Migrating to 0.3.0
0.3.0 changes the output: reconvert your missions. Nothing on this page needs a code change
unless you read PHASE, PHASE_QC, PROFILE_NUMBER, DIVE_NUMBER (now
dive_number), SA/CT, uri, a passed-through variable in cm/s, a data point by its
index around a GPS fix, or the SENSOR_* serial numbers and dates, or still import a pre-0.1.0
name.
PHASE
PHASE is assigned from what the glider recorded it did (see How the phase is assigned). On the same
dive, 0.2.0 and 0.3.0 differ like this:
0.2.0 |
0.3.0 |
|
|---|---|---|
values used |
1 ascent, 2 descent, 3 at the first two GPS fixes |
1 ascent, 2 descent, 3 surfacing, 4 parking (a loiter), 5 inflection, 0 unknown |
descent / ascent split |
at the pressure maximum |
at the glider’s apogee (state table, or the G&C event that moves the VBD to zero) |
bottom of the dive |
descent to the pressure maximum, then ascent |
inflection (the apogee); the deepest sample can fall in the first ascent data points |
end of the climb |
ascent |
ascent through the surface coast (the OG1 table does not name it), then surfacing from begin surface |
GPS fixes |
the first two 3, the last one ascent |
every fix 3 |
data type |
byte, |
byte, |
attributes |
|
also |
Code that selects profiles with PHASE == 2 / PHASE == 1 still works, but the apogee
samples (5) are no longer in either. To keep every sample, split
each dive at its inflection, or at its deepest sample as 0.2.0 did.
xarray reads PHASE with its _FillValue 0 as missing, so PHASE 0 (unknown) comes back
as NaN; xr.open_dataset(..., mask_and_scale=False) or ds.PHASE.fillna(0) keeps the 0.
PROFILE_NUMBERNow from
PHASE, as OG1 defines it: one number per run of descent or ascent, -9999 on every other data point (surfacing, inflection). 0.2.0 split each dive at the pressure maximum, numbered the halves 2n-1 and 2n from the dive number, and gave every data point a number.tools.add_profile_number(and its old nameassign_profile_number), which did that, is removed: calltools.add_phase, thentools.add_profiles.PROFILE_DIRECTION,SEGMENT_NUMBERNew (OG1): 1/-1/0 for descent/ascent/other, and +1 each time the glider enters surfacing.
dive_numberRenamed from
DIVE_NUMBER: it is the Seaglider’s dive number, not an OG1 variable, so it is written in lower case like the other non-OG1 variables.PHASE_QC0 (no QC performed) on every data point, with the OG1 flag scale 0–4 as
flag_valuesandflag_meanings. 0.2.0 wrote 0 with no flag attributes.
Salinity: no SA or CT
Files carry practical salinity (PSAL) only, as TEOS-10 recommends archiving. 0.2.0 renamed a
basestation file’s absolute_salinity and conservative_temperature to SA and CT;
they now pass through under those names, with the basestation’s own attributes (for example
units = "g/kg", no standard_name). POTDENS0 is still written (computed with TEOS-10).
Compute SA and CT when you need them:
import gsw
SA = gsw.SA_from_SP(ds.PSAL, ds.PRES, ds.LONGITUDE, ds.LATITUDE)
CT = gsw.CT_from_t(SA, ds.TEMP, ds.PRES)
Variable attributes
uriis renamedvocabulary(OG1’s name); readvocabulary. See Variable vocabulary.Every variable on
N_MEASUREMENTScarriescoordinates = "TIME LONGITUDE LATITUDE DEPTH".Units are written in one canonical spelling on every variable, including those that keep their basestation names (
g kg-1,kg m-3,degree_Celsius,m,degree). A passed-through variable incm/sis converted tom s-1, withoriginal_unitsand aconversion_unitsentry. Spellings the converter does not recognise are left as written.
GPS fixes: one data point each
Each GPS fix is now one data point, so TIME normally has no duplicate at a fix and
sel(TIME=...), swap_dims and resampling work on it. 0.2.0 wrote each fix twice (the merged gps_info
record and an appended copy carrying LATITUDE_GPS/LONGITUDE_GPS/TIME_GPS); the
*_GPS variables and the fix’s surface LATITUDE/LONGITUDE and DEPTH (0 m) now sit on
the one gps_info data point. A fix logged twice within a dive is written once. The same fix
re-logged at a dive boundary is written once when the later copy carries nothing but the fix
(the dropped copy is recorded in the conversion_dropped_gps_fixes global); if it carries other
data, both are kept, TIME keeps the duplicate there, and the conversion logs a WARNING.
N_MEASUREMENTS is shorter by the removed fixes; TIME_GPS is the fix time as the GPS
recorded it (log_gps_time).
Basestation 2.10 files
Files whose gc_state_secs are seconds from the dive start (basestation 2.10) no longer put
data points in January 1970. Their output name and time_coverage_start change from
sg033_19700101T000012_R to the dive’s real start, for example sg033_20100903T181710_R.
Sensor serial numbers and calibration dates
Serial numbers and calibration dates are read from the calibration comments by named rules (see How calibration comments are read); files differ from 0.2.0 wherever the old parser was wrong.
sample |
0.2.0 wrote |
0.3.0 |
|---|---|---|
sg005 CTD ( |
|
|
sg005 WET Labs (no comment) |
|
|
A serial or date that cannot be read is
UNK;0000,00-00-0000,0000-00-00,Format ErrorandNone Foundare no longer written.Two-number dates are read by separator: slashes month-first (
9/10/08→ 10 September), dots and dashes day-first (12.05.2012→ 12 May). 0.2.0 read every such date day-first.New on each
SENSOR_*:original_calibcomm(the comment as written) and, for multi-channel instruments,sensor_calibration_dates. New global:conversion_sensor_calibration.sensors.<key>.calibration_dateinmission.yamlmust be an ISO date (YYYY-MM-DD).
Removed names
The pre-0.1.0 names that warned or raised through 0.1.x and 0.2.x are removed. Using one gives
Python’s ordinary AttributeError (ImportError for from ... import); Migrating to 0.1.0 gives each one’s replacement:
readers:load_basestation_files,load_first_basestation_file,scan_and_repair_files,repair_netcdf_time_metadata_inplace,profnum_from_filename,glider_sn_from_filename,load_sample_dataset;writers:save_dataset;tools:calc_Z,assign_phase,reformat_units_str,reformat_units_var,convert_units_var,OG1_name_mapping,print_OG1_mapping_summaryand the pipeline internals (og1_name_mapping,gather_sensor_info,add_sensor_to_dataset,merge_datasets_along_time,extract_scalar_parameters,add_scalar_parameters,set_best_dtype);contributors:registry_institutions,normalize_orcid,normalize_role;vocabularies:load_default_contributors,contrib_to_append,unit_str_format,unit1_to_unit2,preferred_units;the
convertOG1module;plotters.show_*no longer accept a file path: pass an open dataset.
tools.add_profile_number (and its old name assign_profile_number) is also removed; see
PROFILE_NUMBER above.