Close window
Close window
This section of the new site isn't ready yet. We've brought you back to the current site.

An overview of global surface temperatures in 2018

February 2019 - Global temperature data from centres around the world show that 2018, together with the previous three years, constitute the four warmest years on record.

Summary

  • 2015, 16, 17 and 18 are the four warmest years on record in all surface temperature data sets.
  • Global mean surface temperatures in 2018 were 4th or joint 3rd warmest on record in all data sets surveyed
  • There is reasonable agreement between data sets in those regions where the data sets overlap, although fundamental choices in the definition of global temperatures are now making a measurable difference to assessments of global average temperature.
  • Differences between the data sets in 2015-2018 are due to how data gaps are dealt with, particularly the Arctic, the treatment of systematic errors in sea-surface temperatures and how sea-ice covered areas are dealt with.
  • The best estimates from all data sets agree that the Arctic was exceptionally warm in 2016 and 2017, but less warm in 2018. The Antarctic was also unusually warm in 2017 and 2018.
  • The current run of HadCRUT4 does not have US land station data for December 2018. The impact of this is likely less then hundredth of a degree in the annual average.

Drivers of recent global temperatures

The long-term increase in global temperature since pre-industrial times is almost entirely due to human activities. However, the increase in global temperature has not been smooth, with periods when warming was faster or slower as well as year-to-year changes (Figure 1).

Figure 1 (top) Global annual average temperature anomalies (°C, relative to the long-term average for 1981-2010, 2018 to November). Six datasets are shown as indicated in the legend. The grey shaded area is the 95% uncertainty range for HadCRUT4. The horizontal grey line with yellow shading indicates the approximate point at which temperatures exceed 1°C above “pre-industrial” levels. The horizontal grey line is the difference from the 1850-1900 average. The yellow shading is derived from Hawkins et al. (2017).  (bottom) differences of each data set from HadCRUT4 on an expanded scale.

One of the main drivers of year to year temperature change is the El Niño Southern Oscillation (ENSO). Cold currents driven by the seasonal winds and fed by deep ocean waters run along the western coast of the Americas and along the equator. During warm El Niño events, the winds shift and the supply of cold water is cut off. As a result, sea-surface temperatures in the central to eastern Tropical Pacific rise. The heat released by El Niño events typically drives a short-lived spike in global temperatures, with the spike lagging a little behind the temperature changes in the Tropical Pacific. During cold La Niña events, the seasonal winds strengthen, the supply of colder water goes up and sea-surface temperatures drop. La Niña events are usually associated with cooler global average temperatures, but like El Niño the global effect usually trails a short way behind changes in the tropical Pacific.

2015 and 2016 saw weak El Niño conditions develop into a strong El Niño event – one of the three strongest of the past fifty years – which has since dissipated. The warming effect of the strong El Niño influenced global temperatures at the end of 2015 and the first half of 2016 (Figure 2). This El Niño influence is the main reason why 2015 and 2016 were clearly warmer than 2014. However, anthropogenic influences are necessary to explain the majority of the overall warming in the global temperature series. In 2017, there was a transition from neutral ENSO or weak La Niña conditions at the start of the year to La Niña conditions by the year’s end (Figure 3).

Figure 2: Monthly global average temperature anomalies (°C, relative to the long-term average for 1981-2010, 2018 to November/December depending on data set). Five datasets are shown as indicated in the legend.

Weak La Niña conditions continued into 2018. Over the course of the year, sea-surface temperatures in the eastern Tropical Pacific increased, nearing and even exceeding El Niño thresholds for a short period. However, the characteristic El Niño response of the atmosphere to warmer sea-surface temperatures was absent. There was no strong signal of ENSO influence on the global temperature of 2018.

Figure 3 Differences from the long-term average of (left) zonal wind (ms-1) i.e wind speed along the equator, (centre) depth at which water temperature falls to 20°C (m) and (right) sea-surface temperature (°C) across the Pacific between 5°S and 5°N. During El Niño events, the sea-surface in the eastern Tropical Pacific is warmer than average, winds blow more strongly from the west than usual and the layer of warm water near the surface thickens in the east. During La Niña the opposite occurs.

The Polar Regions also play an important part in understanding recent global temperatures. Sea-ice extent in the Arctic has been in a long-term decline since the satellite record began in 1979 and it remained low throughout 2016, 2017 and 2018 (Figure 4). Areas of the high-latitude seas which would typically have been ice-covered in the climatology period were open water in these years, a change that is often associated with increased temperatures. Although the area of the polar regions is only a small fraction of the Earth’s surface, temperature variations in these regions can be large and can have an important effect on global temperature. Temperature anomalies in the Arctic in late 2018 were high in the long-term context, but much lower than in 2017 and well below the exceptional year of 2016 (Figure 5). Long-term warming associated with greenhouse gases is expected to be greater at higher latitudes. Unlike the Arctic, long-term temperature changes in the Antarctic have been smaller relative to the year to year variability. However, temperatures in 2018 were at the upper end of the long-term record in a number of datasets.

Figure 4 Daily sea-ice extent (million km2) for the Arctic (top) and Antarctic (bottom) during the satellite record (1979-present). Individual years are shown in grey with 2018 highlighted in red. 2017 is highlighted in blue for the Antarctic. 2012 is highlighted in blue for the Arctic.

Figure 5 Annual average temperature anomalies (°C, relative to the 1961-1990 average) for the Polar Regions: (top) Arctic (North of 65°N) and (bottom) the Antarctic (South of 60°S). Note different temperature scales in the two plots. 2018 is plotted to November. The Antarctic series starts in 1958, which is the first year of long-term routine measurements in the continental interior.

Estimating global average temperature

Surface temperature data are gathered by a global network of weather stations, ships and buoys. This network measures air temperatures over land and sea-surface temperatures over the oceans. A number of groups use these data to produce global temperature data sets. The data are carefully processed to account for changing measurement methods and instrumentation, and for the uneven distribution of measurements around the world. The methods that the groups use are different and so the estimates of the global average that they produce are also slightly different.

In addition to traditional data sets based on surface-temperature measurements, there are atmospheric reanalyses, which use a much wider range of observations including satellite data in combination with a weather-forecasting model to produce a globally complete temperature analysis.

Figure 1 shows the estimated global annual average temperature differences from the long-term average, as calculated from six data sets: HadCRUT4 (produced by the Met Office Hadley Centre in collaboration with the Climatic Research Unit at the University of East Anglia), GISTEMP (produced by the Goddard Institute of Space Studies at NASA), NOAAGlobalTemp (produced by the National Centers for Environmental Information at NOAA), Berkeley Earth and Cowtan and Way (CW). It also shows global temperature series calculated from the ECMWF ERA-Interim reanalysis and the Japanese JRA-55 reanalysis.

In general, the agreement between the data sets is very good. They all display a similar increase in estimated global average temperature over time and the short ups and downs in global temperature of around 0.1 to 0.2°C that signify the pulse of El Niño and La Niña can be clearly seen in each of the data sets. However, some differences are to be noted, particularly in the 19th Century where data are sparse and in the mid 20th Century when there were large changes to the way that sea-surface temperatures were measured.

We are focusing here on differences in the period 2015-2018, which can be seen more clearly in Figure 6. For most data sets, 2018 is nominally the fourth warmest year on record. In the ERA-Interim data set 2018 is very close to 2015 and is effectively tied for third place.

Figure 6 Global annual average temperature anomalies (°C, relative to 1981-2010) for each of the data sets (as labelled) from 1970 to 2018. The heavy, dotted line shows the dataset in the label and the fine grey lines show the other data sets for comparison. Data sets in the top row use HadSST3 for their SST data set. Data sets in the middle row use ERSST (v4 for NOAAGlobalTemp and v5 for NASA GISTEMP). The bottom row shows the two reanalysis data sets. Horizontal lines are spaced in 0.2°C increments.

Why are there differences?

One major difference between the data sets is the way they deal with geographically-uneven sampling – there are more weather stations in the mid latitudes of the northern hemisphere and fewer in the tropics and near the poles. There is a difference in the degree of sophistication with which they attempt to fill gaps in the station network. All the data sets perform some degree of interpolation and the reanalyses are globally complete by virtue of using a global weather forecasting model. Of the traditional data sets GISTEMP, Berkeley and CW data sets do the most interpolation. HadCRUT4 and NOAAGlobalTemp estimates do less and when calculating a global average, they implicitly fill the gaps with the average for the rest of the world. Most importantly for the discussion here, they do not interpolate extensively into the Polar Regions which have been particularly warm – at both ends of the Earth – in the past three years (Figure 5).

In 2018, the Arctic was exceptionally mild during the Northern Hemisphere winter and autumn, but closer to average during the summer. Figure 5 shows that annual temperature anomalies for areas north of the Arctic Circle (approximately 65°N) were nominally the highest on record in 2016, with temperatures around 2.0 to 3.5°C above the 1961-1990 average. In 2018, Arctic temperatures were below those of 2016 but still unusually high in the longer-term context, exceeding anything recorded prior to 2005. The Arctic average was considerably higher than the average for the rest of the globe so the data sets with sparser coverage in the Arctic will – all else being equal – tend to underestimate the global average temperatures in recent years including 2018. Unusual temperatures were also recorded at the other end of the Earth in 2018 with a particularly warm year recorded in the Antarctic.

Temperature anomalies in the Arctic exceeded 5°C in January to April 2018, with some areas exceeding 10°C above the long-term average (in February in particular). Temperature departures in other parts of the world were generally less extreme. This combination was sampled differently by each data set and it led to a large divergence in estimates of the global average between late 2017 and April 2018 (Figure 2).

How much difference does this make?

We can explore the size of the “coverage effect” by successively reducing the global coverage of all the data sets to that of HadCRUT4, which performs the least extensive interpolation. In Figure 7a, we display the global average anomalies estimated from each of the data sets with their full coverage after first converting them to the same grid as HadCRUT4 – the same basic pattern can be observed as in Figure 1. Differences (Figure 7b) between global averages calculated from the data sets and HadCRUT4 exceed 0.05°C (the shaded region shows the ±0.05°C range) in a number of years, including 2016, 2017 and 2018.

Next, we reduce the coverage of each of the datasets to be the same as HadCRUT4 north of 60°N and south of 60°S (Figure 7c). This has the effect of reducing the differences (Figure 7d) such that even the largest differences between the five traditional data sets do not exceed 0.05°C with the exception of GISTEMP. Finally, we reduce the coverage to be equal to that of HadCRUT4 across the whole globe (Figures 7e and 7f) which makes only a small additional difference.

Figure 7 (a) Global average temperature anomalies for the five global temperature data sets at full coverage (HadCRUT4 black, Cowtan and Way red, GISTEMP pale blue, NOAAGlobalTemp orange, Berkeley Earth dark blue) °C relative to the 1961-1990 average. Colours are as in Figure 1. (b) Differences between individual data sets and HadCRUT4. The grey shading is a visual aid to indicate the ±0.05°C range in the difference plots, which is the approximate magnitude of the uncertainties arising from sources other than coverage. (c) Global average temperature anomalies calculated from data sets matched to HadCRUT4 coverage in the Polar Regions. (d) Differences between individual data sets and HadCRUT4 after data matched to HadCRUT4 coverage in Polar Regions. (e) Global average temperature anomalies calculated from data sets matched to HadCRUT4 coverage everywhere. (f) Differences between individual data sets and HadCRUT4 after data matched to HadCRUT4 coverage everywhere. 2018 is plotted to November.

The differences since 2016 are larger than for other years since 1990. There are three likely components to this whose contributions are not easy to separate. The first is that NOAA adjust their data so as to remove an artificial drop in sea-surface temperatures measured by ships since the early 2000s. The adjusted sea-surface temperature data are used in both NOAAGlobalTemp (using version 4) and GISTEMP (using version 5). This adjustment is not applied in the Met Office Hadley Centre analysis, HadSST3, which is used in HadCRUT4, CW and the Berkeley Earth analysis and could explain some of the differences. Differences in global sea-surface temperatures amount to around 0.03°C since 2005 and the effect will be proportionately smaller when combined with land temperatures to calculate the overall global average.

The second reason is related to the problems associated with polar observational coverage. Some analyses – GISTEMP and Berkeley Earth – use air temperature anomalies across regions that are usually covered by sea ice, others – NOAA and HadCRUT4 - use sea-surface temperature anomalies. In the past three years, air temperature anomalies estimated for these regions have typically been higher than the coincident sea-surface temperature anomalies (though not at all times, in some months and places it was colder) and, as the ice retreats, the size of this area increases. Recent research (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2015GL064888) has shown that simulated global temperatures calculated using a blend of air temperature over land and sea-surface temperatures over ocean areas warms more slowly than a blend of air temperatures across both surfaces. In effect, different data sets are defining global temperature in slightly different, but now consequential, ways.

The third reason is that there are always small differences between data sets due to the processing applied. The algorithms used to homogenise observations from different sources (for example the adjustments for the previously mentioned ship and buoy biases) and correct for non-climatic changes in weather station measurements, grid the data, blend land and sea observations and calculate the global average vary from one dataset to another.

US Data in December 2018

Due to the late arrival of US land station data for December 2018, HadCRUT4 was updated without it. With modern data coverage, the expected effect of this gap in coverage on the annual average – which can be estimated by masking US data from previous Decembers – is around 0.003°C.

Last updated:

Follow us on

Facebook, Twitter, YouTube, Instagram, Snapchat, or LinkedIn Facebook Follow @metoffice on Twitter YouTube Instagram Snapchat LinkedIn