Penn State NASA

Temperature: Instrumental Records


Temperature: Instrumental Records

Temperature is probably the most important observation regarding the global climate, but how to measure the temperature of something as large as the Earth is complex. Climate scientists have taken a variety of approaches to answering this question, depending on the timescale of interest; we’ll have a look at the results of these different approaches in this section.

The first approach is the most obvious — you use thermometer data; this is usually called the instrumental record of climate. Let’s take a look at what some of these data look like for a place familiar to the authors — State College, PA. These data come from the US Historical Climatology Network, where you can find data from stations around the US. These data are the monthly mean temperatures from 1849 to 1994, so there has already been some averaging of the data to remove the day-to-day variability.

Graph of monthly mean temperatures in State College, PA from 1849 to 1994
The thin blue line is the monthly mean temperature, which shows a great deal of variation related to the annual cycle of temperature change here — a bit more than 20°C.
Credit: David Bice © Penn State University is licensed under CC BY-NC-SA 4.0

What we are seeing in the above figure is weather, which is "noisy"; what we want is the climate record from this station, which is not obvious, but we will find it in the first lab exercise for this module. The data for this one station can give us a climate record for the immediate surroundings, but going from this record at one point to the global temperature requires a bit more work.

One approach is shown in the figure below. Say you have an array of weather stations on a map:

map showing different weather stations and the area they cover based on halfway points to the next closest weather stations
Schematic representation of a method to figure out the area represented by each weather station's temperature record. This is one of several different strategies.
Credit: David Bice © Penn State University is licensed under CC BY-NC-SA 4.0

The general approach is to draw lines between a station and all of its nearest neighbors, then find the midpoints of these lines (circles in the figure) then make a polygon that connects the circles, giving an area (in gray). This area (in km2) represents a tiny fraction, fi, of the Earth’s surface that is associated with the temperature, Ti, of this station. If you do this for all stations, and sum all the Ti x fi values from each station, you would have a global temperature (all of the fi values would add up to 1.00).

You can see from the above example that if you have good weather stations spread uniformly across the planet (land and sea) and they have been recording continuously for a long time, then one can take the mean annual temperature of each station and calculate a simple global average for each year, and thus the history of temperature change for our planet. But, as you might imagine, the stations are not uniformly distributed — they are clustered in populated countries — and the number of stations declines as you go backward in time, so the actual process of assembling an instrumental record takes some care. A variety of groups have done this using slightly different data sets and approaches. The trick here is in how you combine the individual temperature records to come up with a global average. This is complicated by the fact that some weather stations may have problems related to things like the "urban heat island effect." Man-made materials retain heat better than open land and the lack of trees also amplifies warming in cities, which are currently warming at double the rate of the global average! Thus, if urban development encroaches on a weather station, the urban heat island effect will make the local temperature rise for reasons that are unrelated to any regional climate change. Researchers have found ways of ensuring that this effect does not skew the results, and many different groups come up with results that are nearly identical, giving us confidence that the data analysis is sound. Just in case you are wondering, the mean surface temperature of the Earth as a whole is 15o C (59o F)!

There are a number of good estimates of the recent history of global temperature change, and they are shown, plotted at the same scale, in the figure below.

temperature reconstructions for the past 170 years from a various groups. all show increase in temp anomalies starting around 1975
Temperature reconstructions for the past 170 years from a variety of groups
Credit: Timothy Bralower © Penn State University is licensed under CC BY-NC-SA 4.0

The figure above shows anomalies relative to the mean for 1960-1980. GISTEMP is from NASA, CRUTEM4 is from the Climate Research Unit of the University of East Anglia in England, Berkeley is from the Berkeley Earth Surface Temperature project from the University of California, and NOAA is from NOAA (no surprise here). These different groups use essentially the same data, but they have slightly different approaches to selecting which data to use and how to convert the station data into global temperatures.

It is interesting to see how similar the curves are given that they use different strategies for averaging the data, and some of the records are based on slightly different sets of weather stations. In particular, note that none of these estimates show a general cooling trend over this length of time — they all show warming. Back in the 1800s, there were fewer weather stations, and so it is more difficult to estimate global temperature back then (see figure below), but it gets steadily better as time goes on, and for the last few decades, we have excellent data due to the satellites that now circle the globe taking temperature measurements of every spot on Earth (more on this in a bit). What we see in the figure above is a detail of the blade of the "Hockey Stick" — beginning about 1900, the temperature starts to rise, then it flattens out a bit in the 1950s and early 1960s, and then it increases again at a faster pace since that time. The total warming since 1900 is about 1.1°C as a global average.

Looking in more detail at the Berkeley temperature estimate, which is based on about 1.6 billion measurements, we can see that the uncertainty, indicated in the figure below by the green band surrounding the red line, gets progressively larger as we go back in time, but the uncertainty is practically zero for more recent decades.

Graph of Instrumental Temperature Record 1800-2011. Green in wider than red line in 1800s but narrows to match the red line around 1950
This figure shows the Berkeley temperature reconstruction with the uncertainty indicated by the green zone, with the global mean temperature (as an anomaly) shown in the thick red line.
Credit: David Bice © Penn State University is licensed under CC BY-NC-SA 4.0

This is a good point to explore a question about these records. Why does the annually averaged temperature rise and fall in such a complicated fashion? The sun does not vary in its brightness in such a dramatic fashion (the solar cycle related to sunspots can account for a global temperature variation of about a tenth of a degree), and the greenhouse gases that keep our planet warm do not vary in their concentration this much. Instead, it appears that a good deal of the variability seen in these records is related to things like volcanic eruptions and climate system oscillations like the El Niño – La Niña Southern Oscillation (ENSO), which is discussed in detail in Module 6. In short, ENSO is essentially a huge, sluggish, sloshing back and forth of warm water along the equator in the Pacific Ocean — it is like a wave that reflects back and forth between the two edges of the Pacific, and it has a global reach in terms of climate. During the El Niño phase of this oscillation, the warm water is pooled up on the eastern side of the equatorial Pacific and this has the effect of making the whole Earth warmer (the reasons for this are complex, but the effect is quite clear). Conversely, during the La Niña phase, the warm water is pooled up at the western edge of the equatorial Pacific and the whole globe tends to be cooler. The El Niño stage causes flooding rains in California, wet conditions in Florida (recommend you visit Disney during the La Niña!), but crippling drought in Australia and southern Africa.

surface temp 1980-2005. Increasing yearly mean, annual average. Monthy average spikes when it's el nino, and dips when its la nina
The last 25 years of globally averaged instrumental surface temperature measurements from the HadCRUT3 dataset.
Credit: Robert Rohde from Wikipedia, licensed under CC BY-SA 3.0

The above figure shows the last 25 years of globally averaged instrumental surface temperature measurements. Also shown is the recent history of fluctuations in ENSO and the period of atmospheric disturbance due to the eruption of Mount Pinatubo in the Philippines in 1991, one of the largest of the 20th century; the volcano injected ash and sulfur gases into the upper atmosphere, where they blocked enough sunlight to cool the global climate for a period of about 3 years.

Satellites offer another way of studying temperature changes and they are not subject to the same problems associated with weather station data — they provide a complete coverage of surface temperature on land and at sea. But, as can be seen below, there is a very good agreement between satellite measurements and the weather station data (NOAA surface in the figure below). The only problem is that the satellite data only go back to about 1980.

Graph comparing global instrumental temperature record to two versions of satellite data of lower atmospheric temps. See text below
Global Temperature Anomaly chart by year relative to 1979-2000
Credit: David Bice © Penn State University is licensed under CC BY-NC-SA 4.0

The figure above shows the global instrumental temperature record in blue (NASA GISTEMP) is compared to two versions of the microwave sounder satellite (MSS) data of lower atmospheric temperatures (UAH from Univ. of Alabama, Huntsville; RSS from Remote Sensing Systems, Inc.). The timing of the ups and downs in the satellite record are a near-perfect match with the instrumental record, but the magnitude of change is greater according to the satellite measurements. For comparison, we show the history of the El-Niño La-Niña oscillation and periods of volcanic eruptions that load the atmosphere with tiny particles of sulfuric gas that block sunlight and cool the planet. The eruption of Pinatubo in the Philippines had a big effect, and strong El-Niño periods lead to warming — together, these two variables (volcanoes, and El-Niño) along with small fluctuations in sunlight, account for the majority of the "noise" in these records. Another important point of this figure is that it confirms that the instrumental temperature record does a good job of representing what actually happened.

Next, we look at the spatial variations in the temperature over different spans of time.

Mean temperatures 1951-1980. Overall increase in temp, greatest in northern 1/3 of globe over land. small decrease over some of pacific
Temperature Anomalies comparing the 2023 Mean Temperatures versus 1951-1980
Credit: Berkeley Earth:, licensed under CC BY-4.0

The figure above shows the difference in instrumentally determined surface temperatures between the period January 1999 through December 2008 and "normal" temperatures at the same locations, defined to be the average over the interval January 1940 to December 1980. The average increase on this graph is 0.48 °C, and the widespread temperature increase is considered to be an aspect of global warming. The most striking feature of this map is that the temperature changes have not been uniform across the globe; the high latitudes (above about 50 degrees) in the Northern Hemisphere have warmed more than any other part of the Earth, while the tropics warmed far less. But Antarctica has been warming significantly too, and, most recently in 2022 there have been record temperatures 20 °C warmer than normal! 

We now turn our attention to the spatial pattern of temperature change over a much longer range of time — back to 1884. Below is an animation of the temperature change based on the instrumental record. It is worth remembering that the quality and quantity of the data get better and better as time goes on, so the early parts of this animation have more uncertainty connected to them.

The movie below is from NASA’s reconstruction of surface temperature since 1884, and it shows how Earth has warmed over the last century plus in a very, very graphic and indisputable way. Just in case you can't see this, 2016 was the warmest year on record, and 16 of the 17 warmest years have occurred since 2000!

Video: Global Warming: 1880-2021 (00:31) This video is not narrated.

Global Warming: 1880-2023
Credit: Global Warming from 1880 to 2020, NASA's Scientific Visualization Studio. Data provided by Robert B. Schmunk (NASA/GSFC/GISS) (Public Domain)

Click here if the video above does not play

Play this movie and watch as the globe becomes dominated by the yellow, orange, and red colors signifying warmer temperatures. Note that the warming is not uniform across the globe, nor is it steady through time, but the warming trend is nevertheless clear to see.

Another way of looking at this history of warming is by taking the average temperature at each latitude for each year and then stringing those along the horizontal axis, as below:

Graph of annual latitude averages relative to 1951-1980 mean. see text below
Latitude-averaged temperature anomalies from the GISTEMP data shown in the movie referred to in the last figure. Here, time runs along the x-axis and each vertical swath of color represents the temperature anomalies for that time, averaged along lines of latitude.
Credit: Tim Bralower © Penn State University with NASA GISTEMP, 2021 and licensed under CC BY-NC-SA 4.0

In the figure above, as in the movie above, the temperatures are given as anomalies, or differences relative to a mean established from some arbitrary period of time (1951-1980 in this case). One thing that is clear is that the polar region of the Northern Hemisphere is the area that has warmed the most — more than 6°C during this time period, when the mean global temperature has risen by a bit less than 1°C. Also clear is the fact that starting around 1990, nearly all of the globe is warming. It's pretty hard to argue with this plot, isn't it?

But if you are still unconvinced, we have another dataset up our sleeves that is completely independent of all of the atmospheric data we have shown so far: the ground has also warmed up!