HadSST4¶

Met Office Hadley Centre sea surface temperature data set, version HadSST.4.2.0.0

Available

At a glance¶

Note on period: Updated monthly. On 2026-10-12 the download page showed "Last updated: 15/09/2026" and the annual time series ran through 2025; the most recent month was not checked directly. Observations are very sparse in the 19th century, during the two world wars and in the Southern Ocean, so many grid boxes are empty.

Overview¶

A monthly global field of sea surface temperature (SST) anomalies on a 5° latitude by 5° longitude grid from 1850 to the present, built by the Met Office Hadley Centre from ICOADS ship and buoy observations and Copernicus Marine buoy data. The data are adjusted to reduce the effects of changes in instrumentation (buckets, engine room intakes, buoys and so on) and are presented as a set of 200 interchangeable realisations that sample the bias adjustment uncertainty, with measurement, sampling and coverage uncertainties supplied separately. The data are not interpolated: unlike gap-free analyses such as HadISST, ERSST or COBE-SST, empty grid boxes stay empty. HadSST4 is the ocean component of global temperature datasets such as HadCRUT5 and is widely used to study long-term SST change.

What changes with the successor¶

HadSST.4.1.1.0 → HadSST.4.2.0.0¶

Transition: 2025-09

A minor science release. Refinements to the pre-1950 engine room intake (ERI) bias estimates reduce the anomalous warmth during the Second World War; it also adds truncation bias corrections for the Japanese Kobe Collection and fixes a bug in the uncertainty calculation. A new version means the full dataset must be downloaded again.

Sources: HadSST.4.2.0.0 Product User Guide (September 2025) · Sandford & Rayner (2026), International Journal of Climatology

Available variables¶

Based on the HadSST.4.2.0.0 Product User Guide and the download page. For files whose variable names are not given in the guide, only the file name is listed.

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

SST anomalies (bias-adjusted)¶

Uncertainties¶

Observation counts¶

The full variable list is in the official documentation.

How to get the data¶

• Met Office Hadley Centre HadSST4 download page — Web download (NetCDF / zip / CSV) No login needed
  • Provides the median file (HadSST.4.2.0.0_median.nc), the ensemble (four zips of 50 members), uncertainty and error covariance files, observation counts, and monthly and annual time series (CSV) for the globe, the Northern and Southern Hemispheres and the tropics. No account needed; MD5 checksums are listed.
• Previous versions (HadSST.4.1.1.0, HadSST.4.0.1.0) — Web download No login needed
  • 4.0.1.0 ended in December 2024 and 4.1.1.0 was to be updated monthly only until the end of 2025. Use the latest version for new analyses.

Example code¶

Downloads the median (anomaly) file and computes the Niño 3.4 area mean from observed grid boxes only; empty boxes are not filled.

# pip install xarray netCDF4 numpy requests
import numpy as np
import requests
import xarray as xr

url = "https://www.metoffice.gov.uk/hadobs/hadsst4/data/data/HadSST.4.2.0.0_median.nc"
with open("HadSST.4.2.0.0_median.nc", "wb") as f:
    f.write(requests.get(url, timeout=600).content)

ds = xr.open_dataset("HadSST.4.2.0.0_median.nc")
tos = ds["tos"]                                   # anomaly vs 1961–1990 (K); NaN where unobserved
box = tos.where((ds.latitude >= -5) & (ds.latitude <= 5) &
                (ds.longitude >= -170) & (ds.longitude <= -120), drop=True)
w = np.cos(np.deg2rad(box.latitude))
nino34 = box.weighted(w).mean(["latitude", "longitude"])   # NaN boxes are skipped automatically
n_obs = box.notnull().sum(["latitude", "longitude"])       # also check how many boxes went into the mean
print(nino34.sel(time="2025").values, n_obs.sel(time="2025").values)

Practical tips¶

• HadSST4 is not infilled. The set of grid boxes behind an area mean changes from month to month, so always look at the number of contributing boxes, and apply the same coverage mask before comparing with gap-free analyses such as HadISST, ERSST or OISST. Area means carry a coverage uncertainty.
• To get the bias adjustment uncertainty, compute your area mean for each of the 200 members and use the spread across members; the distribution is not Gaussian, so inspect it first. The guide treats the four components — uncorrelated measurement and sampling, correlated measurement (covariance), bias and coverage — as independent and combines them as a root sum of squares.
• Anomalies are relative to 1961–1990. To compare with anomalies on a 1991–2020 baseline, subtract the mean over a common period first.
• HadSST.4.2.0.0 changed 1941–1945 values substantially. Re-check any result based on versions before September 2025 (for example, the wartime warm anomaly) against the new version.
• HadISST is a separate product that uses HadSST4's unadjusted in-situ observations as input (from January 2021 data), and HadSST4 itself contains no sea ice information. For gap-free SST or sea ice concentration, use HadISST or another analysis.

Notices and change history¶

References for your paper¶

Reference papers¶

HadSST4 reference paper (required citation) — DOI 10.1029/2018JD029867

@article{kennedy2019hadsst4,
  author  = {Kennedy, J. J. and Rayner, N. A. and Atkinson, C. P. and Killick, R. E.},
  title   = {An Ensemble Data Set of Sea Surface Temperature Change From 1850: The {Met Office Hadley Centre} {HadSST.4.0.0.0} Data Set},
  journal = {Journal of Geophysical Research: Atmospheres},
  volume  = {124},
  number  = {14},
  pages   = {7719--7763},
  year    = {2019},
  doi     = {10.1029/2018JD029867}
}

HadSST.4.2.0.0 update paper (Second World War warm anomaly) — DOI 10.1002/joc.70388

@article{sandford2026hadsst42,
  author  = {Sandford, Caroline and Rayner, Nick},
  title   = {Addressing the World War 2 Warm Anomaly in {HadSST.4.2.0.0}},
  journal = {International Journal of Climatology},
  volume  = {46},
  number  = {8},
  pages   = {e70388},
  year    = {2026},
  doi     = {10.1002/joc.70388}
}

Acknowledgements¶

The terms and conditions require a credit line such as "HadSST.4.2.0.0 data were obtained from http://www.metoffice.gov.uk/hadobs/hadsst4 on [date downloaded] and are © British Crown Copyright, Met Office [year of first publication], provided under an Open Government Licence, https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/" (the Product User Guide gives the year as the year of download — copy the exact wording from the terms and conditions page). Both Kennedy et al. (2019) and Sandford & Rayner (2026) must be cited.

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¶

• HadSST4 home page (Met Office Hadley Centre)
• HadSST4 download page
• HadSST.4.2.0.0 Product User Guide (PDF)
• HadSST4 previous versions
• HadSST4 terms and conditions
• HadSST.4.2.0.0 diagnostic diagrams (observation counts, biases, uncertainties)