Guest post by Willis Eschenbach
Bob Tisdale has discussed a variety of issues with the hemispheric and basis-by-basin Levitus summary of the ARGO data in his excellent post here on WUWT. I wanted to take a larger look at at the global ocean data, to provide it with some context. After following a variety of blind paths and dry holes, I arrived at Figure 1. This shows the Levitus data, along with several datasets for both the sea surface temperature (SST) and the land surface air temperatures.
Figure 1. Three land surface air temperature datasets, three ocean surface temperature datasets, and two measures of the deeper ocean. Datasets from KNMI, except the Levitus ocean data. Levitus data have been converted from ocean heat content values to temperature values.
Unfortunately, we only have pentadal (5-year centered average) data for the deeper ocean layer from 1955, no annual data for the deeper layer is available before 2005. So I have shown the other data as pentadal averages as well, so that they can be compared directly. Now, I have no profound conclusions from this, but there are some curiosities about these results.
The most obvious oddity involves the sea surface temperature data. Two of the datasets, the ERSST and the HadISST, are quite similar, but are offset over much of their length by an amount that varies in an odd stepwise fashion. In particular, they diverge significantly in the period 1997-2001. They have the same small peaks and valleys, but the land is going up while the sea temperature is dropping. Figure 2 shows this in more detail:
Figure 2. Sea surface temperatures (SSTs) from three different sources. The vertical axis scale is different from Figure 1.
On the other hand, Figure 2 also shows that the ICOADS dataset runs almost exactly in parallel with the HadISST data up until 1997, when it shifts gears and goes up to agree with the ERSST dataset.
I have absolutely no idea what happened in 1997, or in earlier periods, which allowed the three datasets to vary so much. Always more questions than answers.
The next oddity involves the changes in the land and ocean surface temperatures. Figure 3 is like Figure 1 but leaves out the sub-surface temperatures:
Figure 3. Three land surface air temperature and three ocean surface temperature datasets.
The oddity here is two-fold: what happened around 1977 or so, and what happened around 2002 or so? From 1977 on, after twenty years with no significant land or sea temperature change, both land and sea temperatures start to rise. In addition, the land temperature rises much more rapidly than the SST. Then around 1998, SSTs start to level off, and they peak in 2002 and start to fall. The land temperatures start falling shortly thereafter. Why, on all counts?
Now, I don’t know the answer to these questions either. However, in this modern world there’s a term called a “WAG”, which means a wild-assed guess. However, scientists don’t provide those. Instead, there is a refined version called a “SWAG”, or a scientific wild-assed guess.
So here’s my SWAG about what we’re seeing here. I think at least some of this is the effect of the Pacific Decadal Oscillation, or PDO.
The PDO is a large-scale reorganization of the flow of the North Pacific Ocean. I think that, like other large-scale patterns such as the El Nino/La Nina alterations, these all function to regulate the temperature. It is obvious that one of the two major patterns (positive or negative PDO) must be more efficient than the other one at cooling the planet. The transfer of heat via the oceans from the equator to the poles is one of the two major pathways by which the tropics moves heat (the other is via the atmosphere). One of the two PDO states must be better at cooling the planet.
I hypothesize that this alteration between PDO states is one of the ways by which the planet maintains such a regular temperature. Unfortunately, I have no idea how to support that hypothesis using observational data.
A final oddity is the decoupling of the land temperatures from the ocean temperatures. Figure 4 shows a longer version of the same data, this time extending back to the 1870’s:
Figure 4. Long-term land air temperature and sea surface temperature records, expressed as an anomaly around the 1900-1980 average.
The oddity to me in Figure 4 is that since about 1974, the land temperatures have been rising much faster than the sea surface temperatures. The SST records and the land temperature records run quite close to each other up until about 1975 or so, but after that land temperatures have been rising much more quickly, while the sea surface temperatures (and to a greater extent the sub-surface temperatures) have not followed suit.
Once again, I’m out of explanations for the decoupling of the temperatures. Clearly it’s not CO2. Urban heat islands? Aerosols reducing the clouds over the land? I have no idea. All suggestions welcome. That’s the beauty of settled science, there are always unexplained questions to
Best regards to everyone,
w.
THE DATA
Levitus provides two different global pentadal datasets. One covers the layer from 0 to 700 m depth, and the other covers a thicker layer, from 0 to 2000 m depth. I wanted to see what was happening in the “missing link”, which is the data covering the deeper layer, from 700m to 2000m depth.
The good news is that since we are dealing in oceanic heat content, the heat content of the lower layer is simply the difference between the two datasets.
My spreadsheet with the results is here.
Does the difference between land and ocean after the 70s perhaps involve the ‘adjustments’ made to land station data about which much controversy yet rages?
I’ve been convinced for years that the PDO is indeed the key. And it’s so poorly understood!
“Unfortunately, I have no idea how to support that hypothesis using observational data.”
You’d never make a good Climatologist….
Need to add the satellite post 1979 data as well
CRU, HAD, GISS, and Best.. take with a grain of salt. Highly manipulated ex-data
“The SST records and the land temperature records run quite close to each other up until about 1975 or so, but after that land temperatures have been rising much more quickly, while the sea surface temperatures (and to a greater extent the sub-surface temperatures) have not followed suit”.
I think it has much to do with the different amounts of energy (differential heat inertia?) it takes to heat land and water. It takes a lot of energy to heat water, and much less to heat land.
I think you will find that generally, during cold periods, land temperatures decouples from ocean temperatures downwards, as in around 1880 in Figure 4, and during warm periods it decouples from the ocean upwards, as after the 1970s in Figure 1.
Afternoon sea breezes are a good illustration, where winds increase in the afternoon towards the land because the water takes longer to heat up than the land. There is also an opposite effect at many places at night when winds blow offshore, as well as often also in many places in winter-when offshore breezes are more common. I used to watch these effects growing up on Sydney Harbour (where I also note there is virtually/no change in high water level over the last 30 years). Katabatic breezes from mountains are the same thing, they blow down at night and up during the day.
Because seabreezes are by definition, only where the land meets the sea, this shouldn’t really show up in regional wind or temperature trends. Sea breezes die off only a few kilometres inland.
All this could have been pointed out to the IPCC, if any of them had done some simple observations whilst growing up near the sea, rather than panicking over warming temperatures since the 1970s, from delayed effects of high solar activity during the 20th century.
ps, Nothing odd about the Land temps coinciding in figure 4, its called “collusion”
Anyone care to discuss the advantages and disadvantages of using a five-year centered mean to compare the above data?
Ian Bryce
Willis, it may be a good idea to look at the temperatures for small towns, medium sized towns, and large cities as Goodridge did for California, and Warwick Hughes for Australia. It may also be a good idea to separate the maximums from the minimums.
Regards,
Ian Bryce
Hmmm anomaly graphs. Maybe its the changing base we measure it all from (hey, does this mean we are actually cooling!!?)
Our global temperature is often depicted as an ‘anomaly’ ie +0.7 C …. so much above or below the mean global temperature.
But, the accepted ‘mean global temperature’ has apparently changed with time:
1988: 15.4°C
Der Spiegel, based on data from NASA.
http://wissen.spiegel.de/wissen/image/show.html?did=13529172&aref=image036/2006/05/15/cq-sp198802801580159.pdf&thumb=false
1990: 15.5°C
James Hansen and 5 other leading scientists claimed the global mean surface temperature was 15.5°C. Also Prof. Christian Schönwiese claimed the same in his book “Klima im Wandel“, pages 73, 74 and 136. 15.5°C is also the figure given by a 1992 German government report, based on satellite data.
1999 14.6°C
Global and Hemispheric Temperature Anomalies – Land and Marine Instrumental Records Jones Parker Osborn and Briffa http://cdiac.ornl.gov/trends/temp/jonescru/jones.html
2004: 14.5°C
Professors Hans Schellnhuber and Stefan Rahmstorf in their book: “Der Klimawandel”, 1st edition, 2006, p 37, based on surface station data from the Hadley Center.
2007: 14.5°C
The IPCC WG1 AR4 (pg 6 of bmbf.de/pub/IPCC2007.pdf)
2010: 14.5°C
Professors Schellnhuber and Rahmstorf in their book: Der Klimawandel, 7th edition, 2012, pg 37 based on surface station data.
http://cdiac.ornl.gov/trends/temp/jonescru/jones.html
2012 14.0 °C
Press Release No. 943 World Meteorological Society Globally-averaged temperatures in 2011 were estimated to be 0.40° Centigrade above the 1961-1990 annual average of 14°C. http://www.wmo.int/pages/mediacentre/press_releases/pr_943_en.html
2013 Wikipedia: 14.0°C
Absolute temperatures for the Earth’s average surface temperature have been derived, with a best estimate of roughly 14 °C (57.2 °F).[11] However, the correct temperature could easily be anywhere between 13.3 and 14.4°C (56 and 58 °F) and uncertainty increases at smaller (non-global)
http://en.wikipedia.org/wiki/Instrumental_temperature_record
http://notrickszone.com/2013/04/21/coming-ice-age-according-to-leading-experts-global-mean-temperature-has-dropped-1c-since-1990/
The decoherence of the land and sea surface temperature data sets is quite interesting. I wonder, might it have something to do with the source of heating? I would guess that it’s a lot easier for a warm ocean to heat a cold atmosphere (one heat reservoir, after all, is massively larger than the other) than for a warm atmosphere to heat a cold ocean (where direct thermal transfer has a minute effect, and heating must rely on a decrease in convection of heat away from the ocean, with impinging visible light pumping most of the extra energy in to the water). Might it be, then, that during the first half of the 20th century the ocean was the entity that was warming and that was driving the rise in atmospheric temperatures, while in the latter half it’s the atmosphere that’s been the initial receiver of extra heat?
…’Course, that doesn’t explain the sharp drop in land temperatures at the very beginning of the record, but on the other hand that might be either a result of the 1883 eruption of Krakatau or simply of BEST’s huge error bars that far back in the reconstruction.
Barclay E MacDonald says:
May 5, 2013 at 7:56 pm
Anyone care to discuss the advantages and disadvantages of using a five-year centered mean to compare the above data?
Is what I have plotted below what you mean? It is interesting that when each month is plotted, there is a huge spike in 1998, but with the mean of 60, there is actually a small dip in 1998 and slightly higher areas on either side; in other words, we get a completely different picture.
http://www.woodfortrees.org/plot/rss/from:1980/plot/rss/from:1980/mean:60
Willis: I’m surprised you missed one very fundamental difference between the land-based record since 1970, and the sea-temperatures since 1970: CO2 itself IS a part of the increase in land temperature change because of albedo: greater plant growth earlier in the season over larger land areas WILL darken the plant, and that lower albdeo WILL tend to heat up the air masses flowing over the land. Look at desert (rocky) land thermometers? have they gone up as much a wetter, more plant-friendly areas? If EVERY plant on the face of the planet is growing 15 to 27% faster, greener, higher, taller, wider and more brushy, is not less solar energy reflecting back from previous light-colored rocks, dirt, and grassland?
On the other hand, sea-growing plants are also benefiting from higher CO2 levels sicne the mid50’s and 60’s, but they will not change the sea “color” and albedo. Hence, little change in sea temperatures.
Dr Spencer has said all of the measured temperature change could be explained by a 2% change in cloud coverage worldwide? could not a 1% darker land albedo be equally influential?
What limits reported temperature rise? Being directly constrained in divergence from satellite-measured temperature rise is not particularly the answer, as those CRU/BEST land temperature plots claiming around 1 degree Celsius temperature rise since the late 1970s already differ vastly from more independent satellite data (even for likewise land and not the whole planet) showing under half that in temperature rise since then.*
* For example: http://woodfortrees.org/plot/rss-land/from:1979/mean:12
While moderating daring during times like the 2010 winter or this spring, overall they have been getting away with that divergence quite successfully, down to only a small minority of people even noticing the difference, despite it being huge in context.
Nor is vast contradiction to data published before the global warming ideological-political movement much of a limit. Especially if quantitatively reading the scales on the graphs, the plots at the top of this article are grossly different from, for instance:
data in a 1976 National Geographic:
http://img240.imagevenue.com/img.php?image=40530_DSCN1557_nat_geog_1976_1200x900_122_75lo.JPG
temperature history towards the bottom of this (click to enlarge):
http://s7.postimage.org/69qd0llcr/intermediate.gif
So a curious question is why hasn’t temperature change in the ocean data been adjusted to be multiple times higher too? The answer might be that, when the public has been told about great harm supposed to result upon multiple degrees Celsius global temperature rise, a claim of being most of the way there already would be too blatantly contradictory.
Tell the average person in the general public that a certain subpart of the world rose in temperature by a degree or two, and that person isn’t likely to look up other data sources to discover the claim to be wrong. Yet that doesn’t mean even the average person could be gotten to simultaneously believe in terrible harm at, for instance, a point of 2 degrees Celsius global temperature rise while meanwhile believing global temperature rose by most of that over the past 3 decades.
Of the portion of the public which believes in primarily-manmade global warming, a majority of them believe such is likely to destroy human civilization within a century (as seen in polls, since, while around 23% of all U.S. voters believed that in a 2009 poll,** the 23% is a bit more half of the percentage who believe in predominately AGW). Maintaining beliefs like that, the fruit of much labor by media outlets and by activists flocking to institutions superficially labeled as for environmental study, is partially mistake-tolerant but still requires a little care in management to not too overstate claims too soon. To take an extreme example, for instance, it wouldn’t work to just suddenly say global temperatures already rose by 5 degrees in the past decade.
** http://www.rasmussenreports.com/public_content/politics/current_events/environment_energy/23_fear_global_warming_will_end_world_soon
Hi Willis,
Once again some very nice work with a humility that is rare in science these days.
I was asking myself about possible jet-stream relationships that would shift weather patterns north/south… and cause temperature readings to swing… so I examined US agricultural output at:
http://www.ers.usda.gov/data-products/agricultural-productivity-in-the-us.aspx#28247
Table 4 (Crop output- if you add your own “Total for All States” column in a pivot table) shows a period of flattened output starting around 1980-1990… but it does not look that exciting.
Then I went off on a tanget … agricultural energy use… when you look at Table 14, (Energy Input) there is a significant drop starting around 1980, but then a gradual rise from 1986 through 2004… so these might be associated with total economic output for the country… if CO2 is “to blame” then we should see a sharp spike in temperatures related to energy use for this major industry… but again, not that exciting…
Russ
A land surface temperature of the order of half a degree due to land-use change and associated changes in hydrology does not seem impossible. These changes have been big enough to render other explanations highly uncertain.
Clearly it’s not … Aerosols reducing the clouds over the land?
Reduced aerosol seeded clouds would, by increasing solar insolation, increase land surface temperatures faster than SSTs, because water has a much greater thermal capacity than air. That is, the same increase in solar insolation would increase air temperatures by much more than water temperatures.
The twentieth century experienced two periods of warming – about 1910 to 1940 and about 1975 to 2000 or so. Each warming episode involved about 0.5K at similar rates, and each coincided with polarity changes in the PDO index.
Monckton (among many) has argued that the similarity of the two events suggests a similar cause, and that both must be fundamentally natural rather than human-CO2-caused because CO2 did not seriously change until WWII and post-war development.
Figure 4 [above] shows a much steeper rise in BEST and CRUTEM air temps where both deviate strongly from sea temps in about 1990.
Given past behavior, I’m inclined to infer the deviation reflects biases in the adjustments of both BEST and CRUTEM datasets (partly due to changes in numbers/distributions of reporting stations alone, not necessarily malice).
It is hard/impossible to believe that air temps could increase faster than ocean heat content on decadal scales.
thingodonta says:
May 5, 2013 at 7:46 pm
Because seabreezes are by definition, only where the land meets the sea, this shouldn’t really show up in regional wind or temperature trends. Sea breezes die off only a few kilometres inland.
How far sea breezes penetrate inland depends on the ocean land temperature differential over time. In many areas during summer, the land is warmer than the ocean for most of the 24 hours, and sea breezes will penetrate many hundreds of kilometers inland, and are called monsoons.
Land/ocean temperature differentials can produce global scale effects. Winter cold in Siberia contributes to the strength of the Australian summer monsoon.
Yes, thank you Werner. I was thinking about 1998 and the effect of smoothing on that and similar, possible, outliers, and how such smoothing might affect our ability to accurately discern relationships between the different data sets that are influenced by different cycles or different lag times. No clever insights, just trying to better understand.
Werner Brozek says:
Barclay E MacDonald says:
May 5, 2013 at 7:56 pm
Anyone care to discuss the advantages and disadvantages of using a five-year centered mean to compare the above data?
Is what I have plotted below what you mean? It is interesting that when each month is plotted, there is a huge spike in 1998, but with the mean of 60, there is actually a small dip in 1998 and slightly higher areas on either side; in other words, we get a completely different picture.
http://www.woodfortrees.org/plot/rss/from:1980/plot/rss/from:1980/mean:60
========
THANK YOU Werner.
This is something I’ve been pointing out for years. These damned running mean “smoothers” that everyone is to keen on cause serious, unintended distortions of the data.
Just look at all the distortions by running mean filter in the following plot of some randomly created synthetic data: inverted peaks like the one you pointed out above (1940), peaks bent left and right (1958), even a whole series of oscillations getting turned upside down ( circa 1970 )
http://tinypic.com/view.php?pic=351v6a1&s=5
I understand Willis’ reason for using them here in order to compare like with like but you then end up discussing artefacts of the filtering instead of the data ( as if the data did not have enough problems itself ).
The actual change in N. Pacific happened in 1974/75 but seems like 77 here due to the filtering.
In view of the above it would probably have been better if Willis had used his usual gaussian filter for everything except the first plot where he included OHC data that was only available as “pentads”.
Adding a triple running mean, with a base period of 30 months instead of 60, to Werner’s plot.
http://www.woodfortrees.org/plot/rss/from:1980/plot/rss/from:1980/mean:60/plot/rss/from:1980/mean:30/mean:22/mean:17
The key to this filter is to shorten the running mean by a factor of 1.3371 at each step to eliminated the negative lobes in a straight running mean. This has a decent frequency response similar to a gaussian.
It is interesting to note that much of the variation in the running mean is actually upside down !!
Willis
I noticed something interesting related to this in the form of sea level rise.
Check this out for the latest sea level rise from the university of Colorado.
http://sealevel.colorado.edu/files/2013_rel3/sl_ns_global.png
Now look very closely at the previous version.
http://sealevel.colorado.edu/files/2013_rel2/sl_ns_global.png
The subtle difference is that the range of variation of the red line, representing the TOPEX data now has considerably less uncertainty.
Anything you know about?
Ah, this is from the Colorado folks…
2013 Release 3 (2013-04-20):
Added Jason-2/OSTM GDR cycles 168-169.
Reprocessed TOPEX cycles 104 and 292 which were marked missing in previous releases.
Fixed bug that misapplied tide model correction to TOPEX cycles. Result is reduction of corrected SSH variance during TOPEX.
Willis: “Once again, I’m out of explanations for the decoupling of the temperatures. ”
The idea that they decouple after a certain point may be spurious to a degree.
The two are aligned here simply by definition of 1900-1980 reference average for calculating ‘anomalies’ which _defines_ them as both being around zero in that period. Looking closer, air temperatures are more volatile in 50s and 60s as well. BEST shows a lot more variation pre 1900 too.
Where the oddity may actually be, is in the period where SST and air temps show the _similar_ variability rather than where they differ.
It gets a little hidden in your figure 3 but ICOADS has more variability in the pre-WWII data than the more processed data sets.
One of my main points in looking at Hadley “bias corrections” to SST (that are largely incorporated in other SST datasets) what that the adjustments remove much of the variability from the climate data:
http://judithcurry.com/2012/03/15/on-the-adjustments-to-the-hadsst3-data-set-2/
This is definitely a good point your raise but the divergence may be a bit different to the way you are reading it.
Maybe plotting rate of change of each would make it clearer what is happening. (Without running mean, of course).