Guest Post by Willis Eschenbach
Like Jason, I proceed into the unknown with my look at the Argo data, and will post random notes as I voyage.
Come, my friends, 'Tis not too late to seek a newer world. Push off, and sitting well in order smite The sounding furrows; for my purpose holds To sail beyond the sunset, and the baths Of all the western stars, until I die.
I have no great insights at this point, just some interesting results. Thanks to a commenter who pointed me to where to get the Argo data in one block. It’s at the Asia-Pacific Data-Research Center.
I downloaded it, and I’ve looked first at the file containing the surface data. It’s where I swim, so it’s the most interesting data to me. Figure 1 shows all Argo measurements of the ocean surface temperature taken to date.
Figure 1. All Argo ocean surface temperature data. There have been 696,872 Argo measurements to date of the ocean surface temperature.
So far, so good. The results look real, which is always good to see, it means I’ve graphed them up properly. You can see the warm ocean along the coast of Europe, for example. But there is one curiosity about the Argo data.
Here’s the oddity. I took the data arranged by latitude as shown in Figure 2. I averaged it by 1° latitude bands, and then took an area adjusted average to give a global mean. The mean is 19.7°C ± 0.02 (95% CI).
Figure 2. All Argo ocean temperatures, sorted by latitude. NOTE: several people commented correctly below that I had not included the variation in ocean area by latitude band in the calculations. They are correct, I was wrong, and the actual corrected 60N-60S average is slightly higher, at 19.9°C.
Note that there is an obvious upper limit to the ocean temperatures, the “flat-top” on the graph at just above 30°C. No matter how much incoming solar there is, the ocean doesn’t get any warmer than that. This provides a “cap” on how hot the ocean can get. Above that temperature, any extra incoming energy is converted to latent and sensible heat, rather than warming the surface.
But I digress, that part’s just interesting. It’s not the curiosity.
The curiosity is the other ocean data sets give the following values for the average ocean surface temperature 2000-2011:
Hadley Center HadISST1 60N – 60S: 20.5°C ± 0.02°C (95%CI)
Reynolds Optimally Interpolated SST 60N – 60S: 20.4°C ± 0.02°C (95%CI)
NCDC Extended SST 60N – 60S: 20.3°C ± 0.02°C (95%CI)
The curiosity is that the Argo average ocean surface temperature data is significantly cooler than the other datasets, half to three-quarters of a degree …
Always more to learn. I do love real data. Look how much colder and more uniform the Southern Ocean is than the northern oceans, for example. Fascinating stuff.
Best to everyone,
[UPDATE]
The data I used is available at the website listed above, identified as “Near-real time Argo profile data interpolated on standard levels”. It’s the largest file on this page, 895 Mb, titled “Argo_TS.tar”.
The info sheet detailing the arrangement of the data is here.
It’s a tarball containing all of the depth files, one for each layer. The one I used was the zero depth file, “Argo_TS_0000.dat”. I downloaded them all, because I wanted the full set. If you only want surface temps you can download just that one file.
To read it in once it was downloaded (in the “R” computer language), I used:
depthcolumns=c("Longitude", "Latitude", "Level", "Depth", "Julian", "Temperature", "Salinity", "Potential Temperature", "Potential Density", "Dynamic Depth Anomaly", "Spiciness", "Extrapolation", "Error Temperature", "Error Salinity", "Error Potential Temperature", "Error Potential Density", "Error Dynamic Depth Anomaly", "Error Spiciness", "Ocean Code", "Region Code", "Argo Float ID", "Cycle Number", "Dynamic Depth", "Dynamic Depth-2")
depthwidths=c(9, 9, 3, 7, 10, 9, 9, 9, 9, 9, 9, 2, 11, 11, 11, 11, 11, 11, 2, 3, 8, 4, 9, 9)
depthinfo0=read.fwf("/Users/willis/Argo_TS/Argo_TS_0000.dat",depthwidths, col.names=depthcolumns)
You’ll need to change the filepath in the final line to wherever you have put the “Argo_TS_0000.dat” file.
w.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
I have often thought of measuring daily the water temperature from my taps. I know it varies with the season and a bit less after a hot spell of eather or a cold spell. But if I did go to the trouble of making daily measurements, what use would they be if they could only be compared with the air temperatures, which vary during the day by up to 20º C?
Fifth, you then at the end of your response set out a summary of the ARGO data reviewed by you. This summary suggests that nearly 10,000 points/sets of data show a temperature exceeding 30degC. That alone, would indicate that even based upon the data that you yourself have reviewed your assertion of a 30degC cap for surface sea temperature cannot be correct.
Those greater than 30C temperatures are clustered around 30 to 35 degrees north, which is where the Earth is mostly land and there are enclosed and semi-enclosed seas, the Med, Yellow Sea.
Without a geographic breakdown, I’ll suggest most of those +3OC temps are close to land.
30C does appear to be an upper limit for SSTs in the open ocean.
The cause is likely to be the hydrological cycle limiting the rise in air temps over the ocean.
And it doesn’t matter how hot your swimming pool gets, its not going to affect the hydrological cycle.
“John, either you are missing something or I am. The Argo temps Willis has used are not summer readings – they are ALL the data, summer and winter, and, I suppose, everything in between.”
Yes it is all the data that is the problem. You’re effectively overlaying several bell curves sightly offset from each other by the changing angle the sun indents on the Earth.
Plot the temperatures for a single month and I’ll wager you the “flat-top” all but disappears. There is no mystery here I’m afraid it’s just an artifact from including the whole years data in 1 graph.
Philip Bradley said @ur momisugly February 10, 2012 at 2:40 pm
IIRC it was Roger Pielke Sr who pointed us to a slide show where one of the slides said that humans now sequester 40% of rainfall. A single swimming pool ain’t gonna make a lot of difference, but 40% of rainfall will definitely influence the hydrological cycle.
John –
You are seeing bell curves where it is not a bell curve application. These are not “normal” distribution, but are based on latitude, and it is instead a case of a sine curve (because of the roundness of the Earth) on either side of the Equator. Yes, it LOOKS like a bell curve, but you are applying the wrong idea to it. Don’t be deceived by the flat top/cap. There are several curves that turned that way could be misconstrued as “bell curves.” But bell curves don’t apply to spherical applications.
Steve Garcia
Whoa Willis.
I decided out of curiosity to look at the data and as is my way I eyeball before trusting.
Looked at 5 metre daa. Choose a float and extract the data. Looks sane for one float, running about 40N in the Pacific.
On checking the date, ouch. According to the Read_me “5 – Julian time (days) relative to 2000-01-01 00:00 UTC”, the float start date is Julian -588.69 and increments from there. Cycles increment too. Web site says “Deployments began in 2000 ”
Does not compute. (why I actually look)
Temperature… careful. Look at time of day for data. This poses a problem, is this time of satellite pass for collection or time of temperature reading, it doesn’t say. Try plotting the time of day, very odd result and a warning about diurnal, nyquist etc., remember the latitude moves.
Doubt this will work on WUWT
http://daedalearth.files.wordpress.com/2012/02/argo-29000-a.png
Note the large annual cycle. Is it safe to average all floats without taking time of day and year into account?
Unless I have done something wrong.
[Reply: Fixed link. Just copy & paste the URL, nothing more. ~dbs, mod.]
Richard Verney –
Some of you aren’t reading carefully. Willis stated THIS, not 30.00000C:
“Note that there is an obvious upper limit to the ocean temperatures, the “flat-top” on the graph at just above 30°C.”
Then, when someone picked nits about 30C, Willis came back with:
Most of your linked caps are at about 31C, very few at 32C, with onesy-twosies at 34C. Willis DISPLAYS outliers, so bringing in those doesn’t change his assertion that there is a flat top at “just above 30C,” which, when I look, looks more like “just about” 31C. and I don’t mean 31.000000000C.
Where I come from stating something in whole numbers means +/- 0.49999, and when Willis says he just did this out of curiosity, I don’t expect him to then parse his words as carefully as if it was an academic peer-reviewed paper. Can you guys give him a break and quit playing auditor?
YES, there seems to be a flat top. YES, Willis speculated that some mechanism was possibly doing that. It does NOT seem to be just an artifact of the methodology, because there ISN’T much of a mechanism to begin with, just a loosey-goosey “Let’s see what this does” look. That is the data off the buoys, whatever the TOB and whatever the missed zones. The data is what the data is. And it doesn’t depend on what your definition of “is” is.
Steve Garcia
That down-welling, back-radiated LW infrared really gets busy around the Equator. Look, it has increased the average temperature of the equatorial ocean by about 12C relative to the “average” ocean. I’m amazed!
@Philip Bradley says: February 10, 2012 at 2:40 pm
Fifth, you then at the end of your response set out a summary of the ARGO data reviewed by you. This summary suggests that nearly 10,000 points/sets of data show a temperature exceeding 30degC. That alone, would indicate that even based upon the data that you yourself have reviewed your assertion of a 30degC cap for surface sea temperature cannot be correct.
Those greater than 30C temperatures are clustered around 30 to 35 degrees north, which is where the Earth is mostly land and there are enclosed and semi-enclosed seas, the Med, Yellow Sea.
Without a geographic breakdown, I’ll suggest most of those +3OC temps are close to land.
30C does appear to be an upper limit for SSTs in the open ocean.
The cause is likely to be the hydrological cycle limiting the rise in air temps over the ocean.
And it doesn’t matter how hot your swimming pool gets, its not going to affect the hydrological cycle
///////////////////////////////////////////////////
Philip
As you will have noted from my post. I have reviewed thousands of ship’s logs containing hundreds of thousands of entries and I can assure you that 30degC is not the upper limit for SSTs whether in enclosed, semi-enclosed or open oceans. Incidentally, the Med only reaches above 30 deg C close to Alexandria and very near to the Northern coast off Egypt.
Warms seas can be commonly encountered in and around Indonesia, Thailand, off the coast of China and Japan, in the Indian ocean both up the west side of India and the east side of Africa, in the approaches to and through the Red Sea (see my earlier post which referred to a wikipedia entry recording the survey done of Sudan revealling 32 degC temps) the Atlantic ocean around north east Africa, in and around the Gulf of Mexico.
Just to give you a slight flavour, see the following links:
Off the coast of UEA, this months water temp 32degc, long term max temp 35degC . See http://www.sea-temperature.com/country_water/united%20arab%20emirates/127
Off the coast of Ghana, this months maximum temp 34 degC, long term max temp 33 degC (the maximum cited is not the absolute maximum but more the average maximum there will always be examples when a higher figure is unusally observed) . See http://www.sea-temperature.com/country_water/ghana/77
Indian Ocean say off the coast of India, this month’s maximum 31 deg C, long term max temp 31 degC. See http://www.sea-temperature.com/country_water/india/110 In practice, this is far to general, warmer temperatures are not infrequently recorded off the west coast of India and say off the coast of Madagascar, see http://www.sea-temperature.com/country_water/madagascar/64 .
Off the coast of Thailand, this months max temp 32degC, long term max temp 32 degC. See http://www.sea-temperature.com/country_water/thailand/119
In an around Indonesia, eg this month’s max temp 31degC, max long tem temp 31 degC. See http://www.sea-temperature.com/country_water/indonesia/115 and East Timor this months max temp 29 degC, long term max 31degC. See http://www.sea-temperature.com/country_water/east%20timor/114
In and around Mexico, for example http://www.sea-temperature.com/water/tapachula/536 and Guatemala, see: http://www.sea-temperature.com/country_water/guatemala/49 noting max temps of 31 degC.
I make no comment upon the distribution of the ARGO buoys, My comments are based upon a review of thousands of logs compiled by ships plying trade through the shipping routes through the major oceans of the world. You will note from the above links (which is by no means an exhaustive list) that the distribution of temperatures above 30degC is not as narrow as you suggest nor is it limited to enclosed and semi-enclosed oceans.
If you read the third numbered paragraph of my post of 8:12 am you will note that I accept that the hydrological cycle has a role to play but it is not the only process at work.
The hydrological cycle does not cap SST in open ocean at 30 degC.
PS. I am not comparing my swimming pool to the open ocean. My observation was meant merely as an indicator upon which many people will have some experience so that they can get a grasp as to general merits of the assertion made by Willis that ‘no matter how much incoming solar there is the process of evaporation caps temperature at 30 degC’ (my paraphrasing). Obviously, not many people will have reviewed thousands of ships logs and will therefore have little feeling for what ships report as prevailing ocean temps.
ALSO: Looking at Willis’ breakdown, if he’d had said 31C instead of 30C, which is what he should probably have done, there are, as he points out, quite a bit fewer than 1,000 data points out of nearly 700,000 above 31C, so YES, there is an apparent ceiling. With about 1 in 700 points being outliers.
SG
feet2thefire says:
February 10, 2012 at 1:00 pm “…At the same time, yes, it does appear that some mechanism must be occurring to limit the “All-Argo” temps at that level. Right now anything anyone says is just a guess, maybe an informed guess, but still just a guess….”
That strip of warm ocean, let’s say from 20˚S to 20˚N, receives at noon on a clear day, an instantaneous solar heating input equivalent to about 88˚C. That these measured temperatures max out at around 30˚C is related to several things: 1) that maximum solar input at zenith doesn’t pertain all day; 2) half the time the strip of ocean is in the dark, receiving no solar input; and 3) a great deal of the heating input when it’s being received is being dissipated by evaporation and convection, among probably numerous other things.
And that doesn’t even begin to account for a massive influx of down-welling LWR that goes on 24/7. It’s a wonder the sea doesn’t boil!
Richard Verney,
Your examples, coast of Thailand (Gulf of Thailand), coast of UAE (Persian Gulf), coast of Sudan (Red sea and Gulf of Aden), etc are all examples of semi-enclosed seas adjacent to large land masses.
My point, which I could have explained better was that these areas are subject to air masses, typically dry, originating over land, and therefore would not experience the same hydrological processes as oceans far from substantial land masses, which constitute the bulk of the oceans.
@Steve Garcia
I used bell curve just to describe the shape not that it was a statistical distribution but yes sine wave would be the correct.description
However, the flat-top is a result of plotting the whole year data on 1 graph. Plot the individual months in the year overlay them and you’ll see I’m right and how the flat-top arises.
I haven’t scanned hundreds of thousands of data points. However, as I recall, the Red Sea off the coast of Sudan has the warmest peak temperatures, at around 34C. Further north near the Suez, sea air temperatures peak at around 40C. In the Mediterranean, evaporation lowers the sea surface by enough to cause a surface flow from west to east throughout the Mediterranean. Sea surface summer month average temperatures reach 28C near parts of Turkey and Israel.
Is it not apparent that as tropical energy surface insolation increases, an ever higher percentage of the increased energy goes into latent heat of evaporation? Is it not logical that as the rate of evaporation increases so will cloud formation and the speed of conduction? It it not apparent that clouds, especially the under clouds, plus increased evaporation will reduce surface insolation and T? Is it not clear that all across the tropics the high points will be in that particular geograpical locations summer? If the tropical high points at all the ocean locations quickly fall off upon reaching a certain level, and that level is more smooth and lower than land data, then [there] are limiting factors in the ocean surface T not realised in land data? So, is it not fair to say that above 30C, T in the tropical oceans rapidly increase the rate of evaporation , convection and cloud formation, so that the number of recorded T above 30 C rapidly declines to a very low percentage.
>
The curiosity is that the Argo average ocean surface temperature data is significantly cooler than the other datasets, half to three-quarters of a degree …
>
That is one part of the adjustments done by hadSST3, a recent warming is added due to the increased proportion of argo sensed data in the mix. It should be noted that ALL Hadley Centre data is highly “corrected” , it is not to be compared directly as you are doing here. Apples etc. …
They’ve added about 0.07K since 2005. considerably less than you find here but be careful of making simplistic comparisons of incomparable quantities.
>
No matter how much incoming solar there is, the ocean doesn’t get any warmer than that. This provides a “cap” on how hot the ocean can get. Above that temperature, any extra incoming energy is converted to latent and sensible heat, rather than warming the surface.
>
Confirmation of the negative feedback I just suggested in your other thread on argo. 😉 This may explains the asymmetry you noted there.
Alistair Ahs says:
February 10, 2012 at 5:21 am
I’m sorry, but that’s not correct.
I have averaged first by latitudinal bands. If the area with missing data is the same temperature as the adjacent deeper water, including the missing data will not change the average temperature of that band at all.
So indeed, the missing data must be warmer than the adjacent data to change the average. Whether it is warmer than the global average makes no difference.
Thanks,
w.
michael hart says:
February 10, 2012 at 7:30 am
Very nice, bonus points.
w.
steven mosher says:
February 10, 2012 at 10:12 am
Thanks for the question, Steven. I know the ocean is not spatially coherent for two reasons:
1. I’ve spent a lifetime observing the ocean from on the surface, under the surface, and from the air. I’ve lived my live out there, I’ve spent years sailing and motoring and diving and flying over, on, and under the ocean. I’ve driven for days watching a thermometer measuring the sea temperatures, looking for the albacore. I’ve seen the numbers of tide rips and temperature changes on a host of oceans of the world. It is not a coherent mass of water. It is a mix of innumerable water bodies, each with its own characteristics, and with surprisingly sharp boundaries (thermoclines, descending vs ascending water at night, rips and eddy lines between water bodies, etc). and not very spatially coherent either vertically or horizontally.
2. I’ve actually run the Argo data with and without the 3°C-wide color bands. You should try it.
w.
Willis
You are right, any swimmer in northern latitude waters knows there are warm patches of water and cold patches of water. The same goes for the atmosphere but to experience the warm and cold patches you need to be on a bike, not walking
Tonyb
@William Howard M. Connolley
I believe you’re correct. When I use the APDRC’s interpolated data and weigh the latitudinal mean surface temps by the fraction of grid boxes reporting a value multiplied by the circumference of the latitudinal band, I get a weighted average mean temp of 20.37C between 60N to 60S.
Maybe there’s a spatial bias in the data Willis is using?
tchannon says:
February 10, 2012 at 6:53 pm
Thanks, tchannon. You looked, but you haven’t looked hard enough. Yes, there were the first trials of the system beginning in 1996. But the actual system deployment started in about 2000. Here’s the number of profiles per year:
Looking at that, when would you say the deployments started?
w.
John says:
February 10, 2012 at 3:13 pm
Sorry, John, but you definitely lose the wager. See my newer post, “Argo Notes Part 2“, for another view of what is actually happening. There is a clear “flat-top” on the warm end of the annual swings.
w.
feet2thefire says:
February 10, 2012 at 7:26 pm
Thank you most kindly, sir, clearly you understand what I said. I was looking at the chart with tick marks at thirty and thirty-five degrees. I noted that it didn’t get far above 30 degrees.
Now, there is one obvious caveat here. I’m clearly talking about the open ocean, because that’s where the Argo floats are located. So yes, if you take a boat on the Red Sea or the Dead Sea, you can find warmer water.
However, my point remains. In the open ocean there are hardly any examples of water temperatures much above 30°C. Here’s a histogram of the temperatures:
So John and Richard and others, I’m sorry, but there most definitely is a thermostatic mechanism at work limiting open ocean temperatures. This has been known for some years.
w.
Grin, trust Murphy to show me an abnormal set.
I still warn about aliasing etc. No doubt there is a lot which needs considering to do with coherency. Your flat top, various things come to mind, no idea why without legwork.
Oh and thanks for pointing out the data source, I’d looked but never found it.
Anyone interested this is the 1979 paper I referenced earlier, sorry, no direct link.
Evaporation conduction of latent heat may vary far more then realized and set a limit on further tropical temperature increases. (Newell & Dopplick’s (1979) calculations that tropical temperatures cannot rise any further.)