Probabilistic modelling of tidal inlets
Sediment fate estimation in the coastal system using Markov chains
J.F. Laan (TU Delft - Civil Engineering & Geosciences)
S.G. Pearson – Mentor (TU Delft - Coastal Engineering)
José A.Á. Antolinez – Graduation committee member (TU Delft - Coastal Engineering)
Oswaldo Morales Napoles – Graduation committee member (TU Delft - Hydraulic Structures and Flood Risk)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Under tidal currents and wave action, sediment particles show complex transport patterns for coastal systems like tidal inlets. To model the sediment transport in such coastal systems, models have been derived that can predict that transport using Eulerian (grid-based) or Lagrangian (particle-tracking) methods. The sediment transport model SedTRAILS uses a Lagrangian method to predict the pathways that sediment particles will follow. As a process-based model SedTRAILS is limited in how fast it can compute results. Here we show a methodology to make a faster probabilistic metamodel from the process-based model SedTRAILS by creating an Auto-regressive Hidden Markov Model (ARHMM). Creating a probabilistic model of the sediment transport in a coastal system using Markov chains is
a novel approach in that it builds further on the connectivity approach of SedTRAILS, which is still quite a new development in the field of sediment transport at present. For different versions of the model, correspondence with physical properties of the system were found, that the development to estimate sediment pathways accurately look promising. The method of describing sediment transport pathways of a coastal system as an ARHMM is a form of probabilistic Machine Learning allowing for faster computation times by describing the patterns the system can make. This is in contrast to a process-based model that has to compute all of the underlying processes. The autoregressive component in this approach is expected to be essential in describing any type of dynamical system that consists of trackable pathways. Taking into account the position of a particle at the current time step highly limits the position a particle can have at the next time step in favor of accurately estimating sediment pathways. The modelling approach discussed can find uses in engineering applications where particle pathways are of importance like the dispersal of nourishments, but also other dispersion events like dispersed goods from when a vessel loses a container, as long as there is a process-based model available to create the necessary Lagrangian data.