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!
You can install the argopandas
package from PyPI or conda-forge.
pip install argopandas
conda install --channel=conda-forge argopandas
The package depends on pandas
, numpy
, netCDF4
, and pyarrow
, which install automatically if using pip
or conda install
. The argopandas
package requires Python 3.6 or later.
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)
file | profiler_type | institution | date_update | |
---|---|---|---|---|
0 | aoml/13857/13857_meta.nc | 845 | AO | 2018-10-11 20:00:14+00:00 |
1 | aoml/13858/13858_meta.nc | 845 | AO | 2018-10-11 20:00:15+00:00 |
2 | aoml/13859/13859_meta.nc | 845 | AO | 2018-10-11 20:00:25+00:00 |
3 | aoml/15819/15819_meta.nc | 845 | AO | 2018-10-11 20:00:16+00:00 |
4 | aoml/15820/15820_meta.nc | 845 | AO | 2018-10-11 20:00:18+00:00 |
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
PRES | TEMP | |||
---|---|---|---|---|
file | N_PROF | N_LEVELS | ||
aoml/13857/profiles/R13857_001.nc | 0 | 0 | 11.900000 | 22.235001 |
1 | 17.000000 | 21.987000 | ||
2 | 22.100000 | 21.891001 | ||
3 | 27.200001 | 21.812000 | ||
4 | 32.299999 | 21.632000 | ||
... | ... | ... | ... | ... |
aoml/13857/profiles/R13857_005.nc | 0 | 102 | 976.500000 | 4.527000 |
103 | 986.700012 | 4.527000 | ||
104 | 996.799988 | 4.533000 | ||
105 | 1007.000000 | 4.487000 | ||
106 | 1017.200012 | 4.471000 |
551 rows × 2 columns
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
, andargo.prof.history
argo.traj
:argo.traj.cycle
,argo.traj.measurement
,argo.traj.param
, andargo.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
, andargo.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)
dict(float_obj.meta.info.iloc[0])
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
{'DATA_TYPE': 'Argo meta-data ',
'FORMAT_VERSION': '3.1 ',
'HANDBOOK_VERSION': '1.2 ',
'DATE_CREATION': '20181011200014',
'DATE_UPDATE': '20181011200014',
'PLATFORM_NUMBER': '13857 ',
'PTT': '09335 ',
'PLATFORM_FAMILY': 'FLOAT ',
'PLATFORM_TYPE': 'PALACE ',
'PLATFORM_MAKER': 'WRC ',
'FIRMWARE_VERSION': '012797 ',
'MANUAL_VERSION': '012797 ',
'FLOAT_SERIAL_NO': '28 ',
'STANDARD_FORMAT_ID': 'n/a ',
'DAC_FORMAT_ID': 'PALACE_T1 ',
'WMO_INST_TYPE': '845 ',
'PROJECT_NAME': 'ACCE (Atlantic Circulation and Climate Experiment) ',
'DATA_CENTRE': 'AO',
'PI_NAME': 'BOB MOLINARI ',
'ANOMALY': 'n/a ',
'BATTERY_TYPE': 'Alkaline ',
'BATTERY_PACKS': 'board - 1 (s/n: 32); ',
'CONTROLLER_BOARD_TYPE_PRIMARY': 'n/a ',
'CONTROLLER_BOARD_TYPE_SECONDARY': 'n/a ',
'CONTROLLER_BOARD_SERIAL_NO_PRIMARY': '32 ',
'CONTROLLER_BOARD_SERIAL_NO_SECONDARY': 'n/a ',
'SPECIAL_FEATURES': 'n/a ',
'FLOAT_OWNER': 'BOB MOLINARI ',
'OPERATING_INSTITUTION': 'NOAA, AOML, Miami ',
'CUSTOMISATION': 'n/a ',
'LAUNCH_DATE': '19970719145500',
'LAUNCH_LATITUDE': 0.007516666781157255,
'LAUNCH_LONGITUDE': -15.013999938964844,
'LAUNCH_QC': b'1',
'START_DATE': '19970719163000',
'START_DATE_QC': b'1',
'STARTUP_DATE': '19970719103000',
'STARTUP_DATE_QC': b'1',
'DEPLOYMENT_PLATFORM': 'R/V Seward Johnson ',
'DEPLOYMENT_CRUISE_ID': '97-03 ',
'DEPLOYMENT_REFERENCE_STATION_ID': 'CTD 108 ',
'END_MISSION_DATE': ' ',
'END_MISSION_STATUS': nan}