Guest Post by Bob Tisdale
In this post, we’re going to discuss one of the three adjustments with the greatest impacts on the new NOAA “pause buster” sea surface temperature dataset. That is, I’m going to discuss and illustrate that NOAA based one of their recent adjustments (the adjustment with the greatest impact during the slowdown in global warming) on the outputs of one climate model…clarification: one
obsolete earlier-generation climate model…and that if we examine the consensus of the latest generation of climate models, we see that NOAA may have adjusted away a relationship that the consensus of newer models indicates should exist…assuming that yet another of NOAA’s assumptions is correct.
NOAA revised their global surface temperature product earlier this year to show more global warming during the post-1998 period. Those data manipulations supposedly ended the slowdown in global warming over that period. The changes to NOAA’s global surface temperature product were presented in three studies. The greatest changes were made to the sea surface temperature portion, and those changes were discussed in the papers Huang et al. (2015) and Liu et al. (2015). Those two studies published in February 2015 slipped by the mainstream media, unnoticed. The alarmists, however, did not miss the Karl et al. (2015) paper Possible Artifacts of Data Biases in the Recent Global Warming Hiatus (paywalled, full paper here). While the revisions included changes to the NOAA global surface temperature product since its start in the 1850s, Karl et al. (2015) focused on the periods of 1998 to 2014 and 2000 to 2014…periods during which NOAA had adjusted the sea surface temperature portion to the extent that they claimed the slowdown in global warming never existed.
The claims were misdirection of course, because, even with NOAA’s changes to the surface temperature record, there is still an ever-growing difference between observations and climate model-based predictions of global warming. In other words, the manipulations of the NOAA global surface temperature records decreased the difference between models and data, but those modifications did not eliminate them.
U.S. Congressperson Lamar Smith is the chairman of the House of Representatives Committee on Science, Space and Technology. Like many persons, Representative Smith questions NOAA’s changes to their surface temperature record. Recently, Representative Smith has been formally requesting NOAA emails regarding the changes to the global surface temperature record, and, as of yet, Dr. Kathryn Sullivan of NOAA has not produced those emails. There are a multitude of news articles about the Smith-NOAA dispute, but they are not the subject of this post.
This post discusses the primary adjustment that had the greatest impact on the “pause busting” nature of the new NOAA global sea surface temperature reconstruction. My finding is that NOAA may have adjusted away a difference between ship-based sea surface temperature measurements and night marine air temperature that should exist according to the consensus of the climate models used in the latest IPCC 5th Assessment Report. Once again, this assumes that yet another of NOAA’s assumptions is correct.
ADDITIONAL BACKGROUND: OCEAN SURFACE TEMPERATURE BIAS CORRECTIONS
Different methods were used to sample sea surface temperatures since the start of the surface temperature record in the mid-1800s: buckets of different materials (wood, canvas, insulated), engine room intakes, and buoys (moored and floating). Additionally, the temperature of the air above ocean surfaces has been recorded aboard ships…a metric called marine air temperature. Each method has its own sampling errors, uncertainties and biases. (For those new to the subject, see Kent et al. (2010) Effects of instrumentation changes on ocean surface temperature measured in situ. It has a detailed and relatively easy-to-understand overview. Also see Kennedy (2014) A review of uncertainty in in situ measurements and data sets of sea-surface temperature. A copy of the submitted paper is here.)
Beginning in the mid-1990s, data suppliers have been accounting for the biases that come from the different types of buckets and from the transition from buckets to ship inlets. Additionally, the ship-buoy bias has also been discussed in papers since the early 2000s. The ocean temperatures sampled from engine-room intakes (ERI) on ships are slightly warmer than the temperatures sampled by buoys. But data suppliers had not attempted to account for the ship-buoy bias until recently because of the large uncertainties of the ship-inlet and buoy temperature measurements and due to the massive uncertainties of the ship-buoy bias. (See Reynolds et al. 2002. Also see Kennedy et al. (2011) Part 1 and Part 2 for documentation of those uncertainties.)
For their new “pause-buster” data, NOAA:
- adjusted the ship data using the UKMO HadNMAT2 night marine air temperature data as a reference. This of course assumes the night marine air temperature data vary at the same rates as sea surface temperatures (both day and night).
- adjusted the buoy data to account for the difference in temperature between ship inlets and buoys where both exist. The effects of this would vary over the past few decades because the ratio of buoy and ship-based measurements changed with time. There has been an increase in buoy samples and a decrease in ship-based measurements.
- weighted the buoy data more than ship data during periods when both ship and buoy data exist. This was done to account for the different accuracies of the ship and buoy temperature measurements…buoy data have been shown to be more accurate than the ship inlet data.
According to Karl et al. (2015) the adjustments to the ship data (based on night marine air temperature data) had the greatest impact on the trends from 2000 to 2014:
Of the 11 improvements in ERSST version 4 (13), the continuation of the ship correction had the largest impact on trends for the 2000-2014 time period, accounting for 0.030°C of the 0.064°C trend difference with version 3b. (The buoy offset correction contributed 0.014°C dec-1 to the difference, and the additional weight given to the buoys because of their greater accuracy contributed 0.012°C dec-1. See supplementary materials for details.)
In this post, we’re going to overlook the ship-buoy bias adjustment and the weighting of buoys and ship inlet data, they’ve been discussed in numerous other blog posts, and we’re going to concentrate on…
NOAA’S ASSUMPTIONS FOR USING NIGHT MARINE AIR TEMPERATURE DATA TO ADJUST SHIP-BASED SEA SURFACE TEMPERATURE MEASUREMENTS
NOAA made numerous assumptions to justify using night marine air temperature data to adjust the ship-based sea surface temperature measurements. They’ve listed them in Huang et al. (2015) paper Extended Reconstructed Sea Surface Temperature Version 4 (ERSST.v4) – Part I: Upgrades and Intercomparisons. There they write on page 919 (where NMAT is night marine air temperature and SST is sea surface temperature):
The bias adjustment for ship SSTs in ERSST.v4 is originally proposed by Smith and Reynolds (2002) and involves using NMAT as a reference. NMAT is selected because the differences from SST are more stable than daytime marine air temperatures, which can have a large range due to solar heating of the ships decks and of the instruments themselves. To formulate the bias adjustment, however, it is necessary to assume that
1) the difference between SST and NMAT is near constant during the climatological period (1971–2000);
2) the climatological difference of SST and NMAT is constant in other periods;
3) the NMAT is less biased (more homogeneous) than the SST data to which it is being compared;
4) the mix of SST measurement methods (bucket or ERI) is invariant across the global oceans, and the spatial pattern of biases follows the climatological difference of SST and NMAT in the modern time (1971–2000); and
5) biases vary relatively slowly and smoothly with time.
We only need to examine the first two of NOAA’s assumptions.
NOAA then went on two describe how they tested their first two assumptions:
To test the first two assumptions, which are assuming broad physical coherence between two highly correlated but physically distinct measurands, the average difference between SST and near-surface air temperature (SAT) of day and night at 2m is calculated by subsampling monthly outputs of the GFDL CM2.1 coupled model with monthly observation masks from 1875 to 2000 (Fig. 4). The model SAT is used since the model bias is assumed to be the same during daytime and nighttime. It is found that the first two assumptions are valid since the model simulations indicate that the difference of SST and SAT is near constant and its linear trend is weak in all four different latitudinal zones (Fig. 4)…
Huang et al. are not very specific about the meaning of “monthly observation masks”. Does it mean that NOAA has masked all grids that do not contain ship data? I suspect that’s the case.
IMPORTANT NOTE: In the following, I have not taken the step of masking the grids without ship data, because I’m attempting to illustrate the global (without polar oceans) climate model-expected relationship between marine air temperature and sea surface temperature. [End note.]
Curiously, NOAA is assuming the modeled relationship between sea surface temperature and marine air temperature (both day and night) is the same as the relationship between night marine air temperature and sea surface temperature, the latter of which includes both day and night measurements. That is, NOAA is using night marine air temperature (not marine air temperature, both day and night) to adjust sea surface temperature, yet they’re presenting modeled sea surface temperatures and marine air temperatures (both day and night) for justification. Is this addressed by “The model SAT is used since the model bias is assumed to be the same during daytime and nighttime”? That seems to be the case.
My Figure 1 is Figure 4 from Huang et al (2015).
First, note that the climate model-based graph in their Figure 1 ends in 2000. That’s odd because we’re very interested in the period of 2000 to 2014. Second, they listed the trends for 1875 to 2000 in the caption for their Figure 4, but they failed to show the trend for their climatological period of 1971 to 2000 (See their assumption 1 above).
Third, and very important, the GFDL CM 2.1 coupled ocean-atmosphere climate model from the NOAA Geophysical Fluid Dynamics Laboratory (GFDL) is a curious choice for use in a study in 2015. On the GFDL webpage about the new and improved GFDL CM3 model, they begin:
GFDL’s successful model, CM2.1 (Delworth et al., 2006), was used as the starting point to develop the next-generation CM3 coupled model.
In other words, the 10-year-old GFDL CM 2.1 climate model could be considered obsolete inasmuch as it has been replaced by the GFDL CM3. The GFDL CM 2.1 was supported by Delworth et al. (2006) GFDL’s CM2 Global Coupled Climate Models. Part I: Formulation and Simulation Characteristics. The abstract notes it was used for the IPCC’s 4th assessment report back in 2007.
We know from past experience that there can be wide ranges in the absolute values of surface temperatures from model to model and wide ranges in their trends. NOAA’s use of one earlier-generation climate model could lead anyone to believe NOAA selected (cherry-picked) that single earlier-generation GFDL CM 2.1 climate model because it provided them with the results they wanted.
So let’s take a look at the multi-model mean of all the new-and-improved climate models used by the IPCC for their 5th Assessment Report in 2013. Those models are stored in the CMIP5 archive, and their multi-model mean (along with the outputs of the individual models) are available at the KNMI Climate Explorer. The multi-model mean (the average of the models) basically represents the consensus (the groupthink) of the climate modeling groups on how sea surface temperatures and marine air temperatures (in this example) respond to the climate forcings used to drive the models. Due to the wide ranges of climate model outputs, using the average of all of the models in the CMIP5 archive (the multi-model mean) assures that we cannot be accused of cherry-picking one climate model that supports an agenda. And, to give you an advanced warning, we only need to examine the climate model outputs for the global oceans, excluding the polar oceans (60S-60N); that is we won’t need to examine the additional subdivisions presented in Figure 4 from Huang et al (2015) (my Figure 1).
We’ll start with period of 1875 to 2000 shown in Huang et al. Figure 4. The top graph in Figure 2 presents the modeled difference between the simulated global (60S-60N) sea surface temperatures and marine air temperatures during NOAA’s selected period of 1875 to 2000, with the marine air temperatures (TAS, with land surfaces masked) subtracted from the sea surface temperatures. Over this period, only the historic climate forcings were used by the modeling groups. For NOAA’s selected period of 1875 to 2000, the consensus of the latest generation of climate models show the modeled difference between the global (60S-60N) sea surface and marine air temperatures declining at a rate of -0.008 deg C/decade…exactly the same as the -0.008 deg C/decade (-0.08 deg C/century) claimed by NOAA for the period of 1875-2000. That would appear at first to confirm the results of Huang et al. (
2025) (2015). The bottom graph in Figure 2 illustrates, for the same timeframe, the multi-model mean of the simulated sea surface temperatures and marine air temperatures, which serve as the bases for the top graph. I’ve included the model outputs for 2000 to 2014 (dashed curves) in both graphs as a reference as well.
Note, however, in the top graph of Figure 2 there is a noticeable change in trend starting in the mid-1970s, with a sharp decline thereafter.
So let’s examine the trend for NOAA’s climatological period of 1971 to 2000. See Figure 3. For NOAA’s climatological period of 1971 to 2000, the consensus of the latest generation of climate models show the modeled difference between the global (60S-60N) sea surface and marine air temperatures declining at an increased rate of -0.021 deg C/decade…far greater than the -0.008 deg C/decade (-0.08 deg C/century) claimed by NOAA for the period of 1875-2000.
Referring to the bottom graph of Figure 3, the modeled marine air temperatures are warming slightly faster than the modeled sea surface temperatures over NOAA’s climatological period. In other words, the consensus of the new-and-improved climate models contradicts the NOAA (Huang et al. 2015) finding about their first assumption for using marine air temperatures to adjust the biases in sea surface temperatures measured from ships…or any sea surface temperature data.
But we’re really not interested in the period of 1971 to 2000. We’re more interested in the period of 2000 to 2014, because it was the period NOAA (Karl et al. 2015) used for their claim, “the continuation of the ship correction had the largest impact on trends for the 2000-2014 time period, accounting for 0.030°C of the 0.064°C trend difference with version 3b.”
Figure 4 is similar to Figures 2 and 3, but in Figure 4 we’re illustrating the model results for the period of 2000 to 2014. The only difference is,
these models now the models also include projected forcings, in addition to historic forcings. The transition from historic to projected forcings occurs at 2005/2006 in most models. (It’s not a feature I selected.) But I did select the RCP6.0 forcing scenario, because I did not want to be accused of cherry-picking the worst-case RCP8.5 scenario.
Not too surprisingly, in the top graph of Figure 4, the multi-model mean (the consensus) of the new-and-improved climate models shows a drop in the temperature difference between the sea surface temperatures and marine air temperatures. In fact, the -0.027 deg C/decade trend of the modeled temperature difference is comparable to the 0.030 deg C/decade adjustment NOAA made (using night marine air temperatures) to the ship data for the period of 2000 to 2014.
That is, using the newer models, the consensus (the groupthink) of the modeling groups expects that marine air would warm faster than the surfaces of the oceans for 2000 to 2014…at a rate that is comparable to the “correction” applied to the ship data using night marine air temperature. This invalidates the second of NOAA’s assumptions that “the climatological difference of SST and NMAT is constant in other periods”. It clearly is not.
Simply put, NOAA appears to have adjusted out a difference between ship temperature measurements and night marine air temperature that should exist according to the consensus of the newer climate models. That assumes that NOAA’s other grand assumption…“The model SAT is used since the model bias is assumed to be the same during daytime and nighttime”…is correct.
Let’s put that into perspective. NOAA appears to have assumed that night marine air temperatures warmed at the same rate as marine air temperature (day and night). They used marine air temperatures in place of night marine air temperatures to verify their first two assumptions. Also consider that the consensus of the latest generation of climate models indicate that marine air temperatures should be warming at a higher rate than sea surface temperatures since about the mid-1970s. Yet, taking those considerations into account, NOAA’s end product, their new “pause-buster” ERSST.v4 data, shows the opposite relationship. Their new sea surface temperature data show a noticeably higher warming rate than the night marine air temperature data (HadNMAT2) they used as a reference. See Figure 5, which is Figure 1 from the post Open Letter to Tom Karl of NOAA/NCEI Regarding “Hiatus Busting” Paper.
Note: That graph ends in 2010 because the HadNMAT2 data end then. That of course raises the question: how does NOAA (unjustifiably) adjust the ship data after 2010? [End note.]
Bottom line: according to the latest generation of climate models, marine air temperature should warm faster than sea surface temperatures, but not in NOAA’s much-tinkered-with world.
Again, my illustrations show the expected global (60S-60N) relationships between marine air temperatures (day and night) and sea surface temperatures because NOAA used them to verify a relationship between night marine air temperatures and sea surface temperatures.
Also, I have not masked the model results so that they only include the grids with ship-based sea surface temperatures, as NOAA appears to have done. But, according to the consensus of the climate models used by the IPCC for their 5th Assessment Report, the expected relationship globally (60S-60N) is that Marine Air Temperature has warmed faster than Sea Surface Temperatures since the mid-1970s.
I suspect, when Congressman Lamar Smith’s whistleblowers are concerned about rushing the Karl et al. (2015) study “before appropriate reviews of the underlying science and new methodologies used”, they’re discussing:
- the uncertainties of the bias adjustments,
- the uncertainties of the data,
- of course the basic methodologies, including how NOAA distributed those adjustments around the oceans, and
- most importantly, for the “underlying science”, how NOAA appears to have adjusted out a difference between ship temperature measurements and night marine air temperature that should exist according to the consensus of the newer climate models—once again assuming that NOAA’s other grand assumption…“The model SAT is used since the model bias is assumed to be the same during daytime and nighttime”…is correct.
Maybe, in time, Dr. Sullivan of NOAA will produce the emails requested by Representative Smith so that we can confirm my suspicions and the suspicions of many others.