SODA4¶

Simple Ocean Data Assimilation version 4 (SODA4.15.2)

Available

At a glance¶

Note on period: The model starts on 1980-01-01 from initial conditions interpolated from SODA3.15.2. As of 2026-10-12, the 5-day files start on 1980_02_02 for the regridded data and the original-grid sea ice, and on 1983_01_02 for the original-grid ocean (original-grid files for 1980–1982 are permanently missing). The last files are 2024_06_20 for the regridded ocean and 2024_06_25 for the original-grid ocean and regridded sea ice. The portal and paper say "1980–2024", but 2024 covers only the first half of the year, and the 2024 monthly files are half the size of other years.

Overview¶

SODA4 is the fourth version of the University of Maryland Simple Ocean Data Assimilation global ocean/sea-ice reanalysis. It keeps the GFDL MOM5/SIS1 numerics and ERA5 forcing (COARE bulk formulae) of SODA3 but raises the resolution to 1/10° with 75 levels, making it eddy-resolving over most of the globe and eddy-permitting in the Arctic. World Ocean Database 2023 temperature and salinity profiles (including Argo and gliders) and NOAA satellite SST calibrated with ICOADS are assimilated by optimal interpolation on a 10-day cycle with incremental analysis updates, and the sea ice is constrained by relaxing the surface heat flux toward GIOMAS thickness. The 1980–2024 record (on the server, to June 2024) was produced as four roughly decade-long streams.

Available variables¶

Main variables from Tables S2–S5 of the Chepurin et al. (2025) supplement and the SODA4 README. Names are the netCDF variable names. The README and supplement sometimes disagree (w/wt, aice/cn, mx·my/tx_trans·ty_trans), so check the actual files. Variables without a note are in both the original and regridded files.

instantaneous mean accumulated maximum minimum static — instantaneous: value at the time stamp · mean/accumulated/maximum/minimum: over the indicated interval · static: does not change with time

3-D ocean (ocean, 75 levels)¶

2-D ocean (ocean)¶

Sea ice (ice, SIS1 with 5 categories)¶

Transports (transport, original grid only)¶

Analysis increments (1°, upper 33 z* levels)¶

The full variable list is in the official documentation.

How to get the data¶

• SODA4 portal (UMD) — Information page (file links and wget examples)
  • Version description, links to yearly monthly-mean files, wget commands for bulk download, and how to cite.
• UMD data server — SODA4.15.2 regridded (0.1°) — Direct HTTP(S) download No login needed
  • 5-day and monthly netCDF files in ocean/ and ice/. A 5-day ocean file is about 1.5–1.6 GB, a yearly monthly-mean ocean file 17–19 GB, and a yearly monthly-mean ice file about 0.4 GB. On 2026-10-12 the 2021 monthly ice file was missing.
• UMD data server — SODA4.15.2 original grid (tripolar 3600 × 2700) — Direct HTTP(S) download No login needed
  • Folders ocean/, ice/ and transport/. A 5-day ocean file is about 5.2–5.4 GB, a yearly monthly-mean ocean file about 65 GB, a 5-day ice file about 145 MB and a 10-day transport file about 2.2 GB. Original-grid ocean files start in 1983. The portal recommends computing basin-integrated transports (e.g. AMOC) on the original grid, because remapping can introduce truncation errors.
• UMD data server — companion run without data assimilation (SODA4.15.2.0) — Direct HTTP(S) download No login needed
  • Folders ORIGINAL/ and REGRIDED/. Same forcing and bulk formulae with no assimilation, intended for closed-budget diagnostics.
• UMD data server — per-stream output and analysis increments (soda4.15.2_s1 to _s4) — Direct HTTP(S) download No login needed
  • The per-stream folders named in the preprint's data availability statement (e.g. soda4.15.2_s1/ with ORIGINAL/, REGRIDED/ and SODA/). The portal's analysis-increment link (DATA/soda4.15.2/SODA) pointed to a path not present in the server listing on 2026-10-12.

Example code¶

Downloads one 0.1° regridded 5-day file from the UMD server and opens it with xarray. Ice files are tens of MB; ocean files about 1.6 GB.

import urllib.request
import xarray as xr

base = "https://dsrs.atmos.umd.edu/DATA/SODA4.15.2_REG/"
fname = "soda4.15.2_5dy_ice_reg_2024_06_25.nc"         # 5-day sea-ice mean (small file)
urllib.request.urlretrieve(base + "ice/" + fname, fname)

ds = xr.open_dataset(fname)
print(ds)            # check variable names (cn/aice etc.) and the time coordinate first
print(ds["hi"])      # sea ice thickness (m)

# 5-day ocean mean (~1.6 GB): base + "ocean/soda4.15.2_5dy_ocean_reg_2024_06_20.nc"
# Check how deep the depth coordinate goes (~1017 m before 2005).

Practical tips¶

• Regridded files may not cover the full depth. The preprint states that regridded state variables span only 0–1017 m before 2005 and 0–2006 m from 2005, so for deep-ocean work use the original grid or check the depth coordinate first.
• Supplement Table S6 lists original-grid files that are permanently missing: 1980–1982 and some 5-day and 10-day files (1986-06-15, 1993-12-23, part of 1994-12, 2000-02-10, part of 2009-01, part of 2016-07, 2021-09-27, among others). The regridded 5-day listing also lacks 1980-07-21 and 07-26, so check date continuity before building time series.
• The record joins four roughly decade-long streams (starting around 1980, 1990, 2000 and 2010), and the authors note it is "not quite continuous". Treat trends and change points that straddle 1990, 2000 or 2010 with care.
• In the version name, 15 stands for ERA5 forcing and the final 2 for the COARE bulk formulae. SODA4.15.2.0 (with the extra .0) is the run without data assimilation; do not mix it with the reanalysis.
• The original grid is tripolar, with the Arctic cap split into two poles over Eurasia and North America, and its velocities are grid-oriented rather than east/north. Use the regridded files for maps and regional means, and the original grid for section transports.
• The full dataset is about 46 TB. One year of monthly ocean means alone is 17–19 GB regridded and about 65 GB on the original grid, so decide on years and variables first and subset by region and depth right after download.
• Known limitations from the preprint: an artificial zonal temperature gradient near the equator from slightly weak ERA5 trade winds; 300–1000 m salinity differing from EN4.2.2 by about 0.04 PSU; insufficient resolution and no Argo in the Arctic; and Fram Strait inflow and outflow about 20% smaller than observational estimates.

Notices and change history¶

References for your paper¶

Reference papers¶

SODA4 reference paper (citation requested by the SODA project; EGUsphere preprint) — DOI 10.5194/egusphere-2025-3810

@article{chepurin2025soda4,
  author  = {Chepurin, Gennady A. and Carton, James A. and Sun, Luyu and Penny, Stephen G.},
  title   = {{SODA4}: a mesoscale ocean/sea ice reanalysis 1980--2024},
  journal = {EGUsphere [preprint]},
  year    = {2025},
  doi     = {10.5194/egusphere-2025-3810}
}

SODA3 reference paper (basis for SODA4's initial conditions and assimilation) — DOI 10.1175/JCLI-D-18-0149.1

@article{carton2018soda3,
  author  = {Carton, James A. and Chepurin, Gennady A. and Chen, Ligang},
  title   = {{SODA3}: A New Ocean Climate Reanalysis},
  journal = {Journal of Climate},
  volume  = {31},
  number  = {17},
  pages   = {6967--6983},
  year    = {2018},
  doi     = {10.1175/JCLI-D-18-0149.1}
}

Acknowledgements¶

The SODA project asks users to state the version used (e.g. SODA4.15.2) and cite Chepurin et al. (2025). The UMD server data have no dataset DOI. If the preprint is published in a journal the citation will change, so check the EGUsphere page for the final paper before submitting. Model and assimilation code are archived on Zenodo (doi:10.5281/zenodo.16878109, CC BY 4.0).

Information on the datasets used in this study was compiled with the help of the Korean Climate Data Navigator (KCDN) (Sim, 2026; https://doi.org/10.5281/zenodo.23300923).

Links¶

• SODA project home page
• SODA4 portal
• SODA4 README
• Chepurin et al. (2025) SODA4 preprint (EGUsphere)
• SODA4 model and assimilation code (GitHub UMD-AOSC/soda4)
• SODA heat content data (SODA3 and SODA4)
• UMD data server listing