Steve Fitzpatrick writes in with a short essay:
On May 11 you reposted a blog from Dr. Roy Spencer, where he suggests that much of the increase in atmospheric CO2 could be due to warming of the oceans, and where he presents a few graphs that he claims are consistent with ocean surface temperature change contributing more than 80% of the measure increase in CO2 since 1958. Dr. Spencer’s suggestion is contradicted by many published studies of absorption of CO2 by the ocean, with some studies dating from the early 1960’s, long before “global warming” was a political issue. In this post I offer a simple model that shows why net absorption of CO2 by the ocean is most likely the main ocean effect.
If the rise in CO2 is being driven by human emissions, then the year-on-year increase in atmospheric CO2 ought to be a function of the rate of release of CO2, less any increase in the rate of removal of CO2 by increased plant growth and by absorption and chemical neutralization of CO2 by the ocean. Both ocean absorption and plant growth rates should increase with increased CO2 concentration in the atmosphere. To simplify things, I focus here only on ocean absorption.
On the other hand, surface temperature changes ought to have a relatively rapid effect, because the surface of the ocean is in contact with the atmosphere and so can quickly absorb or desorb CO2 as the water temperature changes. In fact, the ocean surface continuously absorbs CO2 where the temperature is falling, mostly at high latitudes, and emits CO2 where the water is warming, mostly at lower latitudes. Cold upwelling water from the deep ocean warms at the surface and desorbs CO2, while very cold water at high latitudes absorbs CO2 before it falls to the deep ocean. An increase in average ocean surface temperature will cause more CO2 to be emitted from surface water, but this effect is limited to a very small volume fraction of the ocean. Effects due to rapid temperature changes (annual time scale and less) are limited to a relatively thin layer, while the gradual absorption/neutralization process takes place at a rate controlled by ocean circulation and replacement of the surface water with upwelling (and “very old”) deep ocean water.
Any change in sea surface temperature should add to or subtract from the atmosphere’s CO2.
Annual change = (Annual emissions) – K1 * (CO2 – 285) + K2 * (delta SST)
Where “CO2” is the atmospheric concentration, K1 is a unitless “ocean uptake constant”, and K2 is a sea surface absorption/temperature constant, with units of PPM per decree C. Delta SST is the year-on-year change in average sea surface temperature. K1 is related to how quickly surface water is replaced by deeper water, and it should be a relatively small number, since ocean circulation and mixing are slow. K2 should be a relatively large number, since surface water temperature changes are relatively fast and we know that there is a strong short-term correlation between the rate of change of CO2 concentration and SST changes.
The model performs an iterative calculation (a step-wise approximation of integration) of the evolution of CO2 in the atmosphere. Each year a change in CO2 is calculated using the above equation, that change is added to the atmospheric CO2 concentration from the previous year, and the process is then repeated. The calculation starts with 1959, using a starting CO2 concentration of 315 (the value from Mauna Loa in 1958).
Measured CO2 values and measured year-on-year changes are from Mauna Loa. Average SST’s are from GISS. CO2 emissions, expressed as PPM potential increase in CO2 in the atmosphere, are based on worldwide carbon emissions (according to CDIAC at Oak Ridge) converted to an equivalent weight of CO2, divided by an assumed atmosphere weight of 5.3 X 10^9 million tons. This result was scaled by a constant factor of 0.7232, which is 28.96/44 = 0.6582 (to convert weight fraction CO2 to volume fraction), multiplied by 1.099 to match up with the range of CO2 emissions that Dr. Spencer used in his May 11 blog post. Note that nobody really knows the total carbon emissions, so different sources offer different estimates of total emissions. The final two years of CO2 emissions I had to estimate beacause the CDIAC data ended in 2006. I assumed an equilibrium ocean CO2 level of 285 PPM. I optimized K1 and K2 by hand so that the model had a reasonable fit with the data; the values were 0.0215 for K1 and 5.0 for K2. So the model equation is:
Annual change = (Annual emissions) – 0.0215 * (CO2 – 285) + 5.0 * (delta SST)
The graph titled “Annual Increase in CO2” compares the measured and calculated year-on-year changes along with the potential increase from fossil fuels.
The graph titled “Correlation: Model Increase vs. Mauna Loa Increase” shows that the model does a decent job of capturing the year-on-year temperature driven change in atmospheric CO2.
I suspect that if the model used monthly data and the 6-month lag between SST changes and CO2 changes that Dr. Spencer used, then the model fit would be better.
The graph titled “Measured CO2 versus Ocean Uptake Model” shows the final result of the calculation.
The evolution of CO2 in the atmosphere calculated by the model between 1958 and 2008 is reasonably close to the Mauna Loa record. The model suggests that about 2.15 PPM equivalent of emitted CO2 is currently being absorbed, or about half the total emissions.
My only objective is to show that the CO2 released by human activities, combined with slow ocean absorption/neutralization and sea surface temperature variation, is broadly consistent with the measured historical trend in atmospheric CO2, including the effect of changing average SST on short term variation in the rate of CO2 increase. Temperature changes in ocean surface waters cause shifts of a few PPM up and down in the rate of increase, but surface temperature changes do not explain 80% to 90% of the increase in atmospheric CO2 since 1958, as suggested in Dr. Spencer’s May 11 post. Because of its relatively high pH, high buffering capacity, enormous mass, and slow circulation, the ocean is, and will be for a very long time, a significant net sink for atmospheric CO2.
With a bit of luck, continuing flat-to-falling average surface temperatures and ocean heat content will discredit the model predictions before too much economic damage is done.
Paul Vaughan (16:58:48)
“Certainly this is an interesting puzzle to work on …”
Indeed so, care to join me?
Your caveats are well founded and I know that the standard explanation is that the “massive circumpolar boreal forest” forms the biological sink for carbon dioxide. However, when Ferdinand Engelbeen last year posted graphs showing this northern hemisphere CO2 summer draw-down signal, I was intrigued. There are two hats I can wear, geoscience (professional) and bioscience (amateur) and I viewed this atmospheric response from my geoscience experience and not the standard bioscience perspective. After all, thinking outside the box is what scientists are suppose to do 🙂
I decided to build on the analysis that Ferdinand presented, so I looked at CO2 data for high and low latitudes from both hemispheres, continental and oceanic locations, & high and low elevations. From inspection it is clear that there is a global pattern to these data. Consider the CO2 variation as a signal, its greatest amplitude and sharpest onset is at the highest Arctic latitudes (e.g. Barrow, Alaska), the signal propagates south through the atmosphere; Mauna Loa, Hawaii is later in time and smaller in amplitude than at Barrow, and it then reverses phase into the Southern Hemisphere, down to the South Pole which has the smallest amplitude excursion of all.
What follows is a description of my simple scoping study to establish if there is a correlatable relationship between the atmospheric CO2 summer minima at Barrow, Alaska and the extent of the Arctic Ocean open water (ice free) area.
Establishing the open water area of the Arctic Ocean is relatively straight forward using published sea ice data, if we accept the idea of a progressive northward zonal melt of sea ice as the boreal summer advances. My assumption that the southerly located frozen waters of the Sea of Okhotsk, Bering Sea, Hudson Bay & Baltic Seas all melt before the Arctic Ocean ice does, is of course a simplification prone to error, but one that could be corrected by using a detailed latitudinal melt history database.
In order to determine the strength of the CO2 summer draw-down signal for Barrow we must establish the notional carbon dioxide concentration that would exist if the summer sink was inactive. The CO2 data base for Barrow extends from 1974 to 2006, by using an appropriately designed filter it is possible to preserve only the winter data and discard all the summer values. By this means the preserved winter data can be curve fitted and the equation of this curve used to establish (despite the rising annual trend) the hypothetical inactive sink summer CO2 concentration at Barrow for all of the 33 years in the record. A cross-correlation of the strength of the CO2 draw-down signal at Barrow versus the Arctic Ocean open water extent for July thru November, fitted with a simple linear trend curve, has an R squared value of 0.664.
The amplitude of the annual Antarctic CO2 signal is small compared to the Barrow data. The South Pole CO2 record is for a high elevation continental location, when compared with the time series data for Syowa, a low elevation coastal Antarctic station, both of these CO2 data sets are coincident in amplitude and phase. The simple sinusoidal form of these data matches the phase of the annual variation in areal extent of the Southern Ocean sea ice. This phase locked match with the surrounding Southern Ocean sea ice area suggests a local geochemical cause for the austral summer CO2 draw-down, rather than a distant land-based biochemical cause for this signal.
Questions & Comments
1. Why, if the Arctic Ocean is the predominant cause of the Barrow CO2 signal, is the Southern Ocean signal so weak in Antarctica?
The Arctic Ocean is a geographical feature with a defined southern coastline, whereas the Southern Ocean is unbounded in its northern latitude. The Arctic is a “ponded ocean” and occurs at higher latitude and has lower surface water salinity than the open Southern Ocean does. The key question therefore is “Does surface water salinity affect the rate of CO2 uptake by cold ocean waters?”
2. Is it possible to determine the total mass of CO2 removed from the Arctic atmosphere by the boreal summer draw-down?
A comparison of the Barrow sea level data with the high altitude Greenland Summit data suggests that the summer CO2 draw-down affects the full vertical extent of the Arctic atmosphere.
The R squared correlation coefficient of 0.664, noted above, gives support to the idea that the ice free Arctic Ocean sea water acts as a CO2 sink during the northern hemisphere summer, in addition to the currently recognised biological sinks.
Philip Mulholland (15:16:17) :
We seemed to have been thinking along similar lines:
A co2 plot from many locations (with ch4 at barrow thrown in)
http://img527.imageshack.us/img527/6153/co2manysitesch4.jpg
A plot of days from 1st jan for Barrow CO2 to reach minimum
http://img30.imageshack.us/img30/4917/barrowlajollatimeformin.jpg
Note that la jolla pier clifornia has almost identical date.
Note also that the Barrow minima have no greatly changed over the record despite temperature and seaice changes. The La Jolla data is daily the barrow is houly. Plot of last 3 barrow minima:
http://img32.imageshack.us/img32/6678/barrowhoulydata.jpg
The minima at barrow occur Not at minimum Ice Not at Autumn onset Not at algal Bloom time
If it were sea ice then why is it so strong over 2/3rds the globe christmas island shows a significant dip.
South Pole station has a ripple 6 months out of phase with NH.
The signal is as strong in central kazakhstan (land locked) as at Barrow (coast)
I cannot explain it!
Philip Mulholland (15:16:17) “Does surface water salinity affect the rate of CO2 uptake by cold ocean waters?”
Sure – since it has a very serious effect on freezing temperature.
Otherwise, the references the chemically-minded WUWT participants have posted haven’t (so far at least) emphasized a (strong) role for salinity in the CO2 equilibrium chains. (I’m eager to learn if anyone has any details &/or links to share to shed deeper illumination on the role of salinity.)
–
Philip Mulholland (15:16:17) “Is it possible to determine the total mass of CO2 removed from the Arctic atmosphere by the boreal summer draw-down?”
I know people who work on this, but I haven’t been following their work; however, recent WUWT threads have made me curious enough to make some enquiries next opportunity I have.
One thing I can share is that ~15 years ago when I was involved with a research group that was modeling biogeochemical (earth-water-atmosphere) cycles, the modelers were happy if they got fluxes to within a factor of 2.
There’s a lot of spatiotemporal variability in the field (& sampling resources are not infinite), so it is unrealistic to expect precise estimates – in many biogeophysical modeling contexts.
–
I am also curious about the roles of fresh water (on land) and rocks …and wind …i.e. beyond vegetation, oceans, & temperature …but for the near term (i.e. until I have time to consult a *lot* of literature and run a lot of analyses) I’ll be content to lump everything together, with the possible exception of wind.
HEAD’S UP:
anna v’s instincts about cleansed data are justified – unbelievably so.
Anyone curious to know more?
Paul Vaughan (00:52:35) :
Anyone curious to know more?
Yes
But are you saying that the raw hourly data is fake
I assume this is made by gas analysers.
There are various other methods which I have posted somewhere (climateaudit?) 2 flasks/single flask etc.
It is inconceivable that someone is going to go through 10MBytes of hourly data to falsify the output.
If you look at the output here:
http://img32.imageshack.us/img32/6678/barrowhoulydata.jpg
I have not removed their flagged errors (the vertical lines are all for total failures (-99 or -999) there are other flags used to remove spurious “errors” – these have not been acted on.
Are you suggesting someone sat down and did this?
On this plot I have added filtering to remove the randomness so that the plot is usable. But there is still large variability.
http://img527.imageshack.us/img527/6153/co2manysitesch4.jpg
I again would be suprised if this error ridden data is falsified.
So pleae tell!
Here is a suggested exercise – using Alert, Nunavut (formerly part of NWT = Northwest Territories – but now separate), Canada:
Compare the following 2 (monthly-resolution) time series:
1) http://cdiac.ornl.gov/ftp/trends/co2/altsio.co2
2) ftp://ftp.cmdl.noaa.gov/ccg/co2/flask/month/alt_01D0_mm.co2
Break the analysis down as follows:
a) annual timescale – i.e. applying 12 month bandwidth moving-average.
b) seasonal timescale – i.e. using differencing to isolate seasonal variations from the (secular) trend.
Cautionary Notes:
i) Be (very) careful with missing values (which are summarized differently in the 2 series).
ii) Be sure to note any systematic patterns in the errors.
My reaction to what I found went something like this:
“This is outrageous – unfathomably & unbelievably so!”
The adjustments seriously & systematically distort seasonal variations.
…So why are the files listed as “Data“?
http://cdiac.ornl.gov/trends/co2/sio-keel.html
The ‘adjustment’ procedures are mentioned in the files (at the bottom), but this does not change the fact that:
Labeling model-output as “Data” is grossly misleading.
For those who skipped the analysis, r^2 (seasonal) is less than 0.33 and model assumptions are severely violated.
Why replace real data with loosely-related severely-seasonally-biased modeled “data”, particularly if defending the model assumptions is like asserting that a colorful polka-dot pattern is indistinguishable from solid-grey?
If one breaks the analysis down by month, 8 of the 12 r^2s go below 0.1 and 4 even go below 0.01.
…and earlier in the discussion attention was already drawn to “daily” data which are not actually daily, so …
Some websites begin appearing ‘fishy’, ‘sketchy’, & untrustworthy. “Data” labeled as data should not be assumed to be unbiased representatives of data. Rigorous interrogation is clearly warranted.
Paul Vaughan (20:48:22) :
My reaction to what I found went something like this:
“This is outrageous – unfathomably & unbelievably so!”
The adjustments seriously & systematically distort seasonal variations.
……
Why replace real data with loosely-related severely-seasonally-biased modeled “data”, particularly if defending the model assumptions is like asserting that a colorful polka-dot pattern is indistinguishable from solid-grey?
…
If one breaks the analysis down by month, 8 of the 12 r^2s go below 0.1 and 4 even go below 0.01.
thanks for this info. So Pamela Pamela Gray (09:27:30) : was right about modeling being inputed to the data.
My suspicion flags were raised when on a search of bibliography I found all the publications were Keeling + somebody else, a graduate student most probably. One could call it the Keeling effect. Lets hope the Japanese do not catch it.
Paul Vaughan (20:48:22) :
Here is a suggested exercise – using Alert, Nunavut (formerly part of NWT = Northwest Territories – but now separate), Canada:
Break the analysis down as follows:
a) annual timescale – i.e. applying 12 month bandwidth moving-average.
b) seasonal timescale – i.e. using differencing to isolate seasonal variations from the (secular) trend.
—-
OK, so try this third breakdown of the data:
Instead of using a running 12-month “average” to smooth data – Use a “5 month seasonal” average: This month is averaged against LAST YEAR’s month, last year (month-2 , month-1, month, month+1, month+2)
Too much impact of just last year’s influence?
Alternative 4: Add one more year to the check:
Smooth this month with (year-2: month-1, month, month+1) and (year-1: month-1, month+1)
Paul Vaughan (01:42:05) :
Philip Mulholland (15:16:17) “Does surface water salinity affect the rate of CO2 uptake by cold ocean waters?”
—
OK, so look at it this way: We know that real-world (air!!!!) temperatures have “only” gone up by 1/4 of one gree. 1/2 of one degree at maximum in 1998.
So, given the aera of the ocean, assume the ocean went up an equal amount. (Or, get actual ocean near-surface (top 10 meters) temperature data for every ten years since 1945 from anti-submarine records of the Atlantic and Arctic. )
Can the real-world changes in real-world ocean surface temperatures explain the change in CO2 levels measured? If not, what is the difference? Is the measured difference inn CO2 concentration GREATER or LESS than what is predicted from the change in ocean temperatures, and – if it is – can that difference be CALCULATED from what is known about man’s ACTUAL carbon mining and drilling?
(No “gueses” or “assumptions” about jungle-clearing effects – cutting trees to clear land only exposes new land for growing NEW trees and vegetation and grass on the newly cleared land. )
Re: RACookPE1978 (10:10:33)
It depends on what one is investigating. For example, to look at annual timescale, use 12mo time-integration. (In detailed analyses, all timescales are investigated.)
Re: RACookPE1978 (10:19:02)
Clarification: Philip was asking about the effect of salinity.
Re: anna v (09:31:57)
The real data is available online too …but one has to be careful because some websites have posted modeled “data”. Watch out for seasonal structure that looks “too perfect”. In the example I gave, the NOAA data ‘seems’ real, but the CDIAC “data” is a ridiculous representation of seasonal variation.
correction: data ‘are’ & ‘seem’ (not ‘is’ & ‘seems’) [Data – plural – (vs. 1 datum)]
Steve Fitzpatrick,
” However, since the ‘age’ of the deep ocean water that is currently upwelling is likely in the range of ~1000 years”
You might want to take a look at this paper:
http://www.nature.com/nature/journal/v459/n7244/full/nature07979.html
The idea that the ocean has currents that hold water at the bottom for 1000 ears may not be correct.
Thanks for your susceptibility, It seems that we are building global warming still