The chart that shows the global temperature record in one image

In September 2016 the University of Reading noted, in its own round-up of press coverage, that the latest visualisation from Dr Ed Hawkins of its Department of Meteorology, showing rising global temperatures, was being picked up by publications abroad. The entry is two lines long. It sits between a note about Arctic shipping routes and one about antibiotic resistance.

That is a modest paper trail for what turned out to be one of the more effective pieces of science communication of the decade. Hawkins is a climate scientist, not a designer, and the work in question was not an artwork commissioned to move people. It was an attempt to show a dataset without the usual apparatus getting in the way.

What the visualisations actually removed

A conventional temperature chart is a line on a pair of axes. It carries a great deal of information: exact values, a baseline, units, a time axis you can read off. It also asks the viewer to do work before anything registers, and most viewers do not do that work.

The visualisations that travelled did the opposite. They took information away. No y-axis. No legend. No numbers. In the version most people have now seen, each year of the record becomes a single vertical band of colour, cool for below the long-term average and warm for above it, laid out left to right in time order. There is nothing to read. There is only a sequence that starts blue and ends red.

It is worth being precise about the chronology, because the two are often conflated. The 2016 coverage concerned Hawkins’ animated spiral of global temperature, which circulated very widely that year. The now-ubiquitous banded version, and the campaign that spread it, followed in 2018. The idea underneath both is the same: strip the chart down until only the trend survives.

The same record, drawn from the published data

Below is that treatment applied to a specific, citable dataset rather than reproduced from anyone else’s graphic. Each band is one year of the NASA GISS Surface Temperature Analysis, version 4, land-ocean global mean, measured as an anomaly against the 1951 to 1980 average.

Two details about that image are worth stating plainly, because they are the kind of thing the original format deliberately hides.

The first is the range. The colour scale runs from the coldest year in the series to the warmest, so the visual span is fixed by the data, not chosen for effect. Here that span is 1.77 °C, from −0.49 in 1909 to +1.28 in 2024.

The second is the start date. This URL’s original headline referred to warming since 1850, which is the beginning of the Met Office Hadley Centre’s HadCRUT record. The GISTEMP series used here begins in 1880. Different long-run datasets start in different years because they depend on when usable instrumental coverage existed, and a chart with no axis label gives the reader no way to notice which one they are looking at. That is the cost of the format.

Decade averages, for the version with numbers

The same series, averaged by decade, in the form the banded chart omits:

Decade	Mean anomaly (°C)
1880s	−0.21
1900s	−0.32
1920s	−0.24
1940s	+0.04
1960s	−0.03
1980s	+0.25
1990s	+0.38
2000s	+0.59
2010s	+0.81
2020s (six years)	+1.06

Read as numbers, the record has a flat stretch in the middle of the twentieth century and then a steepening from the 1980s onward. Read as colour, that same structure appears as a long blue field, a pale hesitation, and then a run of reds that gets darker each decade. The numbers are more precise. The colours get looked at.

Why the format worked, and what it cannot do

The reason a banded chart spreads is that it needs no caption to be understood, which means it survives being screenshotted, reposted and printed on things. It transmits one claim, that the recent past is warmer than the earlier past, with almost no loss.

What it cannot do is anything quantitative. It will not tell you the size of the change, the uncertainty on early values, the difference between land and ocean coverage, or which dataset it came from. Those questions need the axes back.

Which is why both versions are useful, and why the Reading round-up filed it under communication rather than research. The science was already published. What changed in 2016 was how far outside the field it could travel.