New Mystery with the Station Temperature Data

I have been comparing the response of global temperature to particular events that happen to the Earth’s climate.  A quick look at any of the satellite records show that a good starting point is the ENSO cycle (El Nino/La Nina) in the Pacific Ocean.  This led me to looking into how the station data sets (GHCN, CRU) respond to comparable events that cause a change in global temperature anomaly.

On average a significant ENSO event will cause the average temperature of the Earth to change by 0.27 °C.  An El Nino in the positive direction and the La Nina in the negative direction.  In fact the global temperature response is a good indicator in the overall strength of the ENSO event.  The El Nino in 1998 stands out as an enormous event both as an El Nino and an increase in the global temperature, at least in the satellite temperature data.

Since satellite data only goes back to 1979, I was looking at the GHCN and CRU temperature records to look for ENSO events.  I found major swings in temperature in 1957, 1964, 1973, 1974 that happened before satellite data was available.  The average temperature response for these events was 0.28 °C.  That is almost identical to the modern satellite response to ENSO.

the Inconvenient Skeptic

Global temperature response to ENSO events. (Blue) the magnitude of the drop for La Nina. (Red) the magnitude for the increase for a La Nina.

Then I compared to the major ENSO event that took place in 1983-1984.  That was a EL Nino in 1983 to a La Nina in 1984.  Both the satellite and station temperature sets showed an almost identical global temperature response.  The station sets a few percent higher for the El Nino and the satellite a few percent higher for the La Nina drop.  The difference was insignificant as the difference was only a couple of percent.

The Inconvenient Skeptic

(Yellow) Station response to ENSO in 1983-1984. (Green) Satellite response to ENSO in 1984-1984. The 1983 was a warming due to El Nino and the 1984 was La Nina cooling. Both sets give very comparable temperature response.

This is when things got strange.  The 1989 La Nina triggered a average temperature drop of -0.202 °C for the two satellite temperature sets.  The station sets only showed a drop of -0.103 °C.  The satellites showed a response that was 97% greater than the stations did.  What had been an easy detector for ENSO in the station record suddenly did not show the same type of response in 1989.

Okay, one event doesn’t mean much, so I looked at the biggest El Nino ever recorded.  The monster 1998 El Nino.  The satellites clearly show a major response to that El Nino with a temperature increase of 0.46 °C.  The stations decided that it was about the same as the El Nino in 1983 with a temperature change of 0.223°C.  The satellites showed a stunning 106% more response to the largest El Nino ever recorded while the stations indicated nothing unusual.

Jump ahead to the major La Nina of 2008.  The satellites once again responded as expected with a temperature drop of -0.210 °C.  Based on historical responses, that is what is expected.  The station response was that it didn’t happen.  According to the CRU and GHCN temperature response, there was no La Nina that year as the temperature resopnse was only -0.096 °C.  The satellite response was 119% greater than the stations.

Finally there is the 2010 La Nina El Nino.  Certainly the stations would bounce back and pick up this La Nina.  That would be incorrect.  The global temperature response to that La Nina was a trivial and insignificant 0.063 °C.  A total non event.  The satellites disagreed with a global response of 0.238 °C.  A strong El Nino, but nothing compared to the 1998 monster.  The difference in response was an absurd 278%.

The Inconvenient Skeptic

(Yellow) Station response to ENSO in 1983-1984. (Green) Satellite response to ENSO in 1984-1984. The 1983 was a warming due to El Nino and the 1984 was La Nina cooling.

Even more odd is the fact that there is a trend in the percent difference between the satellite and the station responses to ENSO events.  It isn’t just that the stations are showing less response, but in each event they are showing a consistently decreasing response to ENSO events.  In the last 2010 El Nino event the stations response was 74% less than the satellite.

The Inconvenient Skeptic

The stations are showing a trend in decreasing response to ENSO events. In the 1983 El Nino they showed slightly more warming than the satellites, in every event since they have show a decreasing response to climate events.

The satellites continue to show the typical response that the stations historically did, but the global station sets have stopped responding to ENSO events.  They don’t cool, they don’t warm.  They are starting to just stay the same.

The argument that the station temperature records simply don’t respond in the same manner is false.  They used to.  Up until 1984 the major ENSO events show up clearly in the global temperature record.  Going back to 1900 I found the El Nino temperature signal in 1914, 1926, 1930, 1951, 1957, 1973 and 1977.  Every one of those also happened to be part of an El Nino event and all showed more than twice the temperature response that the stations did in 2010.

I also found the major La Nina events in 1907, 1916, 1929, 1954, 1964 and 1974 that took place prior to the satellite era.  All of them showed strong drops in global temperature in response to the cooler Pacific Ocean temperatures.  All of them showed much stronger response that the same station temperature sets do today.

The Inconvenient Skeptic

Historically the station sets have shown a clear correlation to ENSO events. They have stopped being an accurate indicator of ENSO events. The 1998 event was the strongest ever recorded. When compared to past ENSO events it does not stand out in the station data.

Global warming doesn’t explain it either because such warm ocean events do warm the atmosphere.  The satellites clearly show the difference.

This leaves us with a mystery: Why have the station temperature sets stopped responding to ENSO events?


Update:  Here is the monthly response to the 1998 El Nino event.

The Inconvenient Skeptic

(Blue) Satellite response, (Red) Station Response to the 1998 El Nino. They both show the same behavior, but the magnitude of the change is much larger for the satellite.

Posted in Anomaly and Measurement Methods by inconvenientskeptic on March 20th, 2011 at 4:15 am.


This post has 18 comments

  1. Malaga View Mar 20th 2011

    Great posting… good digging… wonderful analysis…
    Another confirmation that The Team is cooking the books.

  2. nofreewind Mar 20th 2011

    Very interesting original thought. You come up with with some interesting stuff.

  3. SoundOff Mar 20th 2011

    TIS, you said: This is when things got strange. The 1989 La Nina triggered an average temperature drop of -0.202 °C for the two satellite temperature sets. The station sets only showed a drop of -0.103 °C.

    Let’s compare your findings for this La Nina episode to publicly available data.

    GISTemp Monthly Anomalies during the
    La Nina episode from May/1988 to Jun/1989

    1988 May = 0.380
    1988 Jun = 0.400 maximum during La Nina (A)
    1988 Jul = 0.260
    1988 Aug = 0.310
    1988 Sep = 0.300
    1988 Oct = 0.290
    1988 Nov = -0.030 minimum during La Nina (B)
    1988 Dec = 0.190
    1989 Jan = 0.030
    1989 Feb = 0.280
    1989 Mar = 0.270
    1989 Apr = 0.160
    1989 May = 0.040
    1989 Jun = 0.050

    B – A = -0.030 – 0.400 = -0.430 drop from A to B

    The GISTemp station drop of -0.430°C is far in excess of your satellite drop of -0.202°C and completely different from your stated station drop of -0.103°C. You are right – this is strange.

    Your satellite drop of -0.202°C does not agree with what I see in UAH and RSS either. What I see is a drop of -0.610°C and -0.523°C respectively, or a satellite average drop of -0.567°C, just a bit higher than the GISTemp figure, which is what I expect since satellites are more sensitive to ENSO changes.

    I also tried a second method: comparing the average anomaly for this 14 month La Nina to the preceding 14 months rather than just looking at the max/min swings. Using this method the GISTemp drop was -0.092°C, UAH was -0.149°C and RSS was -0.161°C, for a satellite average of -0.155°C. Again the GISTemp drop is not radically different, certainly not the 100% difference that you found.

    Mystery solved – your data or methodology must be wrong. 🙂

    (Since we are examining swings only, the differing base periods can be ignored.)

  4. inconvenientskeptic Mar 21st 2011


    I used the CRU and GHCN (NCDC) annual average.

    CRU 1989 = 0.21 C: 1988 = 0.338 C
    GHCH 1989 =0.22C: 1988 = 0.299 C

    Annual change = -0.103 C

    Not a difficult method. The satellite data:

    RSS 1989 = -0.127C : 1988 = 0.067C
    UAH 1989 = -0.201C : 1988 = 0.018C

    Annual change = -0.202C

    So the data and the method are available and consistent.

    The same method in the past worked, why isn’t it working now?

    The 1998 El Nino is really the odd one because it was so incredibly strong.

  5. SoundOff Mar 21st 2011

    TIS, I’ve read several peer-reviewed articles on the ENSO topic. My understanding is that no reliable ENSO records exist for periods prior to the 1950. At least I could not find any peer-reviewed or authoritative sources when I was doing a temperature analysis last year. Can I have a link to your pre-1950 ENSO source? I use the following source.

    Your annual anomalies agree closely with mine, so we seem to be working from the same data and the difference seems to be one of methodology.

    First I would say comparing annual records is not the way to do this analysis. Satellite readings respond quickly to ENSO episodes while the surface response both lags and is spread over a longer period, and is therefore generally less extreme (this is because the surface is moderated by oceans and land masses which the LT atmosphere is not). For example, this means a short duration but intense ENSO episode will spike satellite readings but barely make a dent in surface records. The latter only responds to longer duration episodes.

    Second, due to the time lag and the stretched out response time for surface records, you will get spurious results using annual numbers. The quicker satellite response will usually be contained within one year. The longer surface response could be contained within one year or it could be split between two years. Each alternative would give you totally different annual surface average temperatures. Since ENSO cycles have intensified and lengthened under AGW, this split effect is more likely to affect the recent episodes.

  6. inconvenientskeptic Mar 21st 2011


    As I pointed out in the time lag series of articles it does take a few months for the atmosphere to respond.

    I went ahead and added a chart comparing the 1998 event for both the station and the satellite.

    As you can see from the monthly data there is no reason that the annual is not fine for a response comparison. The satellite response is superior.

    There is more along this line of inquiry coming as well.

    As for the pre 1950 ENSO. I agree that there are not reliable data for the magnitude of the events, but there are records for the existence of the events.

    I looked around and found a few different sources, but here is the best one. Not a lot of detail, but enough to confirm that the delta I found in the global temperature correlated to ENSO events.

  7. SoundOff Mar 21st 2011

    TIS, Thanks for the Storm Fax link. Unfortunately the site doesn’t say where it got its data from and it doesn’t exactly look like a source of original research, plus there seems to be no individuals behind the site willing to state their names *. My fear is someone just looked though a surface temperature record and decided that every single wiggle was an ENSO episode to make the list complete. If so, then of course the dates are going to correlate with temperatures. The list they provide for the modern period is vague, off by a year at times and even missing whole episodes (e.g. La Nina from Dec/1967 to May/1968). This makes it even more doubtful for historical periods,

    * They say “The Site is owned, operated and maintained by STORMFAX, INC. for your personal entertainment and information.” An IP trace says the site is registered to a Boca Raton, FL address but is owned by a web hosting service (Network Solutions) registered in Herndon, VA. No science or weather site connections.

  8. SoundOff Mar 21st 2011

    TIS, I’ve looked at your analysis of the 1998 El Nino. I haven’t verified it against my own numbers but it seems reasonable. Basically it just agrees with what I’ve been saying, which is that the surface record is a muted version of the satellite record but follows the same trend.

    I hope adding the following will be interesting to your readers – it is another dimension of the same topic:

    E means El Nino
    L means La Nina
    = means neutral


    ENSO Pattern 1950 to Present: LLLLLLLLLLLLLLL ==== EEEEE =========================== LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL == EEEEEEEEEEEEEEE ================================================== LLLLL ===== EEEEEEE == LLLLLLLLLL ==== EEEEEEEEEEE =================== LLLLL ====== EEEEEEEE == EEEEE ===== LLLLLLLLLLLLLLLLLLL === EEEEEEEEEEE = LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL === EEEEEE ====== EEEEE =================================================== EEEEEEEEEEEEEE =============== LLLLLLLLLLLL ========== EEEEEEEEEEEEEEEEEEE == LLLLLLLLLLLLL ======================= EEEEEEEEEEEEEEE ===================== EEEEEEEEEEE ===== LLLLLLL ============= EEEEEEEEEEEEE = LLLLLLLLLLLLLLLLLLLLLLLL === LLLLL ============== EEEEEEEEEEE ============== EEEEEEEEE ================= EEEEEE ======= LLLLLLLLL ============ EEEEEEEEEEE == LLLLLLL

  9. inconvenientskeptic Mar 22nd 2011


    I have cross referenced years against coral temperatures from the Pacific. Dates match up well. It is hard to distinguish the scale of the events from a single location, but reconstructions have been done. The same paper also states that ENSO was most active in the 1700’s. When the climate was cooler. That is an interesting bit as well.

    Here is the main ENSO reconstruction that I have found. It matches the dates very well. Especially for when the major transition happens.

    So you are agreeing with my point that the satellites are more sensitive at detecting changes to the Earth’s temperature?

  10. SoundOff Mar 22nd 2011

    If your point is that satellites are more sensitive at detecting changes to the Earth’s Lower Troposphere temperature, then I agree. The Lower Troposphere is all air and air responds overall more quickly to local warming events (such as ENSO episodes) because it mixes more quickly around the planet. That’s why I say satellite records are spikier.

    Surface temperatures take longer to catch up while the oceans and land masses attempt to equalize with the air high above them and they probably never do fully. Therefore, surface records show a muted version of what’s happening above.

    Neither is more accurate, they are just different, but related. Since they are related, I expect to see matching rises and drops in time. I don’t expect to see equal magnitude movements. The real mystery is that you did find equal magnitude movements in some of the older records. This is partly explained by using annual anomalies in your analysis. There may be other factors at work too. Perhaps solar cycles or volcanoes coincided with some of the ENSO episodes thus moving surface temperatures more than they usually do for an ENSO event. The older surface temperature record has less resolution and the fewer stations back then may have been in areas that were more in the path of ENSO effects. There are endless possibilities when explaining weather.

  11. inconvenientskeptic Mar 22nd 2011


    Surface temps are a combination of the air temperature about a meter above the ground and the surface temperature of the ocean.

    Since the SST is part of the station record, the station data should respond quickly as SST’s lead the change in the atmosphere. The atmosphere cannot warm/cool more than the ocean SST changes, so it should also respond more strongly than the atmospheric response.

    If you read the next article you will see that the station data also responds poorly to other events that it should easily detect, but doesn’t.

  12. Billy Liar Mar 22nd 2011

    Is it anything to do with the great thermometer die off? (ie fewer stations reporting)

  13. inconvenientskeptic Mar 22nd 2011


    It is possible that the decrease in stations might impact the magnitude of the sensitivity.

  14. SoundOff Mar 22nd 2011

    TIS, There’s a huge difference between the air temperature in the few meters near the surface, which is warmed by conduction and radiation off the immediate surface and the rest of the troposphere, which is warmed by influences from far away. If the cause of warming is from a widespread phenomenon such as the greenhouse effect, solar cycles or possibly a volcanic eruption that reaches into the stratosphere, then the troposphere will respond in line with SST and land temperatures. However, an ENSO episode is a local release of heat stored in the South Pacific Ocean into the atmosphere (or a local storing of heat causing the atmosphere to cool during the opposite phase). This kind of local warming/cooling depends on the troposphere to be distributed around with sea and land responding in a muted lag. Difference rules apply for climate forcings and heat redistributions.

  15. inconvenientskeptic Mar 23rd 2011


    If there was a time lag associated with the ENSO and the station data I would agree with you, but the problem is magnitude.

    If the bulk of the troposphere (75% of the total atmospheric mass) shows a jump of almost 0.5C due to the 1998 El Nino event, why didn’t the station data show a comparable increase in temperature.

    My argument is that if the station data doesn’t show as much resolution to changes in atmospheric temperature, it is a poorer method of measuring temperature.

  16. SoundOff Mar 23rd 2011


    Let’s try a simple physics experiment involving heat transfer. Suppose someone heats one end of a meter long metal bar with a torch for 15 minutes, then stops for 15 minutes, and then repeats this process for many cycles. At the same time, the temperature of the bar is measured at both ends.

    The thermometer at the heated end would show a quick response to the positive part of cycle when heat is being applied and it would also show large magnitude swings in temperature. The other end of the bar would warm up too but its thermometer would show a temperature peak that’s slightly out of phase with the positive part of the cycle and it would show smaller magnitude swings in temperature. The missing heat is radiated away from the bar somewhere between the ends.

    The thermometer at the cooler end of the bar is not a poorer method of measuring temperature. It is simply showing what the temperature really is at that end. If we are holding on to the bar at that end, then it is the more important measurement of the two. The heat transfer mechanism at work with the metal bar is conduction while other mechanisms are at work in the climate system but the same thing is happening – heat is transferred away from a local hot spot, some of it makes it to the surface in other parts of the world and some of it is lost during the transfer. We live on the surface not up in the troposphere.

    The peak global average temperature is usually achieved between six months to a year after the peak of an El Niño. The troposphere initially absorbs the released heat and spreads it around the planet. Some of that heat increases the surface temperature far away from the source of the heat release but some of that heat is lost by being radiated away into space by the troposphere and never shows up in surface records.

    One interesting fact that I read somewhere is that the temperature peak at the bottom of the Aral Sea actually occurs in local winter, 100% out of phase with summer surface heating. This is a real world demonstration of heat transfer lag at work within a body of water.

  17. inconvenientskeptic Mar 23rd 2011


    I understand you are arguing that the stations are measuring the lagged part compared to the satellite.

    I don’t care why the stations show less response. When trying to detect global warming, the method that is most effective at detecting changes in the Earth’s temperature is the better method.

    If the worlds most sensitive thermometer is placed in the middle of the field, nothing will measure that temperature more accurately, but it will be horrible at detecting what is happening globally. The temperature of that field might be known accurately to 50 decimal places, but that is meaningless.

    If the satellites provide better resolution to changes in the TLT (which is what the anomaly from the satellites is) than station data, they will also be better at detecting global warming.

    I am not saying the station data isn’t useful, but I am arguing that it is not sensitive at detecting changes in global temperature.

    The method of combining and interpolating station data and SST’s from ships and Argo’s data and whatever else they have is a lesser method.

    Despite the problems with the satellite data, they provide better global resolution to overall change in temperature.

  18. SoundOff Mar 23rd 2011


    It’s not clear to me why you want to measure the Earth’s atmospheric temperature. It’s of relevance to weather forecasting but the short term chaotic nature of the atmosphere makes it less suitable for climate change analysis or to detect global [surface] warming.

    As I’ve shown in earlier comments, the atmosphere overreacts to large heat transfers, which are redistributions of heat and not true global warming events. The ideal source would be full ocean heat content measurements but this is not yet possible. So the surface record is the next best thing and satellites don’t do that well.

    If you want to measure atmospheric temperature, then why use TLT, which is strongly influenced by surface temperatures? Why not use TUT or TLS or an average of all these to measure the temperature of even more atmosphere?

    Like you, I am not saying the satellite data isn’t useful, but I am arguing that it is not sensitive at detecting changes in global temperature on the surface, which is what most people care about – you excepted. I’d rather have a thermometer in every field.

    Many of the issues we’ve been discussing here with respect to satellite measurement of temperatures are described in great detail in a new article that was published today at Skeptical Science. It’s a must read for those who want to understand the nature of satellite data.

Web Design & Dev by

Mazal Simantov Digital Creativity