Integration of Geographic Information System frameworks into domain discretisation and meshing processes for geophysical models

Adam S. Candy, Alexandros Avdis, Jon Hill, Gerard Gorman, Matthew D. Piggott2014 · DOI

TL;DR

This paper presents a novel approach to integrate Geographic Information Systems (GIS) with geophysical model meshing, addressing the long-standing challenge of creating accurate, efficient, and reproducible meshes for complex domains. By leveraging mature GIS tools, the authors enable a more rigorous and automated workflow for model initialisation, paving the way for improved simulation accuracy and data provenance.

What Problem Does This Paper Address?

Accurate numerical simulations of geophysical phenomena (like ocean currents or atmospheric flows) critically depend on a precise discretization of the model domain, known as meshing. While simulation codes have become highly sophisticated, capable of handling complex, terrain-following boundaries, the pre-processing tools for creating these meshes have lagged behind. This gap leads to inefficient, error-prone, and difficult-to-reproduce meshing processes, especially for models with intricate geometries. The authors highlight the disconnect between the advanced capabilities of geophysical modeling and the often manual, ad-hoc nature of mesh generation. They propose bridging this gap by integrating GIS, a field with mature tools for handling and analyzing geospatial data, into the meshing workflow.

Methods

The core of the paper is the proposed integration of GIS into the meshing pipeline. This involves:

  • Leveraging GIS for Boundary Representation: Using GIS tools to define and manipulate complex geographical boundaries (coastlines, shorelines, bathymetry) with high fidelity. This includes processing vector data (like polylines) and raster data (like elevation maps).
  • Developing Spatial Metrics within GIS: Creating "metrics" that dictate the desired mesh resolution across the domain. These metrics are derived from various geospatial datasets (bathymetry, proximity to coastlines) and combined using GIS raster calculator tools. This allows for adaptive meshing, where finer elements are placed in areas of interest or high gradients.
  • Automated Workflow and Data Management: The GIS framework provides a unified environment for managing diverse datasets, applying transformations (like projections) on-the-fly, and automating many pre-processing steps. This hierarchy of automation allows for both high-level operations and fine-grained edits.
  • Interoperability: The approach emphasizes using standardized data formats (e.g., shapefiles, GeoTIFF, NetCDF) and interfaces, facilitating the integration of GIS with meshing tools like Gmsh.

Datasets

The study utilizes several publicly available geospatial datasets:

  • GSHHS (Global Self-consistent, Hierarchical, High-resolution Shoreline Database): Used for defining coastlines.
  • GEBCO (General Bathymetric Chart of the Oceans): Used for bathymetry data.
  • Marine Digimap Coastal Bathymetry: A higher-resolution bathymetry dataset for the specific study region.
  • OS Street View: A raster dataset providing detailed street-level imagery.
  • OS VectorMap District: A vector dataset defining tidal water boundaries.

The paper focuses on the Portland Harbour and Chesil Beach region in the UK as a case study, demonstrating the application of these datasets.

Key Insights and Ideas

  • GIS as a Unifying Platform: The paper strongly advocates for GIS not just as a data visualization tool, but as a powerful, integrated platform for the entire pre-processing pipeline, from data ingestion to mesh generation.
  • Provenance and Reproducibility: By using established GIS tools and standardized formats, the approach significantly enhances data provenance and the reproducibility of meshing results, crucial for scientific validation.
  • Bridging the Disjoint Fields: The work effectively bridges the gap between the mature GIS domain and the specialized field of geophysical meshing, bringing the rigor and automation of GIS to complex simulation setup.
  • Hierarchy of Automation: The proposed workflow supports a flexible hierarchy of automation, allowing for both fully automated processes and manual interventions for fine-tuning, which is essential for complex real-world problems.
  • Adaptive Meshing Driven by Geospatial Data: The ability to derive mesh metrics directly from geophysical features (like bathymetry and coastlines) allows for more intelligent and efficient mesh refinement.

Key Results

The paper demonstrates a practical implementation of the proposed GIS-integrated meshing approach. Key results include:

  • Successful Meshing of Complex Domains: The method successfully generated meshes for the Portland Harbour and Chesil Beach region, accurately capturing intricate features like coastlines, man-made structures, and bathymetric variations.
  • Enhanced Workflow Efficiency: The GIS-based approach streamlines the pre-processing workflow, reducing the manual effort and potential for errors associated with traditional methods.
  • Demonstrated Flexibility: The approach proved capable of handling datasets with large disparities in scale and resolution, integrating them into a consistent meshing strategy.
  • Visual and Interactive Development: The use of GIS tools allows for interactive development and prototyping of meshing strategies, making it easier to experiment with different parameters and datasets.

Limitations and Future Work

  • Performance for Very Large Datasets: While the approach is efficient, the authors acknowledge that for extremely large datasets or computationally intensive operations, pushing some processing to High Performance Computing (HPC) environments might be necessary.
  • Integration with Existing Codes: The paper highlights the need for further integration of these GIS-based meshing capabilities into existing geophysical modeling codes to ensure sustainability and broader adoption.
  • Extension to 3D Meshing: The current work primarily focuses on 2D surface meshing, with a mention of extruding to 3D. Further development in full 3D mesh generation from GIS data could be a future direction.
  • Standardization of Interaction: While the paper promotes standardization, further efforts are needed to establish common interfaces and protocols for seamless interaction between GIS, meshing, and simulation tools.

Summary updated

Abstract

Computational simulations of physical phenomena rely on an accurate discretisation of the model domain. Numerical models have increased in sophistication to a level where it is possible to support terrain-following boundaries that conform accurately to real physical interfaces, and resolve a multiscale of spatial resolutions. Whilst simulation codes are maturing in this area, pre-processing tools have not developed significantly enough to competently initialise these problems in a rigorous, efficient and recomputable manner. In the relatively disjoint field of Geographic Information Systems (GIS) however, techniques and tools for mapping and analysis of geographical data have matured significantly. If data provenance and recomputability are to be achieved, the manipulation and agglomeration of data in the pre-processing of numerical simulation initialisation data for geophysical models should be integrated into GIS. A new approach to the discretisation of geophysical domains is presented, and introduced with a verified implementation. This brings together the technologies of geospatial analysis, meshing and numerical simulation models. This platform enables us to combine and build up features, quickly drafting and updating mesh descriptions with the rigour that established GIS tools provide. This, combined with the systematic workflow, supports a strong provenance for model initialisation and encourages the convergence of standards.
BibTeX
@misc{candy2014integration,
  author = {Adam S. Candy and Alexandros Avdis and Jon Hill and Gerard Gorman and Matthew D. Piggott},
  title = {{Integration of Geographic Information System frameworks into domain discretisation and meshing processes for geophysical models}},
  year = {2014},
  doi = {10.5194/gmdd-7-5993-2014},
}
Figure