New estimates of surface temperature change since the late 18th century
Releasing an observation-based global temperature dataset extending back to 1781
Two new papers have pushed back the start of the historical period that we can examine using both measurements of surface temperature and simulations with climate models. These analyses shed light on the causes of global surface temperature variations back to the 18th century and put human-caused climate change into a longer 240-year context.
Both studies are outcomes of the NERC-funded GloSAT project, which has produced: (i) a new dataset of near-surface air temperature change across the globe since the late 18th century, and (ii) performed a series of simulations using the UK’s Earth System Model (UKESM) that start in 1750 rather than 1850.
A new dataset of air temperature change since the late 18th century
The first of these papers presents a new dataset of near-surface air temperature change across the globe since the late 18th century. This new dataset (called the GloSAT reference analysis, or GloSATref for short) extends the instrumental global temperature record back to the year 1781 through use of marine air temperature measurements taken on board ships, alongside weather station measurements over land. This advance is made possible through improved understanding of measurement uncertainties affecting both land and ocean temperature measurements.
Global near-surface temperature is an important measure of climate change, used to chart our changing climate and evaluate the performance of climate models. Typically, near-surface temperature changes are represented as a combination of air temperatures measured at land weather stations with near surface water temperatures measured at sea. Global temperature series based on this approach typically record temperature change starting from the 1850s.
GloSAT takes a different approach, combining weather station data with measurements of marine air temperature (MAT) rather than sea surface temperature (SST). SST observations only became widely measured from the 1850s onwards, fixing the earliest start date of records based on SST. However, many MAT measurements are available in earlier years and they are used by GloSAT to extend the observational record back to the 1780s.
To combine these data, the observations undergo rigorous quality control as well as alignment to adjust for changes in instrumentation and measurement practices. Anomalies are calculated as the difference between the adjusted measurements and typical conditions in the period 1961-1990. These temperature anomalies are then averaged into gridded maps from which changes in global and regional averages over time can be computed to track climate variability and long-term warming.
The marine air temperature (MAT) record
To create a consistent record over nearly 250 years, MAT measurements need to be adjusted to a standard reference height as ships have typically increased in size over this long period. This means that the height above sea level of the measurements has been increasing, which introduces an artificial cooling to the record which we remove. More complicated is the warming effect during the day in sunny conditions when the whole ship can warm up, potentially affecting measurements that are not taken in well-exposed locations.
GloSAT uses an energy balance model to estimate the daytime warming effect. The model is calibrated using available day and nighttime observations and uses cloud cover observations to estimate the incoming solar radiation and the wind speed at the sensor for the cooling effect of airflow. This gives a temperature record equivalent to what would have been recorded at night but based upon measurements taken at any time of the day or night. Spurious daytime warming effects have been removed and estimates of uncertainty resulting from the required adjustments are provided with the data.
Improved land air temperature estimates
The new dataset also introduces refinements to the temperature record over land that build upon the CRUTEM5 database of weather station temperature series. These improvements increase the number of usable stations and partially account for the effects of the early thermometer shelters that were used prior to the adoption of modern enclosures such as Stevenson screens.
The method used to create the land component of the gridded temperature anomaly dataset requires each weather station to have estimates of climatologically normal temperatures in a fixed reference period of 1961 to 1990. For GloSAT, we introduce a new method for calculating these climatological normals that uses information from nearby stations to estimate values when measurements are missing during the 1961 to 1990 period. This approach has increased the number of usable station series from 7,983 to 11,134.
The second advance accounts for changes in thermometer shelters in the early instrumental record. Comparison of parallel measurements made in both early and modern thermometer shelters indicates the presence of seasonal biases associated with early shelter types. For example, some early measurements were biased too warm during summer because the shelters did not fully shield the thermometer from sunshine or thermal radiation. Based on the results of parallel measurement studies and the recovery and digitisation of station meta data, GloSAT includes new adjustments for early instrumental observations.
Comparison with other near surface temperature datasets
The figure below compares the global average temperature from GloSATref with that from HadCRUT5. Because the two datasets use the same mapping and averaging methods to generate the global fields, this comparison highlights the differences due to the use of marine air temperature rather than SST, and the improvements to the treatment of the land observations. GloSATref shows broad agreement in global average temperature change with HadCRUT5 (and other major global temperature products - see paper) from 1850 onwards. Differences are most notable in the early 20th century and in recent decades, with the global air temperature trend in GloSAT slightly lower than that in datasets that use sea surface temperatures.
The extension of GloSATref before 1850 is much more variable than in the later period. This is due to a combination of sparse sampling and because it captures some cooling events caused by several large volcanic eruptions.
Climate simulation comparisons for the early instrumental period
While state-of-the-art climate simulations of the industrial era typically start in 1850, a second recent paper describes new climate simulations starting in 1750 using the UKESM1.1 Earth system model. The combination of these earlier starting climate simulations and the new observation dataset allows an examination of climate variability and change prior to 1850.
The effect of human influences on the climate during the pre-1850 period is found to be relatively small, with approximately 0.1°C of global warming, although the effects are much larger in some regions.
This period also includes the effects of a series of major volcanic eruptions in the early 1800s, such as the 1815 Tambora eruption. Both model simulations and observations indicate substantial cooling associated with major eruptions, with the new observation dataset indicating greater cooling than seen previously in sparse early weather station and climate proxy series.
Comparison between the new simulations starting in 1750 and equivalent simulations starting in 1850 reveals other differences that extend for many decades. These are understood to arise from ocean heat content and vegetation changes persisting from the initial conditions provided to the model, and from persistence in cooling from the early 19th century volcanic eruptions. These results indicate that an earlier start to historical simulations could lead to more representative climate simulations over the historical period.
Summary
The new GloSATref data set provides the first global record of near-surface air temperature changes from the late 18th century to present. This data set extends the historical climate record further back than previous products, allowing researchers to study climate variability and change in the early instrumental period prior to the mid-19th century. New model simulations of this period display a good agreement with this observational dataset and highlight the important role of volcanic eruptions during this period.
This work was undertaken under the NERC-funded GloSAT project in a collaboration led by the National Oceanography Centre, with the Met Office, University of Edinburgh, National Centre for Atmospheric Science, the University of East Anglia, University of York, University of Reading, and University of Southampton. For more information on GloSAT visit the project website: https://glosat.org/
 | A guest post by
|
 | A guest post by
|
 | A guest post by
|
|
 | A guest post by
|
|
|
The 6 new model simulations spanning the time period 1750-2014 which are described in this post are now available to download from here:
https://dx.doi.org/10.5285/9487085a6a3d4ca1aafa7c569e840d2a
Here's a convenient compilation of links to the various GloSATref archives including its land and marine air temperature components and a couple of visualisations:
https://crudata.uea.ac.uk/cru/data/glosat/