CMIP5 (IPCC AR5) Climate Models: Modeled Relationship between Marine Air Temperature and Sea Surface Temperature Is Wrong – (CORRECTION: It’s Not Wrong)

UPDATE 2 (February 27, 2013): A problem was discovered with how the KNMI Climate Explorer had processed the MOHMAT data used originally in this post. KNMI has now corrected the bug.

In turn, the modeled relationship between marine air temperature and sea surface temperature now better agrees with the observations.  See the Revised Figure 4. In looking at the difference between the modeled and observed marine air temperature, consider that the MOHMAT data includes only nighttime readings while the model outputs do not.

Many thanks to blogger lgl, who found the problem with the KNMI version of the MOHMAT data while reviewing the original cross post at WattsUpWithThat.  And many thanks to Dr. Geert Jan von Oldenborgh at KNMI for his immediate investigation and corrections.

Revised Figure 3

Revised Figure 3

I’ve changed the title of the post and most of the text has been given a strikethrough.  I’ve also deleted the original erroneous graphs so not to cause confusion when researchers use search engines in the future.   Climate models have enough problems without adding wrongly to them.



UPDATE  February 26, 2013:  We’re looking into a possible data problem at UKMO or the KNMI Climate Explorer.


This post is part of a series of model-data comparisons that illustrate the failings of the climate models prepared for the upcoming 5th Assessment Report (AR5) from the Intergovernmental Panel on Climate Change (IPCC). The poor performances of the CMIP3 models prepared for the IPCC’s AR4 have been shown in numerous other posts.

In this post, we’ll compare the Marine Air Temperature (MOHMAT4.3) data from the UK Met Office to two of their Sea Surface Temperature datasets, HADISST and HADSST3. The source of data is the KNMI Climate Explorer. And we’ll present the multi-model ensemble mean (the average of all of the climate model runs) of the simulations of Marine Air and Sea Surface Temperature from the climate models stored in the CMIP5 archive, which have been prepared for the IPCC’s AR5. They’re also available through the Monthly CMIP5 Scenario Runs webpage at the KNMI Climate Explorer. For the modeled Marine Air Temperature, I’ve used the model outputs of “Ocean Only” surface air temperature (TAS). In other words, the outputs of land surface air temperatures have been masked. Modeled sea surface temperature are listed as TOS at the KNMI Climate Explorer.

As you will see, observations indicate the sea surface temperatures warm at a faster rate than marine air temperatures, while the models have the relationship backwards. Not too surprising, since the models simulate very little properly.


The MOHMAT4.3 marine air temperature data is available through May 2007 at the KNMI Climate Explorer, it is no longer updated by the UKMO, so we’ll end the comparisons then.

Animation 1 is a series of maps at 10-year intervals. They illustrate the grids (in purple) where the marine air temperature data is available. As you can see, there’s very little data prior to 1950. There’s also very little data south of 30S and north of 60N, so we’ll also limit the comparisons to the data between those latitudes.

Animation 1

Animation 1

In summary, the data and model outputs presented in this post are for the time period of January 1950 through May 2007 and for the latitudes of 30S-60N. The base period used for anomalies is 1971-2000.


Figure 1 compares anomalies of the two sea surface temperature datasets (HADISST and HADSST3) to those of the marine air temperature data (MOHMAT4.3). The data are presented on monthly bases. The sea surface temperature datasets exhibit very similar warming trends over this period. On the other hand, the marine air temperature anomalies warm at a significantly slower rate—warming at a rate that’s about 60% of the sea surface temperature datasets. Or to look at the other way, sea surface temperatures are warming at a rate that’s about 1.7 times faster than the marine air temperatures.

Figure 1

As noted in the introduction, the climate models get the relationship backwards. See Figure 2. Modeled marine air temperature anomalies warmed at a rate that’s about 1.2 times faster than the modeled sea surface temperature anomalies.

Figure 2


HADSST3 is only presented as anomalies, so it cannot be included in this comparison.

Figure 3 compares modeled and observed sea surface temperatures and marine air temperatures. As you’ll note, I’ve switched to annual data. Modeled sea surface temperatures and their trend are reasonably close to one another. The models are slightly cooler than the observations. Then there’s marine air temperature. The modeled marine air temperature is too warm and it warms at a rate that’s about 2.2 faster than observed.

Figure 3

Last is Figure 4. It compares the modeled and observed temperature differences between the sea surface temperatures and marine air temperatures, where the marine air temperatures are subtracted from the sea surface temperatures. The observed temperature difference (SST Minus MAT) averages about 2.9 deg C, and the difference has been increasing. But the modeled difference averages about 1.25 deg C and has been decreasing.

Figure 4


As illustrated above, climate models cannot properly simulate the relationship between marine air temperature and sea surface temperature. That shouldn’t surprise anyone who has been following the model-data comparison posts.

Climate models can’t simulate satellite-era sea surface temperatures, as illustrated and discussed in the posts here, here and here. In another post here, we showed that climate models simulated a 0.4 deg C warming of the sea surface temperatures for the Pacific Ocean over the past 19 years, but the observations show Pacific Ocean sea surface temperatures haven’t warmed in that period. As shown in the post here, climate models do a poor job of simulating land surface air temperatures and a good job in other when the models are compared to data regionally. As shown in the post here, models can’t simulate the polar amplified cooling during the mid-20th Century cooling period, they fail to capture the polar amplified warming during the early warming period of the 20th Century, and they fall short of simulating the polar amplification during the recent warming period. Climate models cannot simulate global surface temperatures since 1900, when the data and model outputs are broken down into the two warming periods and the two flat (slightly cooling) periods, as illustrated in the posts here and here. That is, they cannot simulate the rates at which the temperatures cooled during the early 20th Century (1901-1917) and the mid-20th Century (1944-1976) cooling periods, and they definitely can’t simulate the early 20th Century warming period (1917-1944). And they definitely miss the boat when trying to simulate global precipitation. Refer to the post here. Satellite-era precipitation data indicates global precipitation has decreased since 1979, but climate models simulations say it should have increased.

Further, sea surface temperature data during the satellite era and ocean heat content data do not support the hypothesis of manmade greenhouse gas-driven global warming. Worded another way, satellite-era sea surface temperature data and ocean heat content data indicate the oceans warmed naturally. This was illustrated and discussed in detail in my essay titled “The Manmade Global Warming Challenge”. The introductory blog post is here and it can be downloaded here (42MB). This was also presented in my 2-part YouTube video series titled “The Natural Warming of the Global Oceans”. YouTube links: Part 1 and Part 2. And it was illustrated and discussed, in minute detail, in my ebook Who Turned on the Heat? which was introduced in the blog post “Everything Your Ever Wanted to Know about El Niño and La Niña”. It’s available for sale only in pdf form here. Price US$8.00. Note: There’s no need to open a PayPal account. Simply scroll down to the “Don’t Have a PayPal Account” purchase option.


We’ve published numerous posts that include model-data comparisons. If history repeats itself, proponents of manmade global warming will complain in comments that I’ve only presented the model mean in the above graphs and not the full ensemble. In an effort to suppress their need to complain once again, I’ve borrowed parts of the discussion from the post Blog Memo to John Hockenberry Regarding PBS Report “Climate of Doubt”.

The model mean provides the best representation of the manmade greenhouse gas-driven scenario—not the individual model runs, which contain noise created by the models. For this, I’ll provide two references:

The first is a comment made by Gavin Schmidt (climatologist and climate modeler at the NASA Goddard Institute for Space Studies—GISS). He is one of the contributors to the website RealClimate. The following quotes are from the thread of the RealClimate post Decadal predictions. At comment 49, dated 30 Sep 2009 at 6:18 AM, a blogger posed this question:

If a single simulation is not a good predictor of reality how can the average of many simulations, each of which is a poor predictor of reality, be a better predictor, or indeed claim to have any residual of reality?

Gavin Schmidt replied:

Any single realisation can be thought of as being made up of two components – a forced signal and a random realisation of the internal variability (‘noise’). By definition the random component will uncorrelated across different realisations and when you average together many examples you get the forced component (i.e. the ensemble mean).

To paraphrase Gavin Schmidt, we’re not interested in the random component (noise) inherent in the individual simulations; we’re interested in the forced component, which represents the modeler’s best guess of the effects of manmade greenhouse gases on the variable being simulated.

The second reference is a similar statement by the National Center for Atmospheric Research (NCAR). I’ve quoted the following in numerous blog posts and in my recently published ebook. Sometime over the past few months, NCAR elected to remove that educational webpage from its website. Luckily the Wayback Machine has a copy. NCAR wrote on that FAQ webpage that had been part of an introductory discussion about climate models (my boldface):

Averaging over a multi-member ensemble of model climate runs gives a measure of the average model response to the forcings imposed on the model. Unless you are interested in a particular ensemble member where the initial conditions make a difference in your work, averaging of several ensemble members will give you best representation of a scenario.

In summary, we are not interested in the models’ internally created noise, and we are not interested in the results of individual responses of ensemble members to initial conditions. So, in the graphs, we exclude the visual noise of the individual ensemble members and present only the model mean, because the model mean is the best representation of how the models are programmed and tuned to respond to manmade greenhouse gases.

About Bob Tisdale

Research interest: the long-term aftereffects of El Niño and La Nina events on global sea surface temperature and ocean heat content. Author of the ebook Who Turned on the Heat? and regular contributor at WattsUpWithThat.
This entry was posted in Model-Data Comparison SST. Bookmark the permalink.

7 Responses to CMIP5 (IPCC AR5) Climate Models: Modeled Relationship between Marine Air Temperature and Sea Surface Temperature Is Wrong – (CORRECTION: It’s Not Wrong)

  1. John (in Oz) says:

    Multi-model mean seems counter-intuitive and, no doubt, there are statisticians that will recognise my ignorance at once.

    If I have several radios slightly mistuned to a station and induce (random) static, will I be able to average these to remove the noise and listen to the radio station?

    The average of many bad guesses would not seem to lead to the truth, as is being shown by your article.

  2. batheswithwhales says:

    Slightly off topic here, but I was reading up on the Argo system and came across this article from NASA EO on “correcting ocean cooling”:

    There are 4 pages in the article.

    I was just wondering if you have read the piece and made any comments?

    I found it worrying that the cooling found was corrected apparently because it didn’t agree with the models, and that the “bump” in the 1979-80s was also corrected because it could not be explained and also didn’t agree with models.

    It didn’t leave me any wiser trying to understand the science behind the presented Argo data.

  3. Bob Tisdale says:

    batheswithwhales: Yup, I’ve discussed the corrections/adjustments to the OHC data a couple of times. The most recent is here:

  4. batheswithwhales says:

    Thanks, Bob.

  5. Greg Goodman says:

    Bob, now the graph is corrected, the one thing that stands out to me is how the data puts the lie on supposed volcanic cooling impact.

    el chichon 1982 , something dips a bit later but probably too late to attribute to eruption. Model seems to pre-empt the effect by two years, so it does not add up. If the dip was due to eruption it was compensated for shortly afterwards, there is no permanent offset.

    Mt Pinatubo 1991: another massive drop in modelled expectations that just does not happen in data. Nothing more than the usual ups and downs.

    Mount Agung 1963 eruption : biggest impact in the models and honestly NOTHING in the data, air or SST.

    This is important since the assumed cooling effect allows for an exaggerated CO2 forcing. The fact that there has been no significant volcanic activity since Mt P has brought this to the fore. The lack of warming predicted by exaggerated AGW ‘compensated’ by fictitious volcanic cooling can only work if both continue. Luckily volcanoes have been quiet allowing this false attribution to become evident.

    If remove this unwarranted cooling and put in the real physically based CO2 effect without amplifying it with rigged cloud parameters, the whole thing may start working a bit better.

  6. Bob Tisdale says:

    Greg Goodman: Thanks for the comment. You have to also consider ENSO when looking at the impacts of volcanos. Mount Agung was counteracted by the moderately strong 1963/64 El Nino. El Chichon in 1982 was counteracted by the very strong El Nino of 1982/83. And Mount Pinatubo in 1991 was counteracted in part by the moderately strong El Nino of 1991/92. Therefore, we wouldn’t expect NMAT and SST anomalies to dip as far as the model mean in response to the volcanic eruptions.

    With respect to your closing comment of “If remove this unwarranted cooling and put in the real physically based CO2 effect without amplifying it with rigged cloud parameters, the whole thing may start working a bit better,” you’re looking at global data and assuming the warming is caused by CO2, but there’s no evidence in the sea surface temperature records or ocean heat content data to indicate any influence of manmade greenhouse gases: See:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s