Shoreline and Bathymetry Approximation in Mesh Generation for Tidal Renewable Simulations

Alexandros Avdis, Christian T. Jacobs, Jon Hill, Matthew D. Piggott, Gerard J. GormanarXiv 2015 · arXiv

Abstract

This paper introduces a novel method for simplifying complex shorelines and underwater topography (bathymetry) using Principal Component Analysis (PCA). By treating geographic features as signals to be decomposed, the researchers can automatically generate high-quality computational meshes that are detailed where it matters for tidal energy simulations and smoothed elsewhere to save on massive computational costs. The approach offers a unified framework for handling both line-based and raster-based geophysical data, ensuring geometric smoothness and numerical stability.

Introduction

In geophysical flow simulations—such as modeling tidal turbines in the English Channel—the geometry of the domain is inherently complex. Coastlines and seabeds exhibit fractal-like properties, yet computational models must balance detail with efficiency. To run these simulations effectively, researchers employ unstructured meshes. High resolution is required near turbines to capture power generation and scour patterns, while coarser representations are preferred in distal areas to minimize computational overhead.

Traditional simplification methods, like the Douglas-Peucker algorithm, often produce jagged geometries that can lead to poor-quality mesh elements and numerical instability. Furthermore, standard filtering techniques for bathymetry, such as Fourier analysis, are often difficult to integrate with shoreline simplification. This paper addresses these challenges by proposing a PCA-based smoothing framework.

Methodology: PCA-Based Simplification

The authors propose using Principal Component Analysis (PCA) as a unified framework for both shoreline and bathymetry simplification.

Shoreline Simplification

The algorithm decomposes a coastline into overlapping segments, or partitions. The $x$ and $y$ coordinates of these segments are treated as observations in a multivariate system. By applying PCA, the geometry is decomposed into "modes." The primary modes capture the most significant "energetic" structures—the general shape of the coast—while higher-order modes represent fine-scale details. Reconstructing the coastline using only dominant modes results in a smooth, simplified version of the original geometry.

PCA-based decomposition of shoreline coordinates into overlapping partitions.

Bathymetry Simplification

A similar approach is applied to raster data. The seabed map is partitioned into sub-grids, and PCA filters out high-frequency "noise" while retaining major underwater features like ridges and troughs. To ensure continuity between simplified segments, a partial overlap technique is used, averaging coordinates at segment boundaries. This signal-processing approach ensures a level of smoothness inherently compatible with mesh-generation software like Gmsh.

Datasets and Evaluation

The method was evaluated using standard geophysical datasets, including the Global, Self-consistent, Hierarchical, High-resolution Shoreline (GSHHG) database and the GEBCO 30" bathymetric chart. Testing focused on complex regions such as the island-heavy Western Norwegian coast and the Orkney Islands, a critical site for tidal energy research.

Key Results and Insights

The PCA-based approach allows the number of modes to function as a "tuning knob" for geometric detail. For instance, using a single mode provides a highly smoothed approximation, while increasing the modes recovers finer details.

Comparison of shoreline simplification levels on the Norwegian coast using different PCA modes and partition sizes.

Multi-Scale Meshing

The researchers successfully generated three tiers of meshes for the UK continental shelf:

  • Coarse: ~4,300 vertices (1 PCA mode, 500-point partitions).
  • Intermediate: ~75,000 vertices.
  • Fine: ~3.4 million vertices (5 PCA modes, 100-point partitions).

The method effectively removed small-scale noise while maintaining the integrity of major features like fjords. A significant advantage is the automatic filtering of small islands; if an island is too small to form a full partition, it is omitted, preventing unnecessary high-resolution clusters in the mesh.

Conclusion and Future Work

The PCA framework provides a robust, algorithmic, and reproducible method for preparing geophysical data for mesh generation. Unlike manual CAD cleaning, this approach is scalable and verifiable. Future work will address potential topological issues, such as self-intersecting shorelines during simplification, and explore parallelization for global datasets. Additionally, the authors plan to integrate high-resolution local surveys with global data within this unified PCA framework and release the software as open-source.

Summary updated

BibTeX
@misc{avdis2015shoreline,
  author = {Alexandros Avdis and Christian T. Jacobs and Jon Hill and Matthew D. Piggott and Gerard J. Gorman},
  title = {{Shoreline and Bathymetry Approximation in Mesh Generation for Tidal Renewable Simulations}},
  year = {2015},
  eprint = {1510.01560},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1510.01560},
}
Figure