The Layers of Meaning in Levitus

Guest Post by Willis Eschenbach

There have been a lot of electrons sacrificed on the altar of the discussion of the Levitus ocean heat content data. The oddity seems to be that the deep ocean is gaining heat faster than the upper ocean. Here’s a typical graphic showing the issue:

global ocean heat content levitusFigure 1. Changes in the ocean heat content for two layers, 0-700 metres and 0-2000 metres. Values are pentadal (5-year) centered averages. SOURCE: NOAA/NODC

This week I got to ruminating about this graphic, and a number of similar graphics I’d seen. And yesterday I realized that it wasn’t showing what I thought it was showing. Let me illustrate what I mean.

I’ll start with an overview of the oceanic heat content (OHC) of the three layers that are provided by NOAA. These cover 0-100, 0-700, and 0-2000 metres depth. Figure 2 shows that data.

changes in ocean heat content noaa layersFigure 2. Annual changes in oceanic heat content for the 0-100, 0-700, and 0-2000 metre layers. 

Now, my problem was that when I looked at graphs like Figures 1 & 2, I thought that the deepest layer was gaining heat the fastest. And there’s been a lot of discussion about how that could be, and much speculation about the reason for the big increase in the deeper layers from 2001 onwards.

But yesterday I thought hey, wait a minute … those layers of the ocean overlap! They are not separate layers, they all extend to the surface.  So what we’re seeing in the deep 2000 metre level data is to some extent affected by what’s happening in the other levels. Yeah, I know, I should have seen it earlier, but I’m not gonna pretend.

The good news is that we’re measuring ocean heat content (OHC), so it’s very different from temperature. We can simply subtract the changes in the 700 metre level OHC from the 2000 metre level OHC changes, and what is left is the change in heat content for the layer from 700 metres down to 2000 metres. Can’t do that with temperature. Figure 3 shows the same OHC data as in Figure 2, except split out into distinct and separate layers, at the same scale. as Figure 2.

changes in ocean heat content separate layersFigure 3. Changes in oceanic heat content. The exact same data was used as in Figure 2, except it was split into three separate layers rather than three overlapping layers.

I was quite surprised by this result. Once I split the information up so that I could see the changes in each of the layers separately, much of the apparent change post-2001 disappeared. In Figure 2 there’s not a lot of change in 2001.

I also found it interesting that for most of the time covered by the study, all three layers picked up about the same amount of heat. Only in the last decade has the middle layer (100-700 m) picked up a bit more heat than the other two layers. I hadn’t expected them to warm and cool generally in unison as we see above.

Finally, I calculated the change in temperature for each of the levels. The nice thing about the ocean is that the temperature and heat content are mathematically related by the fact that it takes about 4 megajoules to warm a tonne of water by 1°C. This lets us convert from heat content to temperature and back as needed.

Remember that the three layers have very different volumes. So a terajoule of energy added to the shallow 0-100 metre layer will warm it more than the same terajoule of energy added to the more voluminous 700-2000 metre layer. Fortunately, NOAA also provided the ocean depths on a 1° x 1° grid, so we can calculate the volume of each of the layers. Once we know the volumes, we can calculate the temperature changes. Figure 4 shows the same data as in Figure 3, except expressed as a temperature change rather than as a change in heat content.

changes in sea surface and subsurface tempsFigure 4. Measurements of ocean temperatures at the surface and three sub-surface layers.

There are several interesting things about this plot of the temperature measurements.

First, as one might hope, we have relatively good agreement between the sea surface temperature (SST) and topmost layer (0-100 m). However, the annual and inter-annual swings in the upper 100 metres are larger than those at the surface … which seems somewhat strange to me. I’d have expected the surface to change more than the bulk.

Next, as we’d expect, the nearer to the surface, the greater the changes in temperature.

Finally, I’ve marked the eruptions of El Chichon and Pinatubo for future reference. I’ll come back to this in a subsequent post. For now, note that there is no visible effect of the volcanoes on any of the five different measurements of ocean temperatures. As near as we can tell from these measurements, the effect of the volcanoes on the ocean was below the limit of detection.

So … now that we have had a better overview of what’s happening to the various layers under the sea, are the changes in ocean heat content surprising?

I’d say not particularly. Yes, the middle layer (100-700 m) started warming in 1995. And yes, the lower layer (700-2000 m) followed suit starting in about 2001. But neither of these seem particularly surprising. I don’t have any explanation for them, but they do not seem to be unusual. It is possible, for example, that they represent the sub-surface changes associated with the gradual shift of the Pacific Decadal Oscillation from the positive to the negative phase. We know very little about the ocean depths, including how much we’d expect them to vary. And the records are short, too short to even show two PDO changes. It is clear, however, that the changes in heat content are not caused by CO2, at least directly.

Finally, we have to consider that the changes in the deeper layers may be an artifact. One obvious possible source is the integration of the Argo data into the Levitus analysis. The first Argo floats went in during the early nineties, and were added progressively over the next two decades. It is at least conceivable that some or all of the recent changes in the deeper layers are an artifact of the change in measuring methods.

Best regards to all,

w.

NOTES:

The data is from NOAA , except the ERSST and HadISST data, which are from KNMI.

The NOAA ocean depth data is here.

The R code to extract and calculate the volumes for the various Levitus layers is here.

0 0 votes
Article Rating

Discover more from Watts Up With That?

Subscribe to get the latest posts sent to your email.

81 Comments
Inline Feedbacks
View all comments
david douglass
May 10, 2013 12:44 pm

Willis
Could you be more explicit as to where the values of “0 to 100m ” OHC data shown in your figure 2 can be found?

Nic Lewis
May 10, 2013 1:17 pm

Willis
“the Levitus data contains the total volume of the ocean. They have it at 1.56e+9 cubic km”
According to a detailed analysis by Cotello et al in Environ. Sci. Technol, 2010, the voiume of the world oceans is 1,335,819,297 cubic km. Not that I believe it can be measured with that accuracy!
I’m not sure where your 1.56e+9 cubic km figure is to be found in the Levitus data or papers?

Ed, Mr. Jones
May 10, 2013 1:28 pm

Unknown variable: Heat added by geothermal vents.;

Nic Lewis
May 10, 2013 1:33 pm

Owen in GA says:May 10, 2013 at 9:07 am
“Bob Tisdale:
The adjustments to ARGO have me scratching my head. I don’t see anything in the engineering documents for ARGO that would indicate the system should have a cold bias, but someone adjusted it to correct for that. That someone needs to have a serious, introspective look at their integrity, because this looks like making changes to experimentally collected data for the purpose of supporting ones hypothesis. ”
The Barker et al 2011 paper “Pressure Sensor Drifts in Argo and Their Impacts” is quite interesting on this issue ( here – full paper paywalled, but available on request at http://ecite.utas.edu.au/76152 . The problem seems to be more with drfit in pressure, and therefore depth, measurement than with temperature measurement per se. Their statement that “A uniform depth error of 5 dbar globally produces a temperature bias that is greater than the observed ocean warming during the past 50 yr in the tropical and subtropical ocean and equals almost half of the observed warming in the higher latitudes when averaging between 0 and 300 m” is pretty shattering. That is a depth error of only 5 m, if I’ve got my units right.

Stephen Rasey
May 10, 2013 2:15 pm

Willis, kudos for comparing OHC (Joules) to temperature. Numbers such as 5*10^22 Joules looks enormous, but Joule is so humanly intangible a quantity on such an enormous object that the number is literally academic.
To get a better mental handle on the problem, I like to keep a couple of parameters in mind.
1 ZettaJoule = 1 ZJ = 1*10^21 Joules.
Area of the earth oceans (A) = 3.60*10^14 m^2 = 360 * 10^6 km^2
Mass of Earth’s Oceans to 2000m = (6.7 to 7.1)*10^20 kg (Depending upon how much of Area is deeper than 2000 m)
Heat to raise oceans 0-2000m 1 deg C = (2660 to 2850) ZJ/degK
Or crudely, as a mental conversion factor, almost 30 ZJ per 0.01 deg C temperature change in the top two kilometers of ocean. 0.01 deg C is an amount many of us believe (for a variety of reasons) no bigger than the precision available in the aggregate of data. (I refer to “Decimals of Precision” WUWT Jan 26, 2012.)

[Willis:]So, assuming there are no problems with my math, they are claiming that they can measure the temperature rise of the top mile of the global ocean to within 0.004°C per year. That seems way too small an error to me.

Willis’ Figure 2, shows 200 ZJ added to the 0-2000 meter ocean volume from 1955 to 2013. So that should be only 0.07 deg C. However, by the same chart, only 50 ZJ of heat are shown to have accumulated 0-2000m since the start of ARGO in 2003, which equates to less than 0.02 deg C, or 0.002 deg C/yr, even if you believe the precision of the measurement of ARGO worldwide averages.
Comment about Figure 4: Perhaps the most important events to mark on the chart are not volcanic eruptions but the start and end of the ARGO float constellation growth. Prior to 1/2 of the ARGO constellation launched, the amount of heat added below 700 m is pure “by-guess-and-by-golly”.
Frankly, we are looking at another “Hockey Stick” with undersampled, noisy data from 1955-2006(?) being given magnitudes more weight than they deserve compared to recent data. Realistic error bars need to be put on all the Levitus point data. I expect error bars prior to 1980 to exceed the top and bottom margins of the chart.

Stephen Rasey
May 10, 2013 3:39 pm

More on: I expect error bars prior to 1980 to exceed the top and bottom margins of the chart.
Who really believes we knew the average temperature of the world’s 700-2000 m oceans to +/- 0.1 deg C? That is, without the assumption that the deep water temperature doesn’t change from year to year. If we did, how could anyone justify the cost of ARGO to get the precision down to 0.004 deg C per year?
So let’s assume that we did have a precision of +/- 0.1 deg C.
That equates to +/- 270 ZJ = +/- 2.7*10^23 Joules,
or 150% of the size of the Y-axis in Figure 2.

chris y
May 10, 2013 4:42 pm

Willis- you say-
“So at the end, I get a heat content of 1.62e+18 tonnes times 10.9e+8 tonnes, or 1.8e+27. This is about 2.3 times the value you obtained.”
Thanks for checking my calculations.
I erred in using 1 J/gmC instead of the correct 4 J/gmC for the heat capacity. With this correction, I arrive at the same result as you (if I use your ocean volume estimate).

chris y
May 10, 2013 4:53 pm

Nic Lewis-
“A uniform depth error of 5 dbar globally produces a temperature bias that is greater than the observed ocean warming during the past 50 yr in the tropical and subtropical ocean and equals almost half of the observed warming in the higher latitudes when averaging between 0 and 300 m” is pretty shattering. That is a depth error of only 5 m, if I’ve got my units right.”
Holy cow!

May 10, 2013 4:57 pm

Willis Eschenbach says:
May 10, 2013 at 10:55 am
Sea water has a specific heat of about 4 megajoules per tonne. So to raise a tonne of water from -273 to 0°C require… 
Shouldn’t the specific heat capacity of ice be used from -273 to -2?

richard verney
May 10, 2013 4:58 pm

Interesting. But where are the error bars for all these measurments?
Does anyone seriously consider that there is sufficient confidence in the accuracy of this data to make analysis a worthwhile objective, especially given the shortness of the data set which is rendered worse by the material changes that have taken place in the methodolgy of taking the measurements?

markx
May 10, 2013 5:28 pm

oF interest perhaps

Important notice to Argo users (Pressure drift in APEX floats):
Users are cautioned that
(i) Most Argo Data Assembly Centers (DACs) have not yet corrected APEX profile data for the drift of pressure sensors in real-time (RT) and delayed-mode (DM) files. Steps are being taken by the DACs to apply drift corrections in both RT and DM files. It is estimated the correction will change affected APEX float pressures by an array average of about -2 dbar in 2003 data, decreasing to about 0 dbar in 2008 (due to improved sensor stability). However, a few older individual floats may have profiles with pressure offsets of over 10db. …..[……]….

http://www.argo.ucsd.edu/Acpres_drift_apex.html

markx
May 10, 2013 6:29 pm

Perhaps of interest:
SIO 210 Talley Topic 2: Properties of seawater
Lynne Talley, 2000
How is pressure measured?
….[….]….
(2) Quartz transducer now used with electronic instruments. The accuracy is 3 dbar and the precision is 0.5 dbar.
Accuracy – ability to measure compared with an absolute standard.
Precision – ability to measure consistently within a given data set (variance in the measurement itself due to instrument noise).

http://sam.ucsd.edu/sio210/lect_2/lecture_2.html
SIO 210 Talley Topic 2: Properties of seawater

Manfred
May 10, 2013 6:30 pm

Hi Willis Eschenbach,
I think this data set is rubbish and I wonder, why nobody has noticed.
I refer in particular to figure 2 and the exceptional massive, massive increase (about 1.0 E23) around 2003 in a very short period of time for the 0-2000 heat content. Such an increase of total ocean energy can only be explained by A) massive decrease in cloud cover B) gigantic underwater volcano activity.
B) appears to have not happened, and
A) did not happen as well.
http://www.drroyspencer.com/wp-content/uploads/AMSRE-CLW-est-of-CERES-SW-global-60N-60S-thru-June-17-2010.gif
so that increase is nothing but an error.

markx
May 10, 2013 6:52 pm

Very detailed discussion of pressure drift and the effects of on ocean heat content in Kobayashi And Johnson: (good graphics in this paper) but it dates back to 2007 and discusses Argo floats deployed as of 2003.
Argo Float Pressure Offset Adjustment Recommendations
Taiyo Kobayashi and Gregory C. Johnson
http://prelude.ocean.washington.edu/dmqc3/pub/argo_float_press_offset_adjustment.pdf

Float CTD sensors are subject to calibration drift after laboratory calibration. With a target of an average 4-year lifetime untended in the ocean, the potential for sensor calibration drift is significant.
Argo DMQC work first focused on conductivity sensor drift, because it was thought that calibration drift of this sensor could be the largest. With a lot careful work on this topic, systems have been built, and are being refined, for assessing and correcting those drifts statistically. These systems generally exploit the relative stability of the T-S relationship. Little has been done to detect subtle (not obvious from visual inspection by an experienced researcher) temperature sensor calibration drift. The way forward on such detection, should it be needed, is not clear. Temperature sensors are thought to be sufficiently stable to meet Argo accuracy targets even with typical calibration drift over a 4-year lifetime. Recently DMQC attention has turned to pressure.

and concludes:
….[….]…

The “actual” warming signals estimated by Levitus et al. (2005) show large variations for each ocean, more warming occurs in the Atlantic and less in the Pacific.
The warming signal observed in the Pacific is about 0.1 °C for the upper 300m water column during this 50-years, which may make it very difficult to assess climate changes in the Pacific by the current Argo Project, in which there are many profiles with pressure drift. This calculation gives an upper limit on the size of the effect because not all pressure sensors are biased. However, considering the history of Argo Project, the situation appears quite serious.
False warming and cooling signals could appear in certain areas and periods due to accretions of the following factors:
1. Among the major 3 types of Argo floats (APEX, PROVOR, and SOLO), only APEX still needs additional operation for pressure drift adjustments. However, it is likely that earlier versions of the PROVOR and SOLO floats both had problematic pressure sensors and did not effect this correction automatically.
2. Large pressure sensor drifts (more than5 dbar) are found in Paine and Ametek 1030pressure sensors, which were used for Argo floats with equipped with SeaBird CTDs before 2003. For floats deployed during and after 2003, the pressure sensors for SeaBird CTDs were changed to those made by Druck, in which drift is usually within 1 dbar (Figs. 4-5). All floats equipped with Falmouth Scientific Instruments CTDs used Druck pressure sensors (W. B.Owens, personal communication, 2007).
3. In 2003, the Argo array was still largely a collection of regional arrays, not a global one, and there were many regions devoid offloats. In addition, many regions were occupied by floats of a single country or PI (Fig. 13).
4. At the present, the operation of pressure drift adjustments depends on each country/organization/PI

markx
May 10, 2013 7:14 pm

Philip Peake says: May 10, 2013 at 8:09 am
water is odd in that it gets more dense as it cools, up to about 4C, then it starts to get less dense. So water at around 4C sinks to the bottom. Water from melting ice at the poles will continually feed the depths.
The actual temperature for maximum density depends upon the salt content, its ~4C for pure water.

Amazing stuff really…. here is more:

Freshwater density is 999 kg/m^3. Typical densities for seawater are only slightly higher: 1020 to 1050 kg/m^3, with most of this range being due to pressure. The range of densities at the sea surface is about 1020 to 1029 kg/m^3.
The freezing point of seawater is lower than that of freshwater. As sea water freezes, it forms pockets of salt. The salt (brine) leaches out of the bottom of the ice and the brine drips into the water below the ice. Thus sea ice when melted is considerably fresher than the original water which was frozen. The “brine rejection” process creates dense water below the sea ice formation area. This can be an important contributor to dense water formation in a global sense as the densest waters are formed at high latitudes, and often involve sea ice.
The faster that sea ice is frozen, the less likely that the salt can escape. Thus the saltiest sea ice is formed at the lowest temperatures. Sverdrup et al. (1942 text) tabulate the salinity of ice formed from water which starts at salinity 30. When frozen at an air temperature of -16C, the salinity of the ice is 5.6. When frozen at an air temperature of -40C, the salinity of the ice is 10.2.

http://sam.ucsd.edu/sio210/lect_2/lecture_2.html

Manfred
May 10, 2013 7:46 pm

Just to show how implausible the 1 E23 Joules increase during the approx. 2 years around 2003 is:
Global ocean surface is 361 E12 m2.
2 years have 63 E6 seconds.
That means average heat uptake of oceans must have been
1E23 Watt *s / (361 E12 m2 * 63 E6 s)
= 4.4 W /m2 in the 2 years around 2003.
Compare this with with the IPCC AR4 estimate of ocean heat uptake of 0.2 W/m2.
There is no way that oceans took up as much heat in 2 years around 2003 as in all other years between 1970-2010 combined (see figure 2).

Chipmonk
May 10, 2013 9:24 pm

Willis, thanks again for an interesting article.
Question: what if we are looking at this upside-down… from the human perspective (breathing air, walking on land, etc.) the deep ocean is a mystery. So we tend to think “top-down” as in the air is causing some effect on the oceans… What if the earth lithosphere below the ocean is really driving the recorded ocean heat content?
According to NOAA, light (electromagnetic radiation) penetration into the ocean is quite limited, red-spectrum roughly 50 meters, therefore heat (infrared) transfer from the wispy-thin atmosphere has very limited transmission depth and mixing would seem to be minimal (compared to mass of water). See:
http://oceanexplorer.noaa.gov/explorations/04deepscope/background/deeplight/media/diagram3.html
and
http://oceanservice.noaa.gov/facts/light_travel.html
So, if there is limited transmission (ignoring ocean current mixing for a minute) from air to deep ocean, could the mantle be driving changes to deep-ocean content? I admit being surprised to learn how “thin” the lithosphere is (3-5 miles), depending on age of the deep crust. Example at:
http://en.wikipedia.org/wiki/Oceanic_crust
Has anyone done actual measurements of the lithosphere heat content and change in that heat over time?
In the spirit other blogs discussing “science” for kids, I was also using the following fun thought experiment as a visual aid (works best in a dark room with a flashlight)
1) Turn on a stove-top heating coil and watch it glow (deep mantle)
2) Place wide pan (lithosphere) on heating coil
3) Pour amount of water (say 2 inches) into pan (ocean)
4) Watch it boil (while carefully observing the ring pattern of bubbles) then gently blow across the steam to see “weather”
Obviously this is not a scaled analog of the planet, and you can a wide variety of other inputs like heat lamp (Sun), but it’s fun to consider.
Russ

Manfred
May 10, 2013 9:48 pm

Willis Eschenbach says:
May 10, 2013 at 8:55 pm
However, we’re a long ways from understanding the vagaries of the ocean and the clouds. And it doesn’t take much change in the clouds to give us 2.8 W/m2 … maintaining it over three years, though, that seems unlikely.
——————————————-
If that data would be true, it would mean that half of the warming of the last 40 years happened in just 3 years. Somehow greenhouse gases “conspired” during this ARGO installation period, and opened a massive cloud window.
And I write massive.cloud window, because in AR4, a total forcing of 1.6W/m2 translated into an ocean heat uptake of 0.2 W/m2. Wouldn’t then 2.8 W/m2 heat uptake require a cloud cover forcing increase of 25.4 W/m2 ?

May 10, 2013 9:49 pm

I have found the source of Trenberth’s deep ocean heat …
[img]http://www.physioroom.com/images/products//full/37760_image.jpg[img]

May 11, 2013 8:54 am

Willis, Why doubt the rise after 2001? You should investigate the rise after 2001.
http://redneckphysics.blogspot.com/2013/05/more-rehashing-of-obvious-ohc-and.html
Thanks to MSU data, you have a satellite era global wattmeter. You can compare “global” response, volcanic, with internal redistribution, the 1998/99 regime change. I doubt it is accurate enough to get hard numbers, but it is a good gut check.