PyFlexPlot

PyFlexPlot is a Python-based tool to visualize FLEXPART dispersion simulation results stored in NetCDF format.

Table of Contents

• Features
• Installation
• Run pyflexplot
  • Examples
• Development (CSCS)
• External Links
• License

Key Features

PDF and PNG Files

Pyflexplot allows to visualize data on a map plot and save the output in either PDF or PNG format. To utilize this feature, simply adjust the outfile variable with the appropriate file extension.

Shape File Generation

Furthermore, Pyflexplot provides the functionality to export data into shape files (.shp) to utilize them in GIS programs such as QGIS 3. The output is a ZIP archive containing the essential components of a shapefile: .shp, .dbf, .shx, .prj, and .shp.xml. Key aspects of this feature include:

• Filtering Zero Values: The tool initially removes zero values from fields (e.g., concentration) before processing.
• Logarithmic Transformation: Field values undergo a log_10 transformation to optimize the visualization of data ranges.
• Precision Handling: The transformed field values are recorded with 15 decimal places, accommodating the precision limitations of some GIS software.
• Metadata Storage: Information, such as details about released materials, are stored within a .shp.xml file as metadata.

Scaling the field values

Another feature is to manipulate the field values by scaling with an arbitrary factor. This factor can be set in the preset with the variable multiplier.

Installation

You can install pyflexplot from MCH pypi repository using pip:

pip install pyflexplot -i https://service.meteoswiss.ch/nexus/repository/python-all/simple

Run pyflexplot

The primary command for pyflexplot follows this structure:

pyflexplot [OPTIONS] CONFIG_FILE_DIRECTORY

To see the available options, run:

pyflexplot --help

If you want to run the following examples interactively, you may want to allocate parallel resources with the help of SLURM (if available), e.g. 10 cores:

salloc -c 10

To use all allocated cpus, add the --num-procs option to the pyflexplot command (note that for a complete pyflexplot command, the definition of a preset or input and output need to be added, see below):

pyflexplot --num-procs=$SLURM_CPUS_PER_TASK

Important: Free resources when done!

exit

Examples how to run pyflexplot

Example using default input file. This example assumes you are in the pyflexplot directory.

Default input files are searched for in ./data. If you want to use the files as defined in the presets for your tests, link thm into the root of the repository. At CSCS on Alps, use:

ln -s /store_new/mch/msopr/pyflexplot_testdata data

There are several default config files available under src/pyflexplot/data/presets/opr.

To produce graphics for a specific FLEXPART output, select the corresponding preset from the table below and define the preset variable accordingly:

Model	Type	Define Preset Variable
FLEXPART-IFS	global	preset=opr/ifs-hres/all_pdf
FLEXPART-IFS	Europe	preset=opr/ifs-hres-eu/all_pdf
FLEXPART-COSMO-1E-CTRL	deterministic	preset=opr/cosmo-1e-ctrl/all_pdf
FLEXPART-COSMO-2E-CTRL	deterministic	preset=opr/cosmo-2e-ctrl/all_pdf
FLEXPART-COSMO-1E	ensemble	preset=opr/cosmo-1e/all_pdf
FLEXPART-COSMO-2E	ensemble	preset=opr/cosmo-2e/all_pdf
FLEXPART-ICON-CH1-CTRL	deterministic	preset=opr/icon-ch1-ctrl/all_pdf
FLEXPART-ICON-CH2-EPS	ensemble	preset=opr/icon-ch2-eps/all_pdf

You may use the * wildcard to operate pyflexplot with several presets at once. For example, to run pyflexplot with all presets that produce the graphics in PDF format for a specific NWP model, define the preset variable as one of:

preset='opr/cosmo*/all_pdf'
preset='opr/icon*/all_pdf'
preset='opr/ifs*/all_pdf'

Note however that the preset in this form requires the respective default input files to be accessible through the ./data directory link.

Define an output directory and create it, if it does not exist, e.g.

nwp=$(echo ${preset} | cut -d/ -f2 | sed 's/*//g') # Extract NWP model name
dest=plot_$nwp
mkdir $dest

After selecting a preset, you may run pyflexplot interactively for the default test data:

pyflexplot --preset "$preset" --merge-pdfs --dest=$dest

On the production server at the CSCS, it is however highly recommended to create and run a batch job using the batchPP utility:

batchPP -t 2 -T 10 -n pfp_$nwp -- \
  $CONDA_PREFIX/bin/pyflexplot --preset "$preset" \
  --merge-pdfs --dest=$dest --num-procs=\$SLURM_CPUS_PER_TASK

To use your own or an operational FLEXPART output file in NetCDF format as input for pyflexplot, modify the settings of the preset with the help of the --setup option as follows:

pyflexplot --preset "$preset" --merge-pdfs --dest=$dest --setup infile <netcdf-file>

To use a FLEXPART ensemble as input, the placeholder {ens_member:03} may be used within the path of <netcdf-file>. Instead of 03 (for %03d), another C-style field width for the ensemble member field can be used.

Example using operational Flexpart ensemble output based on ICON-CH2-EPS:

preset=opr/icon-ch2-eps/all_pdf  # Preset for ICON-CH2-EPS
nwp=$(echo ${preset} | cut -d/ -f2 | sed 's/*//g')  # Extract NWP model name
basetime=$(date --utc --date="today 00" +%Y%m%d%H)  # Recent base time
# Get name of first input file
infile000=$(echo /store_new/mch/msopr/osm/ICON-CH2-EPS/FCST${basetime:2:2}/${basetime:2:8}_???/flexpart_c/000/grid_conc_*_BEZ.nc)
infile=${infile000/\/000\//\/\{ens_member:03\}\/}   # Input file definition
dest=plot_${basetime:2:8}    # Output directory with base time of NWP model
mkdir $dest                  # Create output directory
# Submit job with the help of the batchPP utility
batchPP -t 1 -T 10 -n pfp-$nwp -- \
  $CONDA_PREFIX/bin/pyflexplot --preset $preset \
    --merge-pdfs --setup infile $infile --setup base_time $basetime --dest=$dest \
    --num-procs=\$SLURM_CPUS_PER_TASK

The following expamles use FLEXPART output generated with the test-fp script in the test_meteoswiss subdirectory of the flexpart repository of MeteoSwiss. Define FP_JOBS as path to the FLEXPART output files that are to be used as input for pyflexplot, e.g.

FP_JOBS=$SCRATCH/flexpart/job

Write output to a location where you have write access, e.g.

FP_OUT=$SCRATCH/flexpart/job

After additionally defining the preset as above and nwp as the job name (directory name below FP_JOBS), create the output directory with

infile=$(echo $FP_JOBS/$nwp/output/*.nc)
basetime=$(cat $FP_JOBS/$nwp/output/plot_info)
dest=$FP_OUT/$nwp/plots
mkdir -p $dest

and submit the job with the batchPP command as above.

For ensembles, the infile needs to be a pattern rather than a single file:

infile000=$(echo $FP_JOBS/$nwp/output/000/*.nc)
infile=${infile000/\/000\//\/\{ens_member:03\}\/}

After job completion, list and visualize results e.g. with evince:

ls $dest/*pdf
evince $dest/*pdf

Running Pyflexplot with S3 input (and output)

In order to download input NETCDF data from S3, and S3 URI can be specified as the setup parameter infile as below (example):

pyflexplot --preset "$preset" --merge-pdfs --dest=$dest --setup infile s3://<s3-bucket-name>/flexpart_cosmo-2e_2021030503_{ens_member:03}_MUE.nc

In order to output the resulting plots to an S3 bucket, specify the S3 bucket name as the --dest. The plots will still be created locally at the dest dir path defined in the config/settings.yaml

pyflexplot --preset "$preset" --merge-pdfs --dest=s3://<s3-bucket-name>

Development

Prerequisites: Git, Miniconda (for installation of Conda) or Poetry

Install dependencies & start the service locally (CSCS)

Clone the repo and enter the project folder:

git clone [email protected]:MeteoSwiss-APN/pyflexplot.git && cd pyflexplot

Create a Conda (or mamba/micromamba) environment with only the desired Python version and activate:

conda create --yes --prefix ./.conda-env python=3.10
conda activate ./.conda-env

Install Poetry into this environment and configure Poetry to not create a new virtual environment. If it detects an already enabled virtual (eg Conda) environment it will install dependencies into it:

conda install --yes poetry
poetry config --local virtualenvs.create false

Install packages:

poetry install

Run the tests and quality tools

Run tests:

poetry run pytest

Run pylint to check code style of Python files (if any):

poetry run pylint src

Run mypy to check typing:

poetry run mypy

Updating the Test References

Pyflexplot includes a set of functionality tests that compare generated output against predefined reference data. These reference files, which contain summary dicts, begin with ref_ and have the nomal Python file ending .py, and are stored in the directory tests/slow/pyflexplot/test_plots. To update these reference files, uncomment the following line near the end of the file shared.py in the same directory:

_TestBase = _TestCreateReference

Then re-run the (slow) tests to generate the new reference files. After generating the new reference files, comment out the above line again or simply revert the file with git.

External Links

• pyshp - Python module to generate Shapefiles

License

This project is licensed under the terms of the MIT License. The full license text can be found in the LICENSE file. In essence, you are free to use, modify, and distribute the software, provided the associated copyright notice and disclaimers are included.