argopandas

The goal of argopandas is to provide seamless access to Argo NetCDF files using a pandas DataFrame-based interface. It is a Python port of the argodata package for R. The package is under heavy development and we would love feedback on the interface or anything else about the package!

Installation

You can install the argopandas package using pip.

pip install argopandas

The package depends on pandas, numpy, netCDF4, and pyarrow, which install automatically if using pip or you can install also your favourite Python package manager. The argopandas package requires Python 3.6 or later.

Examples

The intended interface for most usage is contained in the argopandas module. You can import this as argo for pretty-looking syntax:

import argopandas as argo

The global indexes are available via argo.prof, argo.meta, argo.tech, argo.traj, argo.bio_prof, argo.synthetic_prof, and argo.bio_traj.

argo.meta.head(5)

By defaut, downloads are lazily cached from the Ifremer https mirror. You can use argo.url_mirror() or argo.file_mirror() with argo.set_default_mirror() to point argopandas at your favourite copy of Argo.

To get Argo data from one or more NetCDF files, subset the indexes and use one of the table accessors to download, cache, and read variables aligned along common dimensions. The accessor you probably want is the .levels accessor from the argo.prof index:

argo.prof.head(5).levels[['PRES', 'TEMP']]

Downloading 5 files from 'https://data-argo.ifremer.fr/dac/aoml/13857/profiles'
Reading 5 files

You can get data from every variable in an Argo NetCDF file using one of these accessors. The variables grouped in each are aligned along the same dimensions and are documented together in the Argo user's manual.

• All indexes have a .info accessor that contains length-one variables that aren't aligned along any dimensions
• argo.prof: argo.prof.levels, arog.prof.prof, argo.prof.calib, argo.prof.param, and argo.prof.history
• argo.traj: argo.traj.cycle, argo.traj.measurement, argo.traj.param, and argo.traj.history
• argo.tech: argo.tech.tech_param
• argo.meta: argo.meta.config_param, argo.meta.missions, argo.meta.trans_system, argo.meta.positioning_system, argo.meta.launch_config_param, argo.meta.sensor, and argo.meta.param

Once you have a data frame you do anything you'd do with a regular pd.DataFrame(), like plot your data using the built-in plot method:

import matplotlib.pyplot as plt
fig, ax = plt.subplots()
for label, df in argo.prof.head(5).levels.groupby('file'):
    df.plot(x='TEMP', y = 'PRES', ax=ax, label=label)
ax.invert_yaxis()

Reading 5 files

You can access all the index files for a particular float using argo.float(), which lazily filters all the indexes for a particular float ID.

float_obj = argo.float(13857)
float_obj.meta.info

Downloading 'https://data-argo.ifremer.fr/ar_index_global_meta.txt.gz'
Downloading 'https://data-argo.ifremer.fr/dac/aoml/13857/13857_meta.nc'
Reading 1 file