Skip to main content
Geosciences LibreTexts

4.1: Measurements of temperature over the past 150 years: How good are they?

  • Page ID
  • Let’s begin with the instrumental record of global average surface temperature. Thermometers were invented in the 17th century, but it was not until the 19th century that people started to make systematic, quantitative measurements around the globe. Naturally, most of these were made from land-based stations, but it was not long before measurements were being taken of the temperature of ocean water at and near the surface. (Benjamin Franklin discovered the Gulf Stream by lowering a thermometer into the ocean from a ship.) Sea surface temperature was measured routinely from buckets of water retrieved from the sea, and then, beginning in the 1960s, by taking the temperature of engine intake water. By the late 1960s, these measurements were being augmented by satellite-based measurements of infrared radiation emitted from the sea surface.

    Decadal Land-Surface Average Temperature

    Figure \(\PageIndex{1}\): 10-year moving average of the global average temperature over land from 1750 to 2012. The blue curve is from the NASA Goddard Institute for Space Studies; the green, from NOAA’s National Climatic Data Center; the red, from the United Kingdom Hadley Center’s Climate Research Unit; and the black curve with gray uncertainty bounds, from the University of California’s Berkeley Earth Project.

    In estimating global mean temperature, one must carefully account for the uneven distribution of temperature measurements around the world, changes in the precise location and instruments used to measure temperature, the effects of growing urban areas that create heat islands that are warmer than the surrounding countryside, and myriad other issues that can bias global mean temperature. Different groups around the world have tackled these issues in different ways, and one way to assess the robustness of the temperature record is to compare their different results, as shown in Figure \(\PageIndex{1}\). The Berkeley Earth estimate, shown in black with gray uncertainty bounds, was undertaken by a group led by a physicist who was skeptical of the way atmospheric scientists had made their estimates. Even so, the four records agree with each other quite well after about 1900 and especially well after about 1950. The better and better agreement reflects the increasing number and quality of temperature measurements around the planet.

    Glacial Cycles in Temperature and Ice Volume

    Figure \(\PageIndex{2}\): World map showing surface temperature trends (°C per decade) between 1950 and 2014. Source: NASA GISS. Also see

    Theory and models predict that the air over land and at high latitudes should warm faster than that over the oceans, and this is observed (Figure 4). Global warming is neither predicted nor observed to be globally uniform, and there are even places where the temperature has dropped over the second half of the 20th century, thanks to changing ocean circulation, melting sea ice, and other processes. Note also in Figure \(\PageIndex{2}\) that some of the fastest warming is in places far removed from cities, like Siberia and northern Canada; in fact, at most 2%–4% of warming can be attributed to urbanization.*

    So the measurements that underlie Figure \(\PageIndex{2}\) are pretty accurate. But how does that record of temperature and CO2 fit with the longer-term climate record? Is it unusual or is it consistent with natural climate variability on 100-year time scales? Since we do not have good global temperature measurements before the 19th century we must turn to the fascinating field of paleoclimate, which seeks proxies for climate variables in the geologic record.


    Jacobson and Ten Hoeve, 2012: Effects of urban surfaces and white roofs on global and regional climate. J. Climate 25: 1028–1043.