Common Climate Misconceptions

Global Temperature Records

Measuring Earth’s temperature is no easy task.

Four different groups produce temperature records that attempt to compile a single global mean surface temperature: NASA’s GISStemp, the Hadley Center’s HadCRU, Remote Sensing Systems’ RSS, and the University of Alabama, Huntsville’s UAH.

NASA and Hadley rely on an overlapping set of surface and ocean temperature measurement stations and span the period from 1880 to present. RSS and UAH use satellite monitoring and include only the period from 1979 to present.

Despite differences in calculation criteria and a host of technical problems that have plagued the satellite-based records in the past, all four temperature records now show a remarkable degree of agreement. No single temperature record exhibits a significant or consistent warming bias relative to the others.

Prior to the late 1970s, surface temperature measurements were primarily done via land stations located mostly in the Northern Hemisphere and with limited coverage outside the 20 to 90 degrees North latitudes. Since then, a growing number of surface temperature measurement stations worldwide, coupled with improved methods for correcting for biases induced through urban heat island effects and other station siting and operational issues, have allowed for the development of accurate global temperature estimates. Ocean temperature data prior to the 1980s had been taken from ship-based temperature measurements, and is now obtained via satellite measurements.

The two major surface temperature records are NASA’s GISStemp and the Hadley Center’s HadCRU. Both mostly use the same network of surface stations, but they differ in how they extrapolate temperature changes in areas with few measurement stations. GISS assumes that the correlation of temperature change is fairly strong for stations separated by up to 1,200 kilometers, especially at middle and high latitudes. GISS uses this assumption to extrapolate for almost the entire land area of the Arctic, despite a relative paucity of measurement stations.

HadCRU uses a different approach, and some areas of the Arctic and Antarctic are not covered in its temperature series. This difference in coverage may help explain the slight divergences in series over time, especially given that the Arctic has been warming faster than the rest of the globe.

The two satellite data sets, RSS and UAH, use the Microwave Sounding Units (MSU) of orbiting satellites to estimate lower tropospheric temperature. They estimate this temperature based on measurements of the microwave emissions of oxygen molecules in the atmosphere, which increase proportional to temperature. Lower tropospheric temperature is expected to be similar to, though slightly higher than, surface temperatures, and the surface temperature record produced using the lower tropospheric temperature measurements is adjusted accordingly. MSU-based measurements also provide little coverage of Arctic and Antarctic regions.

The UAH team pioneered the approach in 1979, combining temperature measurements from multiple satellites to produce an estimate for monthly global mean temperatures. UAH published data showing significantly lower tropospheric cooling from 1979 till 1998, contradicting the warming trend observed by the surface stations.

In 1998, RSS was formed to provide an alternative analysis of the MSU data. RSS pointed out a significant error in UAH’s temperature analysis caused by a failure to accurately correct for the effects of orbital decay on observations across multiple satellites. This correction, along with another one in 2005, brought UAH largely in line with the other temperature records, though it continues to show a slightly lower long-term warming trend. Scientists generally supporting the Intergovernmental Panel on Climate Change (IPCC) findings on climate change see this correction of the UAH temperature analyses as a significant vindication of their findings on this issue and, as such, as a major rebuttal to climate contrarians who long had pointed to the differences in surface and upper atmosphere warming trends as supporting their viewpoints.

As shown in the figure below, all four temperature series align remarkably well when normalized on the same baseline period. GISS and HadCRU both show a warming trend of 0.16 degrees C per decade from 1979 to February 2008. RSS shows a warming trend of 0.18 per decade over the same period, while UAH shows a warming trend of 0.14. The largest divergence between temperature series over the last 30 years occurred in 1998, when both satellite-based series showed almost half a degree higher temperatures than land-based series, though the cause of this is unclear.

Global surface and lower troposphere monthly mean anomalies relative to the 1979-1998 mean temperature. Data from GISS, HadCRU, RSS, and UAH ranging from January 1979 to February 2008.

Journalists dealing with temperature data should keep in mind that there are a number of different global mean temperature series available, and that advocates often tend to pick the one that will best reinforce their perspectives. Over longer time periods, however, the differences between different temperature series’ settle out, and all show comparable warming temperature trends.

Zeke Hausfather

Zeke Hausfather, a data scientist with extensive experience with clean technology interests in Silicon Valley, is currently a Senior Researcher with Berkeley Earth. He is a regular contributor to Yale Climate Connections (E-mail:, Twitter: @hausfath).
Bookmark the permalink.

18 Responses to Global Temperature Records

  1. mick says:

    The difference may appear visually small but the numerical difference is huge.

    The GISS and HadCRUT data shows a strong obvious warming signal, the satellite data displays a stable system broken by the 1998 super el nino.

    The Satellite data shows current temperatures equal to 1980. The GISS Hadcrut shows it is .4 degrees higher.

    There is huge, huge difference between the two data sets.

  2. James says:

    You state “The largest divergence between temperature series over the last 30 years occurred in 1998, when both satellite-based series showed almost half a degree higher temperatures than land-based series, though the cause of this is unclear.” Many people say El Nino is the reason for this…

  3. Zeke says:


    Satellite temperatures show a trend quite similar to that of surface temperature records. As of the end of February, RSS had a 1979-2008 trend of 0.18 C per decade, higher than that of both GISS and HadCRU. UAH had a trend of 0.14 C per decade, slightly below both. Given the high monthly variability, comparing a single cold month in 2008 to a single warm month in 1980 really doesn’t tell you that much. Its much more useful to compare the annual averages. In both GISS and HadCRU, 2008 was the 9th and 10th warmest year on record, respectively. I’d imagine it would occupy a similar position in both satellite records.


    Indeed, I’d suspect El Nino is the primary reason for this, though the exact mechanism that caused readings to differ is unclear. Perhaps El Nino had a strong effect on the mid troposphere, which is not captured in surface record but bleeds into the satellite TLT band. It could also be due to differing methods for measuring ocean temperatures.

  4. Eli Rabett says:

    You have a minor error worth correcting. The UAH temperature analysis was first done in ~1990. The first MSU was launched in late 1978. Indeed it was Spencer and Christy’s major contribution that they realized the MSU data could be used to construct global temperature anomaly series

  5. Alan says:

    I am trying to keep an open mind on this but it’s tough. I am a chemist, not a meteorologist. But I can not help but be skeptical when organizations whose funding and very existance hinges on climate change “correct” the data set that does not support their agenda.

    I found this to be interesting also:

  6. Darrell says:

    Question on the data – when you show monthly mean temperatures is this based on daytime highs as many analysts use – or are you looking at overnight lows, or both. Temperature trends and trajectories in overnight lows can be very different than temperature trends in daytime highs!

  7. Zeke says:


    While our network of surface stations is hardly perfect, it does seem to do a pretty good job of correcting for issues of poor citing and urban heat island effects. See our earlier article,, for example. Additionally, satellites provide a quasi-independent validation of surface records, and no one is going to accuse Spencer and Cristy of bias towards AGW!


    I’m not sure, but I believe its calculate by taking the monthly average of daily average temperatures and calculating the monthly anomaly relative to the chosen baseline period, but I’m not positive. NASA has a bit more information about the difference between mean daily temps and monthly anomalies here:

  8. Zeke says:

    Since this article seems to be garnering a lot of interest of late, I took the initiative to update the main graph (since its becoming a bit dated). Here are the four major temperature indicies through the end of 2008:

  9. Mike says:

    Zeke – I appreciate your effort but was curious why your graph shows a doubling of each of the following years – 1985,1996,2007 and how to interpret this meaning ?? Perhaps its some bizarre feature of my computer ?

  10. Mike,

    You may be confusing anomaly with actual temperature. The graphs show the temperature over the past 30 years relative to the 1979-1998 mean temperature. That means that 0 on the graph represents the 1979-1998 average, 2 represents 2 degrees above the 1979-1998 average, etc.

    So going from one to two degrees anomaly is not doubling, its just increasing from a given baseline.

  11. Mike says:

    Zeke, my comment was aimed at the increments of the x-axis, which appear to be years, yet there is a repeating of every 11th increment starting with 1985. I now assume this was done to correct for some error/adjustment creep over the span of the graph. It just looked odd at first perusal.

  12. Mike,

    Sorry for the confusion, I thought you were referring to the graph in the original article. The odd x-axis label is just an artifact of it being a bit of a pain to give a monthly dataset a yearly axis in Excel. The proportion of the axis is correct, just the label is a bit wrong. is the same graph with corrected labels.

  13. robert says:


    Which organizations do you believe will lose funding if their findings are different?

    Whether applied to global climate change, regional climate change or weather, satellite temperature measurements of the atmosphere — and the expertise to analyze them — will be necessary and funded for the foreseeable future (as will be climate study in general).

  14. Realistic says:


    Rather than depending upon an INCREDIBLY small subset of temperature data why not consider these data sets:


    Both indicate nothing global temperature wise is out of the ordinary.

    Or would you prefer to ignore the true historical record?

  15. ryan says:

    Thanks for the graph. It shows temperatures since 2007 reverting to your (arbitrary) mean. So we can stop worrying now then can’t we?

    Of course plotting the means avoids the necessity of showing the impact of variance in “weather” (which can vary by 20Celsius on any two days) on the tiny climate signal you are trying to detect. Doesn’t anybodyin climatology know anything about statistics?

    Actually, if you want a real hint of some juicy work to do, take the MONTHLY averages, do a Fourier analysis, detect the higher frequency cyclical weather phenomena, high-pass filter them out and take a look at whats left. You won’t get a global warming signal but you might have a fantastic paper on what is really going on in those charts. Ever heard of a sinusoidal oscillation?

  16. Murf says:

    I’d like to do a little analysis to see for myself what the GISS data show. Ignoring any and all issues of data quality, I just want to see what the data, as given and without any further adjustments, show in terms of a single global annual series.

    I have a question: If we look at the GISS dataset (I’m using [raw GHCN + USHCN corrections] at the moment) as a matrix of year-months x stations, how should one go about getting the data into a single global average annual series, given that there’s so many missing values?

    So far, I can think of two ways to produce a single global series without modifying any data:
    (1) average all the available data over all stations by year-month, disregarding any missing values, then average the monthly series by year to get average annual values;
    (2) average each station by year, omitting any years for each station where there are one or more months missing in the station’s data, then average over all the stations by year.

    I’m wondering what anyone here thinks of either of these methods? What other ways are there? What have others done?

  17. Jeff says:

    Your statement “In 1998, RSS was formed to provide an alternative analysis of the MSU data. ” is not correct. RSS was founded in 1974.

  18. Michael says:

    I came across this site while searching for the definitive global temperature readings that could help me decide if global warming has been occurring since the advent of the “hockey puck” graph.

    My personal belief, prior to finding this discussion, is that those who believe that they have a grasp on global temperatures and what affects them are at the height of arrogance in the scientific community. You have just over a hundred years of data, and most of that is pretty lousy data.

    Seeing a little bit of what goes into the development of a global mean temperature, my beliefs are reaffirmed. You guys don’t have a CLUE what the global mean temperature is. Not only that, but when temperatures don’t seem to be increasing like Al Gore said they would, the name change to “climate change” appears.

    Let me ask you “scientists” that believe in global warming this: If you create a hypothesis, and observations then don’t fit your hypothesis, is it intellectually honest to change the hypothesis and still declare yourself correct? What is it–global warming or climate change? Why do you think that you can take this one variable, CO2, and ignore all the other variables and their effects?

    Clearly, the earth is flat. Otherwise, the people on the other side would fall off!