Dataset: OIO2-WOD2018

Name: Optimally Interpolated O2 anomalies based on World Ocean Database 2018
Published: 2023-01-11
Keywords: deoxygenation, oxygen, climate change, oceans

Cite This Dataset

DOI:10.26008/1912/bco-dmo.886218.1

Spatial Extent: N:90 E:180 S:-90 W:-180

Temporal Extent: 1965 - 2012

Project:

Mapping Dissolved Oxygen using Observations and Machine Learning (DO Machine Learning)

Principal Investigator:

Takamitsu Ito (Georgia Institute of Technology, GA Tech)

BCO-DMO Data Manager:

Shannon Rauch (Woods Hole Oceanographic Institution, WHOI BCO-DMO)

Version:

Version Date:

2023-01-11

Restricted:

Validated:

Yes

Current State:

Final no updates expected

Optimally Interpolated O2 anomalies based on World Ocean Database 2018

Abstract:

OIO2 is a gridded data product of dissolved oxygen interpolated from shipboard observations archived in the World Ocean Database 2018 (WOD18). The quality-controlled WOD18 data are averaged for each bin at 1°x1° and monthly resolution where mean, variance, and sample size are recorded from 1965 to 2014 for the bottle data, and from 1987 to 2014 for the CTD-O2 data.

Expand/Collapse All

Archival Copy

Version Date	Archive	Persistent Identifier	Date Assigned
2023-01-11	Marine Biological Laboratory/Woods Hole Oceanographic Institution Library (MBLWHOI DLA)	10.26008/1912/bco-dmo.886218.1	2023-01-27

Description

This is a gridded data product generated from many observational profiles from the World Ocean Database 2018. There are two types of measurement methods included in the data source: bottle O2 data measured by the Winkler titration method and data from the CTD-O2 sensor. The data product is available as a NetCDF file (OIO2_g1x1v47_1967_2012.nc).

Methods & Sampling

The monthly mean climatology is determined by calculating the climatological monthly mean combining the bottle and CTD-O2 data and then filling data gaps. The gridded map of climatological O2 is assembled to estimate oxygen anomalies for all 1°x1° grid cells following the optimal interpolation method which provides the least-square estimate of O2 anomaly field on regularly spaced grid cells. Stationary and isotropic Gaussian covariance is assumed throughout this study, with the e-folding length scale of 1,000 kilometers (km). This particular choice of length scale controls how far an observation can influence the far field.

The optimal interpolation is applied to each basin separately except for the Southern Ocean, for which data points from all basins southward of 30°S are used. As a consistency check, the calculated climatology is compared to the World Ocean Database 2018 and it is confirmed that the optimally interpolated O2 climatology closely matches that of WOA18. Departures from the monthly climatology are recorded as O2 anomalies. The binned data is very sparse at a monthly timescale. For each year, the monthly anomaly data is averaged into yearly anomalies neglecting the months with missing data. This step increases data coverage significantly while averaging out high-frequency variability in the data including changes shorter than the yearly timescale. In addition, a 5-year moving window (pentadal) averaging is applied to the yearly anomaly neglecting the years with missing data. This further increases the data coverage, while averaging out variability on a timescale shorter than 5 years. The resulting pentadal O2 anomaly data covers the 46-year period from 1967 to 2012.

Finally, the optimal interpolation is applied for each year to yield the optimally interpolated O2 anomalies. The resulting data product is saved in the netCDF format, and it includes O2 climatology, O2 anomalies, and standard error.

Data Processing Description

Data Processing:
Optimal interpolation is performed using the Gaussian covariance function with the e-folding scale of 1,000 kilometers (km). Briefly, it produces the best-fit O2 distribution in the least square sense, given the covariance structure in the dataset. The binned data is very sparse at a monthly timescale. For each year, the monthly anomaly data is averaged into yearly anomalies neglecting the months with missing data. This step increases data coverage significantly while averaging out high-frequency variability in the data including changes shorter than the yearly timescale. In addition, a 5-year moving window (pentadal) averaging is applied to the yearly anomaly neglecting the years with missing data. This further increases the data coverage, while averaging out variability on a timescale shorter than 5 years. The resulting pentadal O2 anomaly data covers the 46-year period from 1967 to 2012. Finally, the optimal interpolation is applied for each year to yield the optimally interpolated O2 anomalies.

Data Files

File

OIO2_g1x1v47_1967_2012.nc

(NetCDF, 2.11 GB)
MD5:e5e1248be2f0006559d3b67c83398762

File header:

netcdf OIO2_g1x1v47_1967_2012 {

dimensions:

lon = 360 ;

lat = 180 ;

depth = 47 ;

time = 46 ;

variables:

double lon(lon) ;

lon:_FillValue = -99999. ;

lon:standard_name = "lon" ;

lon:long_name = "longitude" ;

lon:units = "degrees_east" ;

double lat(lat) ;

lat:_FillValue = -99999. ;

lat:standard_name = "lat" ;

lat:long_name = "latitude" ;

lat:units = "degrees_north" ;

double depth(depth) ;

depth:_FillValue = -99999. ;

depth:standard_name = "depth" ;

depth:long_name = "depth from the surface ocean" ;

depth:units = "m" ;

depth:bounds = "depth_bnds" ;

double time(time) ;

time:_FillValue = -99999. ;

time:standard_name = "time" ;

time:long_name = "time" ;

time:units = "days since 1950-01-01" ;

time:calendar = "proleptic_gregorian" ;

double o2(time, depth, lat, lon) ;

o2:_FillValue = -99999. ;

o2:long_name = "objective map of dissolved oxygen anomaly" ;

o2:units = "micro-molO2/L" ;

double o2_standard_error(time, depth, lat, lon) ;

o2_standard_error:_FillValue = NaN ;

double o2_climatology(depth, lat, lon) ;

o2_climatology:_FillValue = NaN ;

// global attributes:

		:_NCProperties = "version=2,netcdf=4.8.1,hdf5=1.12.1" ;

}