Introduction

Nearshore sandbars (alongshore ridges of sand in 2 - 10 m water depth typical of micro- to macrotidal coasts) serve as a natural protection for beaches by causing waves to break away from the shoreline. Due to the interaction between the sediment and the incoming waves, sandbars change in shape and location over time. Predicting sandbar behavior with the help of mathematical models is difficult because such models are sensitive to initial conditions and model inadequacies. In this project I aim to quantify the predictability of sandbar behavior and to determine its relevant sources.

Cross-section animation of the nearshore zone at Hasaki, Japan. Sand and water are shown in sand-color and blue, and the black asterisk indicates the sandbar-crest location. Wave heights are to scale. The animation speed is six days per second.

Methods

I study the predictability of cross-shore sandbar behavior using different modeling techniques on several datasets of observed sandbar locations and the wave climate. The data have temporal and spatial resolutions of one day and several meters, and cover the sandbar zone over several years. This allows me to study predictability on all relevant scales. I developed and used several analysis and modeling techniques that derive the structure underlying the observations, such as neural networks, a multivariate nonlinearity detection algorithm and a data-based stability analysis method. Also, I compare the results of these data-driven methods to the outcomes of numerical models.

Top-view animation of sandbars at the Gold Coast, Australia. White bands of breaking waves reveal sandbar locations. The alongshore (cross-shore) range is 4 (1) kilometers, and the animation speed is two days per second.

Results

The predictability of sandbar behavior depends on the scale on which a model represents the nearshore zone. Models based on collections of small- and short-scale (meters, hours) physical processes have great difficulty predicting the long-term (months to years) cross-shore evolution of sandbars, especially at sites with long-term trends in sandbar locations. The difficulty to predict long-term sandbar behavior with these models is mainly due to their iterative update scheme, causing exponential accumulation of errors over time. Instead of capturing all the small- and short-scale processes, models can also be specified in terms of the larger-scale features that emerge from the underlying physics. Using a data-based stability analysis method, I show that the small- and short-scale processes drive the nearshore zone toward different stable states. As a result, models for sandbar behavior based on the emerging stable features turn out to be better in predicting long-term sandbar behavior. Predictability is therefore not just an intrinsic property of the nearshore system, but also depends on the scale on which a model is specified.

Links

Gold Coast, Australia study site.
Hasaki Oceanographic Research Station, Japan study site.
Egmond, The Netherlands study site.

Please contact Leo Pape for more info on this project.