles) dataset, consisting of a 57 year timeseries was used to create bathymetric datasets and to derive slope gradient datasets for three Wadden islands (Terschelling, Ameland and Schiermonnikoog). A general structure of the shoreface and barrier islands was investigated (the prototype island). The most important prototype island elements in this study were the head and tail of the island. An initial BFAST parametrization was made where: a complete timeseries was used, the seasonality was removed and the default h-parameter of 0.15 was used. The h-parameter defines the minimum segment size over which trend breaks are calculated. After this, a sensitivity study on BFAST analysis has been done on the datasets by using multiple h-parameter settings corresponding to factors influencing the shoreface: h=0.15 (default), h=0.105 (storm-event) and h=0.326 (nodal tidal cycle).

The Magnitude (a) and moment (b) of trend breaks (two BFAST outputs) have been analysed with landscape metrics (LSM) (a) and density plots (b). Only the greatest trend break value found within the BFAST timeseries analysis was used for this study. The Moran’s I index is one of the LSM and is an indicator of spatial autocorrelation. Patch density is another and quantifies how patchy found trend breaks are within the research area. Generally greater trend breaks, often located at head/tails of the islands, were found for lower h-parameters and clear relations occurred for LSM. High h-values led to higher Moran’s I values and lower Patch densities and vice versa. The highest Moran’s I value found was 0.857. Next to this, BFAST results were more pronounced in the bathymetric dataset compared to the slope gradient. Furthermore, clear peaks in moment of change were found using h=0.326, with peaks around 1983 and 2003 for both datasets and all islands.

To conclude, the BFAST model is able to detect significant changes at the shoreface of barrier islands and can detect different patterns for different elements of the prototype island. However, more research should investigate other BFAST models and outputs. Furthermore, usage of datasets with higher measurement frequencies could provide more detail in results.