pyaxiom
An ocean data toolkit developed and used by Axiom Data Science
Installation
Stable
$ conda install -c conda-forge pyaxiom
Development
$ conda install -c axiom-data-science pyaxiom
netcdf4-python
Dataset object
Enhanced A subclass of the netCDF4.Dataset
object that adds some additional features
Safe closing
Vanilla netCDF4.Dataset
objects raise a RuntimeError when trying to close
an already closed file. This won't raise.
from netCDF4 import Dataset
nc = Dataset('http://thredds45.pvd.axiomalaska.com/thredds/dodsC/grabbag/USGS_CMG_WH_OBS/WFAL/9001rcm-a.nc')
nc.close()
nc.close()
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-18-db44c06d8538> in <module>()
----> 1 nc.close()
/home/kwilcox/.virtualenvs/overlord/lib/python2.7/site-packages/netCDF4.so in netCDF4.Dataset.close (netCDF4.c:23432)()
RuntimeError: NetCDF: Not a valid ID
from pyaxiom.netcdf.dataset import EnhancedDataset as Dataset
nc = Dataset('http://thredds45.pvd.axiomalaska.com/thredds/dodsC/grabbag/USGS_CMG_WH_OBS/WFAL/9001rcm-a.nc')
nc.close()
nc.close()
Retrieving variables by attributes and values/callables
from pyaxiom.netcdf.dataset import EnhancedDataset as Dataset
nc = Dataset('http://thredds45.pvd.axiomalaska.com/thredds/dodsC/grabbag/USGS_CMG_WH_OBS/WFAL/9001rcm-a.nc')
# Return variables with a standard_name attribute equal to 'latitude'
print nc.get_variables_by_attributes(standard_name='latitude')
[<type 'netCDF4.Variable'>
float64 latitude()
units: degrees_north
standard_name: latitude
long_name: sensor latitude
unlimited dimensions:
current shape = ()
filling off
]
# Return all variables with a 'standard_name attribute'
variables = nc.get_variables_by_attributes(standard_name=lambda v: v is not None)
print [s.name for s in variables]
[u'latitude', u'longitude', u'depth', u'T_28', u'CS_300', u'CD_310', u'u_1205', u'v_1206', u'O_60', u'DO', u'time']
# Get creative... return all variablse with the attribute units equal to m/s and a grid_mapping attribute
variables = nc.get_variables_by_attributes(grid_mapping=lambda v: v is not None, units='m/s')
print [s.name for s in variables]
[u'CS_300', u'u_1205', u'v_1206']
IOOS URNs
URN Normalization
from pyaxiom.urn import IoosUrn
u = IoosUrn(asset_type='station', authority='axiom', label='station1')
print u.__dict__
{'asset_type': 'station',
'authority': 'axiom',
'component': None,
'label': 'station1',
'version': None}
print u.urn
'urn:ioos:station:axiom:station1'
from pyaxiom.urn import IoosUrn
u = IoosUrn.from_string('urn:ioos:station:axiom:station1')
print u.__dict__
{'asset_type': 'station',
'authority': 'axiom',
'component': None,
'label': 'station1',
'version': None}
print u.urn
'urn:ioos:station:axiom:station1'
NetCDF Integration
from pyaxiom.utils import urnify, dictify_urn
# NetCDF variable attributes from a "sensor" urn
print dictify_urn('urn:ioos:sensor:axiom:station1')
{'standard_name': 'wind_speed'}
print dictify_urn('urn:ioos:sensor:axiom:foo:lwe_thickness_of_precipitation_amount#cell_methods=time:mean,time:variance;interval=pt1h')
{'standard_name': 'lwe_thickness_of_precipitation_amount',
'cell_methods': 'time: mean time: variance (interval: PT1H)'}
# URN from `dict` of variable attributes
attributes = {'standard_name': 'wind_speed',
'cell_methods': 'time: mean (interval: PT24H)'}
print urnify('authority', 'label', attributes)
'urn:ioos:sensor:authority:label:wind_speed#cell_methods=time:mean;interval=pt24h'
# URN from a `netCDF4` Variable object
nc = netCDF4.Dataset('http://thredds45.pvd.axiomalaska.com/thredds/dodsC/grabbag/USGS_CMG_WH_OBS/WFAL/9001rcm-a.nc')
print urnify('authority', 'label', nc.variables['T_28'])
'urn:ioos:sensor:authority:label:sea_water_temperature'
Gridded NetCDF Collections
Binning files
pyaxiom
installs an executable called binner
that will combine many
files into a single file. Useful for cleanup and optimization.
If you have a script that is opening and reading hundreds of files, those open operations are slow, and you should combine them into a single file. This doesn't handle files that overlap in time or files that have data on both sides of a bin boundary.
usage: binner [-h] -o OUTPUT -d {day,month,week,year} [-f [FACTOR]]
[-n [NCML_FILE]] [-g [GLOB_STRING]] [-a] [-s HARD_START]
[-e HARD_END]
optional arguments:
-h, --help show this help message and exit
-o OUTPUT, --output OUTPUT
Directory to output the binned files to
-d {day,month,week,year}, --delta {day,month,week,year}
Timedelta to bin by
-f [FACTOR], --factor [FACTOR]
Factor to apply to the delta. Passing a '2' would be
(2) days or (2) months. Defauts to 1.
-n [NCML_FILE], --ncml_file [NCML_FILE]
NcML containing an aggregation scan to use for the
individual files. One of 'ncml_file' or 'glob_string'
is required. If both are passed in, the 'glob_string'
is used to identify files for the collection and the
'ncml_file' is applied against each member.
-g [GLOB_STRING], --glob_string [GLOB_STRING]
A Python glob.glob string to use for file
identification. One of 'ncml_file' or 'glob_string' is
required. If both are passed in, the 'glob_string' is
used to identify files for the collection and the
'ncml_file' is applied against each member.
-a, --apply_to_members
Flag to apply the NcML to each member of the
aggregation before extracting metadata. Ignored if
using a 'glob_string'. Defaults to False.
-s HARD_START, --hard_start HARD_START
A datetime string to start the aggregation from. Only
members starting on or after this datetime will be
processed.
-e HARD_END, --hard_end HARD_END
A datetime string to end the aggregation on. Only
members ending before this datetime will be processed.
Examples
Directory globbing
binner \
--output ./output/monthly_bins \
--glob_string "pyaxiom/tests/resources/coamps/cencoos_4km/wnd_tru/10m/*.nc" \
-d month \
-f 1
Directory globbing and applying NcML file to each member
binner \
--output ./output/monthly_bins \
--glob_string "pyaxiom/tests/resources/coamps/cencoos_4km/wnd_tru/10m/*.nc" \
-n pyaxiom/tests/resources/coamps_10km_wind.ncml \
-d month \
-f 1
<scan>
element
NcML aggregation reading the binner \
--output ./output/monthly_bins \
-n pyaxiom/tests/resources/coamps_10km_wind.ncml \
-d month \
-f 1
Creating CF1.6 TimeSeries files
TimeSeries
from pyaxiom.netcdf.sensors import TimeSeries
filename = 'test_timeseries.nc'
times = [0, 1000, 2000, 3000, 4000, 5000]
verticals = None
ts = TimeSeries(output_directory='./output',
latitude=32, # WGS84
longitude=-74, # WGS84
station_name='timeseries_station',
global_attributes=dict(id='myid'),
output_filename='timeseries.nc',
times=times,
verticals=verticals)
values = [20, 21, 22, 23, 24, 25]
attrs = dict(standard_name='sea_water_temperature')
ts.add_variable('temperature', values=values, attributes=attrs)
ts.close()
TimeSeriesProfile
from pyaxiom.netcdf.sensors import TimeSeries
times = [0, 1000, 2000, 3000, 4000, 5000] # Seconds since Epoch
verticals = [0, 1, 2] # Meters down
ts = TimeSeries(output_directory='./output',
latitude=32, # WGS84
longitude=-74, # WGS84
station_name='timeseriesprofile_station',
global_attributes=dict(id='myid'),
output_filename='timeseriesprofile.nc',
times=times,
verticals=verticals)
values = np.repeat([20, 21, 22, 23, 24, 25], len(verticals))
attrs = dict(standard_name='sea_water_temperature')
ts.add_variable('temperature', values=values, attributes=attrs)
ts.close()
Pandas Integration
Pandas integration assumes that there is a Series column time
and a Series
column depth
in your DataFrame. Data values are pulled from a column named
'value', but you may also pass in the data_column
attribute for more control.
from pyaxiom.netcdf.sensors import TimeSeries
df = pd.DataFrame({ 'time': [0, 1, 2, 3, 4, 5, 6],
'value': [10, 20, 30, 40, 50, 60],
'depth': [0, 0, 0, 0, 0, 0] })
TimeSeries.from_dataframe(df,
output_directory='./output',
latitude=30, # WGS84
longitude=-74, # WGS84
station_name='dataframe_station',
global_attributes=dict(id='myid'),
variable_name='values',
variable_attributes=dict(),
output_filename='from_dataframe.nc')
df = pd.DataFrame({ 'time': [0, 1, 2, 3, 4, 5, 6],
'temperature': [10, 20, 30, 40, 50, 60],
'depth': [0, 0, 0, 0, 0, 0] })
TimeSeries.from_dataframe(df,
output_directory='./output',
latitude=30, # WGS84
longitude=-74, # WGS84
station_name='dataframe_station',
global_attributes=dict(id='myid'),
output_filename='from_dataframe.nc',
variable_name='temperature',
variable_attributes=dict(standard_name='air_temperature'),
data_column='temperature')