Morice et al. (2025) An observational record of global gridded near-surface air temperature change over land and ocean from 1781

Identification

• Journal: Earth system science data
• Year: 2025
• Date: 2025-12-15
• Authors: Colin Morice, David I. Berry, Richard Cornes, Kevin Cowtan, Thomas Cropper, Ed Hawkins, John Kennedy, Timothy J. Osborn, Nick A. Rayner, Beatriz Recinos, Andrew Schurer, Michael Taylor, Praveen Teleti, Emily J. Wallis, Jonathan Winn, Elizabeth C. Kent
• DOI: 10.5194/essd-17-7079-2025

Research Groups

• Met Office, Exeter, United Kingdom
• National Oceanography Centre, Southampton, United Kingdom
• Department of Chemistry, University of York, United Kingdom
• Department of Meteorology, National Centre for Atmospheric Science, University of Reading, United Kingdom
• Climatic Research Unit, School of Environmental Sciences, University of East Anglia, United Kingdom
• School of Geosciences, University of Edinburgh, United Kingdom

Short Summary

This study introduces GloSATref, a novel global gridded surface air temperature (GSAT) dataset combining land surface air temperature (LSAT) and marine air temperature (MAT) observations, extending the instrumental record back to 1781. It demonstrates that using MAT, with new bias adjustments, allows for an earlier and more comprehensive reconstruction of global temperature changes compared to traditional sea surface temperature (SST)-based datasets.

Objective

• To develop and present GloSATref, a new global gridded surface air temperature (GSAT) anomaly dataset that extends the instrumental record back to 1781 by uniquely combining land surface air temperature (LSAT) and marine air temperature (MAT) observations, rather than the commonly used sea surface temperature (SST). This aims to provide a reference for comparing observed GMST and GSAT, improve understanding of early instrumental climate, and sample structural uncertainty.

Study Configuration

• Spatial Scale: Global (land and ocean), gridded. Analyses cover Northern Hemisphere, Southern Hemisphere, and specific oceanic and continental regions. Gridding uses a 1° latitude × 1° longitude resolution for climatology generation, with spatial decorrelation length scales of 1300 km for LSAT and 1550 km for MAT in the Gaussian process model.
• Temporal Scale: 1781 to present for land-only analysis (GloSATLAT), and 1784 to present for combined land and ocean analysis (GloSATref).

Methodology and Data

• Models used:
  • GloSAT reference analysis (GloSATref), GloSATLAT, GloSATMAT
  • Extended CRUTEM5 station database (for LSAT)
  • Local-expectation Kriging (LEK) method (for estimating missing station climatological normals)
  • Regression-based predictive models for exposure bias in land observations (Wallis et al., 2024)
  • Berry et al. (2004) heating bias model (for MAT diurnal bias adjustment)
  • Gaussian process analysis system (Morice et al., 2012, 2021) for gridding and interpolation, utilizing a Matérn covariance function.
  • Ensemble gridding procedure (Morice et al., 2012)
  • Monte Carlo approach for propagating height adjustment uncertainty in MAT.
• Data sources:
  • International Comprehensive Ocean-Atmosphere Data Set (ICOADS) (principal source for marine observations)
  • Citizen science data digitization (Old Weather initiative)
  • Ship-based Automated Meteorological and Oceanographic System (SAMOS) archive
  • Voluntary Observing Ship Global Data Assembly Centre (VOS-GDAC) data
  • WMO Publication 47 (for MAT measurement height metadata)
  • Various national and international meteorological archives for land stations, including CLIMAT, BoM, Environment Canada, GHCN-Daily, NCEI World Weather Records (WWR), NCEI Monthly Climatological Data for the World (MCDW), and others.

Main Results

• GloSATref extends the global instrumental temperature record back to 1784 (land-only to 1781), providing nearly 70 additional years of data compared to existing SST-based datasets.
• The dataset incorporates new adjustments for diurnal-heating biases in marine air temperatures, enabling the use of daytime observations, and bias adjustments for non-standard thermometer enclosures over land.
• GloSATref global and hemispheric temperature anomaly series show close agreement with SST-based HadCRUT5 for much of their overlapping period (1850-present) but with slightly less overall warming.
• The pre-1850 record in GloSATref captures strong cooling associated with major volcanic eruptions (e.g., Laki in 1783–1784, Tambora in 1815).
• GloSATref is, on average, warmer than HadCRUT5 from 1850 to 1880, and then notably cooler until the early 1910s, especially in the Southern Hemisphere.
• GloSATMAT shows cold temperature anomalies relative to HadSST4 and NMAT-based estimates during World War 2, suggesting improved bias adjustments.
• The smaller warming trend in MAT relative to SST data, starting in the late 1950s, is apparent in GloSATMAT.
• The use of Local-expectation Kriging (LEK) increased the number of usable land stations from 7983 in CRUTEM5 to 11134 in GloSATref.
• Exposure bias adjustments cool LSAT global annual averages in the 19th and early 20th centuries by less than 0.1 °C.
• Prior to 1850, GloSATref shows greater variability than the PAGES2k median palaeoclimate reconstruction, likely due to reduced global measurement sampling and strong volcanic activity.

Contributions

• First global gridded surface air temperature (GSAT) anomaly dataset to use marine air temperature (MAT) instead of sea surface temperature (SST) in combination with land surface air temperature (LSAT).
• Extends the global instrumental temperature record back to 1784 (land-only to 1781), providing nearly 70 additional years of data compared to existing SST-based datasets.
• Introduces novel adjustments for diurnal-heating biases in MAT, enabling the use of all-hour marine observations and extending the marine record further into the past.
• Applies new bias adjustments for non-standard thermometer enclosures in early land observations, improving the quality of the early LSAT record.
• Provides a new reference dataset for comparing observed Global Mean Surface Temperature (GMST) and Global Surface Air Temperature (GSAT), aiding in resolving disagreements between climate models and observational records.
• Offers an additional line of evidence for understanding global near-surface temperature change and variability, sampling a different dimension of structural uncertainty.
• Enhances the CRUTEM5 land station database with additional data and improved climatological normal estimation using Local-expectation Kriging (LEK).

Funding

• Natural Environment Research Council (grant nos. NE/S015647/2, NE/S015582/1, NE/S015566/1, NE/S015698/1, NE/S015574/1, NE/R015953/1, and NE/Y005589/1)
• Met Office Hadley Centre Climate Programme funded by DSIT (Department for Science, Innovation and Technology)

Citation

@article{Morice2025observational,
  author = {Morice, Colin and Berry, David I. and Cornes, Richard and Cowtan, Kevin and Cropper, Thomas and Hawkins, Ed and Kennedy, John and Osborn, Timothy J. and Rayner, Nick A. and Recinos, Beatriz and Schurer, Andrew and Taylor, Michael and Teleti, Praveen and Wallis, Emily J. and Winn, Jonathan and Kent, Elizabeth C.},
  title = {An observational record of global gridded near-surface air temperature change over land and ocean from 1781},
  journal = {Earth system science data},
  year = {2025},
  doi = {10.5194/essd-17-7079-2025},
  url = {https://doi.org/10.5194/essd-17-7079-2025}
}