Contributor’s Guide

The Python ARM Radar Toolkit (Py-ART)

The Python ARM Radar Toolkit, Py-ART, is an open source Python module containing a growing collection of weather radar algorithms and utilities build on top of the Scientific Python stack and distributed under the 3-Clause BSD license. Py-ART is used by the Atmospheric Radiation Measurement (ARM) Climate Research Facility for working with data from a number of precipitation and cloud radars, but has been designed so that it can be used by others in the radar and atmospheric communities to examine, processes, and analyze data from many types of weather radars.


If you use the Python ARM Radar Toolkit (Py-ART) to prepare a publication please cite:

Helmus, J.J. & Collis, S.M., (2016). The Python ARM Radar Toolkit (Py-ART), a Library for Working with Weather Radar Data in the Python Programming Language. Journal of Open Research Software. 4(1), p.e25. DOI:

Py-ART implements many published scientific methods which should also be cited if you make use of them. Refer to the References section in the documentation of the functions used for information on these citations.


The easiest method for installing Py-ART is to use the conda packages from the latest release. To do this you must download and install Anaconda or Miniconda. Then use the following command in a terminal or command prompt to install the latest version of Py-ART:

conda install -c conda-forge arm_pyart

To update an older version of Py-ART to the latest release use:

conda update -c conda-forge arm_pyart

Code Style

Py-ART follows pep8 coding standards. To make sure your code follows the pep8 style, you can use a variety of tools that can check for you. Two popular pep8 check modules are pycodestyle and pylint.

For more on pep8 style:

To install pycodestyle:

conda install pycodestyle

To use pycodestyle:

pycodestyle filename

To install pylint:

conda install pylint

To get a detailed pylint report:

pylint filename

If you want to just see what line number and the issue, just use:

pylint -r n filename

Both of these tools are highly configurable to suit a user’s taste. Refer to the tools documentation for details on this process.

Python File Setup

In a new .py file, the top of the code should have a brief introduction to the module.

An example:

Retrieval of VADs from a radar object.


Following the introduction code, modules are then added. To follow pep8 standards, modules should be added in the order of:

  1. Standard library imports.

  2. Related third party imports.

  3. Local application/library specific imports.

For example:

import glob
import os

import numpy as np
import as ma
from scipy.interpolate import interp1d

from ..core import HorizontalWindProfile

Following the main function def line, but before the code within it, a doc string is needed to explain arguments, returns, references if needed, and other helpful information. These documentation standards follow the NumPy documentation style.

For more on the NumPy documentation style:

An example:

def velocity_azimuth_display(
    radar, velocity=None, z_want=None, valid_ray_min=16,
    gatefilter=False, window=2):

    Velocity azimuth display.

    radar : Radar
        Radar object used.
    velocity : string
        Velocity field to use for VAD calculation.
        If None, the default velocity field will be used.

    Other Parameters
    z_want : array
        Height array user would like for the VAD
        calculation. None will result in a z_want of
        np.linspace and use of _inverse_dist_squared
        and _Average1D functions. Note, height must have
        same shape as expected u_wind and v_wind if user
        provides z_want.
    valid_ray_min : int
        Amount of rays required to include that level in
        the VAD calculation.
    gatefilter : GateFilter
        Used to correct the velocity field before its use
        in the VAD calculation. Uses Py-ART's region dealiaser.
    window : int
        Value to use for window calculation in _Averag1D

    height : array
        Heights in meters above sea level at which horizontal
        winds were sampled.
    speed : array
        Horizontal wind speed in meters per second at each height.
    direction : array
        Horizontal wind direction in degrees at each height.
    u_wind : array
        U-wind mean in meters per second.
    v_wind : array
        V-wind mean in meters per second.

    K. A. Browning and R. Wexler, 1968: The Determination
    of Kinematic Properties of a Wind Field Using Doppler
    Radar. J. Appl. Meteor., 7, 105–113


As seen, each argument has what type of object it is, an explanation of what it is, mention of units, and if an argument has a default value, a statement of what that default value is and why.

Private or smaller functions and classes can have a single line explanation.

An example:

def u_wind(self):
""" U component of horizontal wind in meters per second. """


When adding a new function to pyart it is important to add your function to the file under the corresponding pyart folder.

Create a test for your function and have assert from numpy testing test the known values to the calculated values. If changes are made in the future to pyart, pytest will use the test created to see if the function is still valid and produces the same values. It works that, it takes known values that are obtained from the function, and when pytest is ran, it takes the test function and reruns the function and compares the results to the original.

An example:

def test_vad():
    test_radar = pyart.testing.make_target_radar()
    height = np.arange(0, 1000, 200)
    speed = np.ones_like(height) * 5
    direction = np.array([0, 90, 180, 270, 45])
    profile = pyart.core.HorizontalWindProfile(
        height, speed, direction)
    sim_vel = pyart.util.simulated_vel_from_profile(
        test_radar, profile)

    test_radar.add_field('velocity', sim_vel,

    velocity = 'velocity'
    z_start = 0
    z_end = 10
    z_count = 5

    vad_height = ([0., 2.5, 5., 7.5, 10.])
    vad_speed = ([4.98665725, 4.94020686, 4.88107152,
                  4.81939374, 4.75851962])
    vad_direction = ([359.84659496, 359.30240553, 358.58658589,
                      357.81073051, 357.01353486])
    u_wind = ([0.01335138, 0.06014712, 0.12039762,
               0.18410404, 0.24791911])
    v_wind = ([-4.98663937, -4.9398407, -4.87958641,
               -4.81587601, -4.75205693])

    vad = pyart.retrieve.velocity_azimuth_display(test_radar,
                                                  z_start, z_end,

    assert_almost_equal(vad.height, vad_height, 3)
    assert_almost_equal(vad.speed, vad_speed, 3)
    assert_almost_equal(vad.direction, vad_direction, 3)
    assert_almost_equal(vad.u_wind, u_wind, 3)
    assert_almost_equal(vad.v_wind, v_wind, 3)

Pytest is used to run unit tests in pyart.

It is recommended to install pyart in “editable” mode for pytest testing. From within the main pyart directory:

pip install -e .

This lets you change your source code and rerun tests at will.

To install pytest:

conda install pytest

To run all tests in pyart with pytest from outside the pyart directory:

pytest --pyargs pyart

All test with increase verbosity:

pytest -v

Just one file:

pytest filename

Note: When an example shows filename as such:

pytest filename

filename is the filename and location, such as:

pytest /home/user/pyart/pyart/io/tests/

Relative paths can also be used:

cd pyart
pytest ./pyart/retrieve/tests/

For more on pytest:


When contributing to pyart, the changes created should be in a new branch under your forked repository. Let’s say the user is adding a new map display. Instead of creating that new function in your master branch. Create a new branch called ‘cartopy_map’. If everything checks out and the admin accepts the pull request, you can then merge the master branch and cartopy_map branch.

To delete a branch both locally and remotely, if done with it:

git push origin --delete <branch_name>
git branch -d <branch_name>

or in this case:

git push origin --delete cartopy_map
git branch -d cartopy_map

To create a new branch:

git checkout -b <branch_name>

If you have a branch with changes that have not been added to a pull request but you would like to start a new branch with a different task in mind. It is recommended that your new branch is based on your master. First:

git checkout master


git checkout -b <branch_name>

This way, your new branch is not a combination of your other task branch and the new task branch, but is based on the original master branch.

Typing git status will not only inform the user of what files have been modified and untracked, it will also inform the user of which branch they are currently on.

To switch between branches, simply type:

git checkout <branch_name>

When commiting to GitHub, start the statement with a acronym such as ‘ADD:’ depending on what your commiting, could be ‘MAINT:’ or ‘BUG:’ or more. Then following should be a short statement such as “ADD: Adding cartopy map display.”, but after the short statement, before finishing the quotations, hit enter and in your terminal you can then type a more in depth description on what your commiting.

A set of recommended acronymns can be found at:

If you would like to type your commit in the terminal and skip the default editor:

git commit -m "STY: Removing whitespace from pep8."

To use the default editor(in Linux, usually VIM), simply type:

git commit

One thing to keep in mind is before doing a pull request, update your branches with the original upstream repository.

This could be done by:

git fetch upstream

After fetching, a git merge is needed to pull in the changes.

This is done by:

git merge upstream/master

To prevent a merge commit:

git merge --ff-only upstream/master

After creating a pull request through GitHub, two outside checkers, Appveyor and TravisCI will determine if the code past all checks. If the code fails either tests, as the pull request sits, make changes to fix the code and when pushed to GitHub, the pull request will automatically update and TravisCI and Appveyor will automatically rerun.