Climate Science for Everyone: How do scientists measure the temperature of the Earth?

To read other articles in this series, click here.

Figure 1: Weather station (image credit:

Before we answer the question in the title above, I’m first going to ask and answer a different question: how do you measure the temperature of a glass of water?

I’m guessing that most people will answer the question with some variation of “use a thermometer.” Stick a digital or glass thermometer into the cup, measure the digital reading or look at the colored bar inside the glass, and write down the number. Done, right?


We didn’t measure the temperature of the cup of water at all. We measured the temperature of the water that was next to the thermometer at a time just before we looked at the thermometer. But when we put the thermometer into the water, that act changed the temperature slightly as energy flowed between the thermometer and the water.

We can try again, of course. Let’s put the thermometer into the water again and let it sit for a little while, until we think that the probe and the temperature of the water have equalized, write down the number, and we’ve got it, right?

Nope, I’m afraid not. We’re still only measuring the temperature of the water that’s next to the thermometer.

We can add more thermometers, or we can stir the water, or we can even buy an expensive infrared camera, but it won’t matter – we’ll never actually measure the temperature of the whole cup of water.

The reason is that the cup of water isn’t the exact same temperature everywhere. Heat doesn’t flow through water perfectly, and so heat builds up in some places, resulting in warmer and cooler spots in the cup.

Here’s what I mean.

Figure 2: Cup of water sitting on a table

Let’s assume that the cup of water is standard tap water, at room temperature, sitting on a table, as shown in Figure 2 at right.

In this case, the top surface of the water is in contact with the air. Since water evaporates, and evaporation takes a lot of energy, the top surface is going to be slightly cooler than the rest of the cup of water. If we rest our hand on the table next to the glass, then the water in the bottom of the cup will get slightly warmer as heat is conducted from our hand, through the table to the cup, through the glass base of the cup, and into the water. We can make these effects small, at least for a little while, by stirring up the water some, but we can never get rid of them completely.

And as I mentioned above, adding a thermometer to the water changes the temperature of the water that’s around the thermometer. The thermometer absorbs a small amount of energy from (or gives a small amount of energy to) the water. If it didn’t, the thermometer wouldn’t work in the first place. And thermometers aren’t perfectly accurate either. Commercial digital thermometers can be accurate to 0.1 degrees Fahrenheit or so.

So if we can’t perfectly measure the temperature of a cup of water, we have to ask – do we really care? For most things, the effects I described above are so small that we don’t care, and so we can say that we measured the temperature of the water in the cup even though we know only got close. In those very strange cases where we might care about how the temperature of the water differs, we can use complicated computer models to model evaporation, conduction of heat from the table through the cup’s bottom into the water, convection of lower density warm water up to the surface where it cools, increases density, and falls again, and so on.

Now, back to the original question: how do scientists know what the temperature of the Earth is? As with the example of the cup of water, the pedantic answer is “they don’t, but they don’t need to know the Earth’s temperature.” The more scientifically useful and answer is “there are several different temperatures to measure, and they’re measured at least three different ways.”

Albatross (image credit: RSBP.UK)

Let’s talk first about the global surface temperature, which is the temperature that most people think about when they’re talking about global warming (we’re not albatross or naked mole rats, which live most of their lives in the air or underground, respectively). For the global surface temperature, scientists generally use thermometers. There are literally thousands of very accurate thermometers scattered around the world. As you can probably imagine, most of them are on land, but there’s a lot of floating buoys and ships traveling the oceans that also have thermometers and measure the temperature at various heights above the surface of the ocean. Scientists combine the measurements and correct them for known errors using mathematics, and from that they can calculate the temperature of the Earth’s surface1. That data is available from sites like these: NASA Goddard Institute for Space Sciences (GISS), NOAA National Centers for Environmental Information (formerly the National Climatic Data Center, or NCDC), and the UK Hadley Center.

Of course, it take a long time to gather the data and do the calculations, so scientists don’t know what the global surface temperature is, only what it was several days or weeks ago. Thankfully, for the purpose of determining whether global warming (aka climate change or industrial climate disruption2), changes in daily and even monthly temperatures don’t matter much, while the changes in temperature over the course years and decades matters a lot. And while microclimates (small pockets of warmer or colder temperatures, often due to local terrain) aren’t usually captured in the thermometer data, it’s pretty rare to have a situation where that much detail is needed. Just like it’s pretty rare to need to know how the temperature changes within a cup of water sitting on a table.

One issue of potential concern is this phrase that I used: “correct them for known errors.” Some of the known errors that can affect temperature records include changes in the time of day that measurements are taken, the movement of a thermometer from one place to another, the addition (or subtraction) of nearby buildings or streets, urbanization around the thermometer, and changing the type of thermometer from glass to digital. Generally speaking, scientists are aware of how these errors affect the global surface temperature data, and so scientists corrected for them. In fact, failing to correct the data for known errors would generally be considered scientifically irresponsible.

A second way to measure the surface temperature of the earth is from a satellite using what’s called an “advanced microwave sounding unit,” or AMSU. An AMSU looks at the microwave portion of the electromagnetic spectrum and measures how many microwaves are being emitted by oxygen molecules, which vary with altitude and temperature. These microwave measurements are then fed into a computer program that does the math required to correct for known errors and to come up with an estimate of the global temperature of part of the Earth’s atmosphere. That data is available from sites like Remote Sensing Systems and the University of Alabama-Huntsville, among others.

Figure 4: AMSU weighting functions (image credit: Remote Sensing Systems)

One of the problems with this method is that the broad swaths of atmosphere that the AMSU measures overlap, as shown in Figure 4. A few channels do reach all the way down to the Earth’s surface, but even the lowest-reaching (TLT) AMSU channel peaks at about 2-3 km above the Earth’s surface and is somewhat affected by temperatures as high as 10 km (for reference, Mount Everest is almost 9 km tall), an altitude at which people die from lack of oxygen. This makes the AMSU data less than ideal for measuring the lower parts of the atmosphere (or the surface temperature).

Another problem is that the satellite that the AMSU is on, the Aqua satellite, has what’s known as a “return period” of 233 orbits, or about 16 days. That means that the satellite can indirectly measure the entire Earth’s temperature only once every 16 days. Most months that will be twice in a month, but some months it’ll be only once. Again, though, measurements used for the purpose of monitoring global warming wouldn’t be affected by this.

One more problem is that satellites drift over time, moving earlier or later in the day, and gradually lower in altitude. These drifts result in errors in the calculated temperature that can vary in a complication fashion as the orbit of a satellite changes.

And a final problem is that each satellite has only a limited life in orbit. Since the first MSU was launched in 1978, there have been 14 different satellites that have carried an MSU or an AMSU. The different satellites didn’t always overlap with each other, and they suffered different types of errors, all of which have to be corrected in order to create a single, continuous dataset from 1978 through to the present.

Figure 5: radiosonde (image credit:

There’s a third way to measure the Earth’s temperature – weather balloons carrying instruments known as “radiosondes.” Balloon-borne radiosondes are released twice daily from hundreds of locations around the world and they measure their position, altitude, air temperature, humidity, etc. as they rise through the atmosphere. Eventually the balloon pops and the radiosonde falls, but the data is radioed down to the ground or a satellite before that happens. The measurements give scientists direct thermometer data for the temperature of the atmosphere as they rise to altitudes in excess of 100,000 feet (or over 30 km). Again, scientists take the measurements and apply corrections for known sources of error and then generate temperatures for the various altitudes in the atmosphere. Some sources of radiosonde data are the NOAA ROAB and RATPAC datasets.

Of course, there are problems with radiosondes as well. Radiosondes are released every day, so there’s good coverage of the Earth’s temperature with respect to time. But they’re released from a relatively small number of locations on the Earth, and nearly always from land, so the coverage of the Earth’s surface isn’t as good as, for example, satellites can provide. And they’re carried by the wind as they rise through the atmosphere, so it’s impossible to guarantee that you’re getting good, repeatable data from the same locations each time we launch a radiosonde.

Of course, this makes radiosondes ideal for studying wind, so what might be a “bug” for one scientist may be a “feature” for another.

So which way is the best way to measure the Earth’s temperature? The one that makes the most sense for a scientist’s research. If a scientist is interested in the temperature of the Earth’s surface, either the thermometer or the satellite records would work. If the scientist is interested in the temperature of the atmosphere, then either the satellite or the radiosonde measurements are probably the best.

And which method gives us the temperature of the Earth itself? All of them, and none of them. Each method has its own set of errors that need to be corrected, and each method tells part of the whole story. So while it’s true that we don’t know the temperature of the Earth, as with the cup of water above, most of the time it doesn’t matter. We don’t need to know the Earth’s temperature to a super high accuracy (say, better than 0.18 °F, or 0.1 °C)

Generally speaking, though, if what you’re interested in is the places on Earth where people live, where crops are grown, where livestock are raised, where we catch fish, then I’d recommend using the thermometer record. It goes back to the 1880s across most of the Earth (and back to the 1860s in a few places). The various types of thermometers that have been used over the years are understood well enough that scientists can correct for the differences across types. The effects of urbanization, of moving monitoring stations, and of changing the time of day that measurements were taken are all well understood, and thus the errors they each add can be corrected. And scientists have repeatedly analyzed the difference between corrected and uncorrected temperatures and found that the difference is small since about 1900.


  1. Scientists actually discovered something interesting when they first did this. The discovered that thermometers that were close together tended to go up and down together. For the purpose of measuring global warming, the temperature difference from one year to the next is what’s important. So if two thermometers go up and down the same amount from one year to the next, then we don’t really need both thermometers to measure the change – one will do. When scientists analyzed this, they found that stations within about 1000 km of each other tended to change together, meaning that we could theoretically measure the Earth’s surface temperature with a relatively small number of ideally-placed thermometers. Doing some quick calculations and I estimate that, if we could place the thermometer’s ideally, we could monitor the surface temperature of the entire Earth with as few as 1000 thermometers. Scientists are presently using 1219 land-based stations plus hundreds of ship and buoy-based thermometers as well.
  2. Industrial climate disruption is scientific is defined as the consensus position is that the climate is changing, that the emission of greenhouse gases by human industry is the dominant driver of those changes, and that the changes will almost certainly be disruptive to human society and global ecology.

3 replies »

  1. About 200 thermometers, well spread over the Earth, would theoretically already be enough. However, that nearby stations measure about the same climate also means that we can compare stations with each other and detect and remove artificial changes in the observations. For example because on the stations changed their thermometer or because there was building activity near the station. Thus we need more stations than 200 for *reliable* data.

    Furthermore, we would like to know a lot more than just the global mean temperature. That gets most attention in the public debate, but the spatial map of the warming is scientifically very important for understanding the reasons for the warming. More stations are also necessary to understand regional climatic variations.

  2. Kevin Cowtan created a temperature tool for the MOOC “Denial101x – Making sense of climate science denial” and it is freely available on Skeptical Science:
    As explained in the short blog post:
    “You can look at temperature records at any scale, from a local weather station to national and global records. You can investigate common myths, such as the impact of urban heat islands and adjustments to the data. And you can use simple statistical tests to examine the accuracy of the record for yourself.”
    It can for example help to answer the question how many thermometers are needed to get reliable measurements?

  3. Thanks, Brain. This is fantastic. I’ll be showing this to the kids in my science class about precise and accurate measurements in science. And thanks for the link, Baerbel.