GloSAT Research

GloSAT has developed a longer global surface temperature climate record: underpinning this record is a wide range of data rescue and research to improve the marine and land observations. Alongside the observation-based research GloSAT has produced and analysed a small ensemble of earth system model runs covering the period since 1750.

1. Marine Data

GloSAT capitalised on the longer record of near surface marine air temperature measurements to extend the observation-based record back to the 1790s. To improve consistency in the measured air temperatures over time three main sources of inhomogeneity in the data need to be addressed, illustrated in Figure 1.

Figure 1: Illustration of some sources of bias in observations of marine air temperature. Figure produced for the GloSAT project (Kennedy, J., Kent, E., & Becker, A. (2025). Illustration of some sources of bias in observations of marine air temperature (Version 2). figshare. doi: 0.6084/m9.figshare.28046009)

• Changing measurement height
  • Typically the sea surface is warmer than the air above so there is temperature gradient near the surface. Under these conditions, observations made nearer the surface will be warmer than those made higher up. The size of ships, and therefore the height of measurements has increased since the late 18th century so we apply an adjustment to account for changing measurement heights. This adjustment acts to increase our estimate of how much global warming has occurred: early in the record the temperatures are reduced and later temperatures made higher in the atmosphere are increased. The adjustment methodology is based on Kent et al. (2013) and Cornes et al. (2020).
• Daytime heating bias
  • During the day the ship heats up as it absorbs incoming solar radiation. The amount of warming depends on the characteristics of the ship and the location of the sensor, and is decreased when there is strong air-flow over the sensor. This spurious diurnal variation is in addition to the normal daytime variations in temperature, and for most ships is larger than the real daily variations. GloSAT estimated the effect of the spurious warming signal ship-by-ship using a model developed by Berry et al. (2004). Implementing this adjustment means that the GloSAT dataset starts in the 1790s, nearly 100 years earlier than other datasets based on marine air temperature, and 60 years earlier than SST-based estimates of global surface temperature.
• The effects of non-standard observing practices
  • During World War 2 observations of both marine air temperatures and sea surface temperatures are too warm. The main cause of the bias is a shift in the time of day that observations were made. More observations were made in day-light hours to avoid the use of lights at night. Similarly any observations made at night will likely have been made in very sheltered locations that may have been warmer than more exposed locations.  

Associated publications and data

2. Land Data

As for marine data observations of air temperature made over land also need to have adjustments applied to improve their consistency over time. Figure 2 illustrates some of the issues that can affect the exposure of the measurement sensor and hence the quality of the observations. The adjustments to remove biases in the land observations caused by the transition from historical screens to Stevenson screens were developed by Wallis et al. (2024).

Further improvements to the GloSAT air temperature over land observations were required to estimate a climatology (normal) for stations where there were no observations during the climatological period. Missing station normals were estimated by using Kriging interpolation (Taylor et al., 2025, in revision) and allowed the use of observations that would otherwise have been excluded from the GloSAT analysis.

Figure 2: Illustration of how changing thermometer screen designs affected measured temperatures historically. Figure produced for the GloSAT project (Kennedy, J., Kent, E., Wallis, E., Osborn, T. J., & Becker, A. (2025). Illustration of how changing thermometer screen designs affected measured temperatures historically. (Version 2). figshare. doi: 10.6084/m9.figshare.28045271).

Associated publications and data

3. Generating a global climate record

The air temperature observations were aggregated on monthly 5x5 degree grids, separately for marine data (following Cornes et al. 2020) and for measurements over land (following Osborn et al. 2021). These are then combined using the same methodology as used for HadCRUT5 (Morice et al. 2020) to generate the GloSAT dataset, known as the reference dataset (GloSATref).

Associated publications and data

4. Models and analysis

GloSAT also funded an ensemble of runs of the joint Natural Environment Research Council (NERC) and Met Office climate model global climate model UKESM1, but starting in 1750 rather than 1850, the usual start data for CMIP historical simulation experiments used extensively in IPCC reports. This was to enable a comparison with the early part of the GloSATref dataset using an "IPCC-class" model. These climate model runs were analysed to provide a greater understanding of climate variability and change over the full industrial period, investigating the effect of a period of pronounced strong volcanic activity and quantifying the anthropogenic influence during the early industrial period.

Associated publications and data