Skip to content

Latest commit

 

History

History
119 lines (73 loc) · 6.72 KB

readme.md

File metadata and controls

119 lines (73 loc) · 6.72 KB
Raw

pyDelft3D-FLOW

pyDelft3D-FLOW is a small Python package that were developed to help with 1) reading and writing of several input files of a Delft3D4-FLOW model, 2) auto-generating multiple successive runs, 3) (interactive) plotting of the Delft3D-FLOW output (NetCDF format only) as and 4) overwriting values in NetCDF output.

These were developed to view the output of modelling turbidity currents in Delft3D(4)-FLOW, which means 3D models that are entirely submerged. I will attempt to make everything I developed for my thesis to a clean and reusable package here. Check out the Jupyter notebooks to get up-and-running. If there's sufficient interest, I'll document and expand it a bit more.

Installation

Clone this repository: git clone https://github.com/JulesBlm/pyDelft3D-FLOW.git

Install pyDelft3D-FLOW with Poetry (preferred), Conda or pip.

Python Poetry

Why poetry?

Run poetry install.

Conda

Make a Python environment with Anaconda, Miniconda, Poetry or whatever you like.

Post-installation

There are three commands to run after installation to get the widgets and plot figures to play nicely with JupyterLab

JupyterLab

  1. To get ipywidgets working run the following two commands
jupyter nbextension enable --py widgetsnbextension

jupyter labextension install @jupyter-widgets/jupyterlab-manager jupyter-matplotlib@
  1. To get matplotlib plots working nicely in JupyterLab run the following command after installing
jupyter labextension install @jupyter-widgets/jupyterlab-manager
  1. To get HoloViews widgets working properly in JupyterLab, run
jupyter labextension install @pyviz/jupyterlab_pyviz
  1. (Optional) to get ITKwidgets working, run the following command after the previous ones
jupyter labextension install jupyterlab-datawidgets itkwidgets

Viewing Output

The scripts were developed with a focus on modelling sequential turbidity currents in Delft3D-FLOW, so the focus is on 3D models and cross-section of the 3D output. However, the scripts are valuable for any type of Delft3D(4)-FLOW output.

xarray

The core functionality is provided by xarray. Delft3D-FLOW writes output in the NetCDF3 64-bit format with metadata following the Climate & Forecast Conventions. xarray is Python package for opening and analysing multidimensional gridded data sets. xarray is particularly tailored to working with self-describing NetCDF files and has native support for CF convention metadata, like that written by Delft3D-FLOW. xarray introduces labels in the form of dimensions, coordinates and attributes on top of raw multidimensional arrays, which allows for a more intuitive, more concise, and less error-prone developer experience.

xarray integrates tightly with Dask under the hood to facilitate out-of-memory and parallel computations on large datasets that do not fit into memory. Dask arrays allow handling very large array operations using many small arrays known as chunks. This enables many successive simulations to be combined and opened as one dataset.

HoloViews/hvPlot

HoloViews is a a Python library that enables visual exploration of multi-dimensional parameter spaces using auto-generated widgets that read the datasets metadata. Thanks to built-in Dask and Datashader integration HoloViews scales easily to millions of datapoints. hvPlot is a convenience wrapper around xarray for plotting data with HoloViews for more advanced, interactive visualizations.

Three-dimensional

PyVista is a pure Python library wrapping the VTK library's Python bindings for a streamlined and intuitive toolset for 3D Visualization and mesh analysis and processing. PyVista can be used across platforms, has extensive documentation and is open-source with a permissive MIT Licence. Structures created in PyVista are immediately interoperable with any VTK-based software.

Optionally, install ITKwidgets to interactively visualize a PyVista mesh within a Jupyter notebook

poetry install --extras "itk"

Prior Work

To Do

I'd like to get to these

  • Move colorcet and imageio-ffmpeg to extras
  • Code formatting with Black
  • Proper package with correct imports and all that
  • Documentation with Sphinx with RST

Ideas

I have plenty of ideas for improvement that I most likely won't get to, here's some

Input

  • Automatically read parameters into right types (float, int). For example in Sed.py and Mor.py, now everthing is read as a string
  • Tighter integration with xarray
  • Match read file contents against dict of description so you know what the keywords are for.
    • Check if selected options are valid
  • Expand D3Dmodel class to include all model's information.
    • Plot all model times in one chart, discharge time, spin-up time, smoothing time
  • Better handling of ensembles of models, ie multiple scenarios.
  • Visualise vertical grid of input files (in 3D).
    • Visualise 3D vertical boundary profile in 3D
  • Testing
  • Support for unit-aware arrays with pint to define, operate and manipulate physical quantities. Pint is already integrated in xarray, this would be very useful in calculating sediment volumes of in-flow boundary conditions for example.

Output

  • Support for plotting non-uniform grids with enclosures in HoloViews.
  • Out-of-core (lazy) operations for summing vector components like velocity and bottom stress.
  • Efficiently plotting vector fields (quiver plots) without loading the all required data into memory at once.
  • Interactive 3D plotting with Panel or itkwidgets