Rasmus, the Chevalier and Bucket Adjustments

We’ve heard a lot recently from RC about whether solar can account for temperature variations in the 20th century. While my primary interest in these ruminations is the possibility of pinning down a set of standards by which they agree that a study is “bad” – since it’s hard for me to determine any sins in Courtillot et al 2007 that are not committed more egregiously in MBH, there are some intriguing issues connecting SST bucket adjustments to attempts to assess solar correlation. Needless to say, Chevalier Pierrehumbert does not himself worry about bucket adjustments – presumably the Chevalier regards that as a chore for stablehands. But there are some interesting issues and connections. Today I want to draw your attention to Paltridge and Woodruff (1981) online here , which provides an interesting counterpoint to bucket adjusted SST.

In several past posts here here here , I’ve observed that there is a very substantial and speculative adjustment of SSTs for changing bucket measurements, amounting a sharp adjustment in December 1941 on the hypothesis that buckets were converted to engine-inlet overnight all over the world. While this hypothesis is deeply embedded in SST values, it is also untrue as shown here where the majority of measurements in 1970 were said to still be done by buckets. It seems very likely to me that the 0.4 deg C adjustment implemented in Dec 1941 is almost certainly implemented far too early and that the adjustment, if it is indeed a valid one, should come into play only after 1970. The implication is that SSTs for the period 1941-1970 may have been adjusted too much. I don’t know this: but I certainly think that any bucket adjustments should be made statistically and not a priori.

Anyway, it caught my eye when Rasmus objected to Scafetta and West on the basis that their hypothesized solar correlation did not account for the 1940-1970 dip in temperatures, which he expressed as follows:

S&W maintain that the climate response is greater for longer time scales (which is reasonable) as illustrated in their figure 4 (reproduced below), and assisted by the simple model illustrated in this figure, they argue that the present warming is a delayed response to past solar changes (presumably before the 1950s). But it is unclear why the temperature then flattened out and even dropped a little between 1940-1970 at the time when it really should have increased fastest. One could argue that something else also happened then, but for an unknown reason, this forcing then seemed to have a shorter relaxation time. Why such an interference would give a quicker response than a solar signal is unexplained (the response to volcanoes is fairly prompt, however).

If some portion of the flattening out and dropping was due to overzealous bucket adjustments, then this particular issue would obviously need to be re-visited.

The Chevalier made an opposite complaint against Courtillot et al. Courtillot had used a Jones temperature version (see my earlier post on the French farce), which was a 20N version with a larger downdip in the 1940-1970 period than the corresponding global series.

Courtillot’s “Tglobe” curve did not look much like the curve published by Jones. Jones’ curve, plotted from his actual data files, is shown in Bard and Delaygue’s corrected version of the figure; they also show the NASA reconstruction for comparison. These two curves are in agreement, but neither shows the sharp rise/dip pattern between 1940 and 1970 which is seen in Courtillot’s figure.

It’s odd that the 1940-1970 period should occur in both criticisms: in the one case, Rasmus thought that the S&W mechanism failed to explain the downturn; in the other case, the Chevalier thought Courtillot had intentionally substituted a temperature version showing a greater downturn than applied in the GLB version. One was too hot and one was too cold: I guess it’s the realclimate version of Goldilocks and the Three Bears.

Paltridge and Woodruff 1981
Garth Paltridge is an eminent climate scientist, perhaps best known for highly provocative articles seeking to explain the distribution of poleward heat transfer – articles that are credited in most recent discussions involving Maximum Entropy Production. But today I’m going to discuss a more humdrum topic – his estimates of SST from 1880 to the 1970s – estimates that preceded and are independent of the Folland bucket adjustments that have affected all SST estimates from 1984 on – to such a degree that the original data is now pretty much lost from view.

On the left is a Figure from Paltridge and Woodruff 1981 comparing their SST estimates to a then contemporary (pre -Phil Jones) estimate of land temperatures. On the right is a figure using CRUTEM3 (for land) and HadSST2 (for SST) covering comparable periods, drawn in a similar style: CRUTEM dashed, HadSST in points (designated solid in the legend as I don’t know how to do a legend differentiating lines and points). I’ve drawn the post-WW2 values in red for emphasis.

 paltri27.gif  paltri29.gif

Left: Figure from Paltridge and Woodruff 1981 comparing their SST estimates to a then contemporary (pre -Phil Jones) estimate of land temperatures. Right: using CRUTEM3 (for land) and HadSST2 (for SST) covering comparable periods, drawn in a similar style. 0.4 deg has been added to the later series to adjust their centering to more or less match the centering in the earlier diagram. A WW2 value is shown without filling since the figure seems to be missing in the Paltridge version. The SST are in points though they are shown as a solid line in legend.

There are obviously some pretty startling differences between the two figures. In the Paltridge and Woodruff version, SSTs continue to rise through the 20th century until around 1960 when they decline slightly. On the other hand, in the bucket-adjusted Jones-Folland version, SSTs decline after WW2. Vis a vis the two incidents described above, the 1940-1970 history entirely depends on bucket adjustments. Although Paltridge and Woodruff wrote several years before Folland 1984, they argued that bucket adjustments were not required for their data set as follows:

The main problem with SST data is that there have been changes in observing technique over time. The change which has been most investigated is that from bucket to condenser intake observations – the latter supposedly yielding temperatures higher on average by about 0.3 K (James and Fox 1972). Fortunately the basic source for the present work (the Historical Sea Surface Temperature data) is based only on bucket temperatures so that until 1960 at least the problem does not arise. The years after 1960 are an unknown mix of bucket and condenser-intake temperatures (at least for the Indian Ocean, Atlantic Ocean and Mediterranean Sea FCDS information) and are a problem. Note, however that a change to condenser-intake temperatures would presumably lead to the recording of higher temperatures. There is no evidence of any sudden apparent rise in temperature (or even a change of slope) which occurs at or after 1960 and which is common to all the curves discussed here. Indeed the NH summer temperatures fall dramatically after that time.

However, Folland and Jones thought otherwise and adjusted anyway. In 1993, James Hansen, of all people, sharply criticized the adjustments as highly speculative (discussed here).

Developing a thorough analysis of the SST record is obviously a herculean task, but needs to be done. The SST adjustments have been calculated in a very ad hoc method and it would be nice to see at least a modicum of statistical rigor.

The Chevalier, following Bard and Delaygue, excoriated Courtillot et al for relying on a tuned version of the Mangini speleothem as evidence for solar relationships. But they are all silent on the tuning of SSTs to air temperatures, a process somewhat confusingly described in Folland al 1995 as follows:

The first quantitative corrections to SST data were tentatively indicated by Folland and Kates (1984), and were closely followed by those of Folland et al. (1984), who applied a constant positive correction of 0.3 deg. C to all data before April 1940, one of 0.25 deg. C to data between April 1940 and December 1941, and no correction thereafter. This largely removed the jump from time series of global and hemispheric SST relative to corresponding series of night marine air temperatures (NMAT) measured by ships and independently corrected by Folland et al. (1984). Jones et al. (1986) used screen air temperatures in coastal locations to correct SST data up to 1945. This assumed that the land air temperatures were free from time-varying biases such as urbanization

We use the corrected NMAT data set of Bottomley et al. (1990) to help tune the SST bias corrections developed here for the late nineteenth century and to independently assess the corrections for later dates. Bottomley et al. (1990) give full details of the corrections to NMAT. In the Atlantic up to 1885 and the Mediterranean and North Indian Oceans up to 1893, their corrections to NMAT depend on comparisons with SSTs corrected according to the Bottomley et al. method. We have therefore avoided the use of NMAT corrected this way when tuning late nineteenth century SST corrections. The pre-1893 NMAT data have not yet been re-corrected to reflect the small differences between corrected SST data derived here and those in Bottomley et al. (1990). These differences are generally less than 0.2 deg. C locally, less than 0.1 deg. C over the above ocean basins, and up to only a few hundredths of a degree globally.

So the SSTs are corrected based on NMATs which, according to the above, are themselves “re-corrected” to SSTs. Does the Chevalier himself have any idea how these corrections were done? I doubt it. That’s a job for stable hands, I guess. The net result, however, is that SST and land temperatures, which supposedly reinforce one another may not be “independent”, as a result of the above tuning. From a statistical point of view, it would be nice to have access to the unadjusted data as well as the adjusted data so that the effect of the bucket adjustments could be kept squarely in the foreground – and, if necessary, estimated concurrently with other effects.

Given the range of SST outcomes that derive from using the Folland-Parker-Jones approach on the one hand or the Paltridge-Woodruff approach on the other hand, I must say that I’m having trouble understanding how one can use this record to either support or preclude solar influence.

Reference: Paltridge, G. and S. Woodruff, 1981. Changes in Global Surface Temperature from 1880 top 1977 derived from Historical Records of Sea Surface Temperature url


86 Comments

  1. Bernie
    Posted Dec 23, 2007 at 8:58 PM | Permalink

    Steve:
    Merry Christmas.
    Is the legend correct for the CRUTEM & HadSST? I see no solid line.

    It is truly unnerving how all these adjustments lack transparency.

    Steve:
    The points are the SST, dashed – land. I didn’t know how to denote this in the legend software, and have added clarifying comments.

  2. steven mosher
    Posted Dec 23, 2007 at 9:00 PM | Permalink

    Odd. today I was looking at GISS land versus Land-sea.

    And I was wondering

    what was the relationship betwetten the Land temps and the land-sea BEFORE the bucket adjust
    and what is the relationship since the bucket adjust.. and could such a change be detected.

    UC and I have been playing with a regime shift detection program ( mostly on unthreaded )
    he thought the approach could be used to detect errors in in method.. If you can’t find his
    post or mine on the matter, just holler and I’ll go hunt the stuff down.. Just look for
    UC on the current unthreaded ( he was slummin I guess)

  3. Steve McIntyre
    Posted Dec 23, 2007 at 9:05 PM | Permalink

    I’ve been using Achim Zeilis software in R for regime detection in hurricanes. Very easy to use. But I think that you’re trying to read too much into the software: the Dec 1941 adjustment is already known to exist. The issue is the validity. I don’t think that this has to be determined through examination of the details and I doubt that a changepoint test will say very much,

  4. Geoff Sherrington
    Posted Dec 23, 2007 at 9:19 PM | Permalink

    Seve, you are closer to the action that most of us but I think I speak for many when saying that we do not know which past temp data can be trusted or used for futher work. There has been so much fiddling, so many adjustments, so many questionable assumptions that we know not where to turn.

    Your post is on SST, but I feel that the larger corrections of TOB on nearby land have to be made more acceptable before a solution can be found to your SST questions. I repeat here part of an email from Phil Jones to me in March 2006 – about when the IPCC was writing that the science was settled.

    > Back to the early Australian data: I mentioned the other day that there are problems with the early Australian data. If you look at the pdf I sent earlier (Figure 2, panel f (GHS_EMAIL ME FOR A COPY IF NEEDED)you’ll see that climate models given SSTs can’t reproduce Australian land temps (the black line) prior to around 1910. The models can in other continents of the world.

    Most importantly for Australia, they can over NZ.

    I would suggest you look at NZ temperatures. The attached paper sort of does this, but only as a part of other regions of the S. Pacific. There are earlier papers by Folland and/or Salinger on NZ temperatures. What is clear over this region is that the SSTs around islands (be they NZ or more of the atoll type) is that the air temps over decadal timescales should agree with SSTs.

    This agreement between SSTs and air temperatures also works for Britain and Ireland. Australia is larger, but most of the longer records are around the coasts.

    So, NZ or Australian air temperatures before about 1910 can’t both be right. As the two are quite close, one must be wrong. As NZ used the Stevenson screens from their development about 1870, I would believe NZ. NZ temps agree well with the SSTs and circulation influences.

    My point is that SST can be adjusted up or down to force agreement with nearby land, or vice versa. The nearby land can’t be trusted too much because the TOB correction lacks resolution and generalises. I think it has to be fixed first. The required max and min are usually in the data, but one has to go through and assign the obs to the correct day. Becuse the TOB correction can be larger than SST, it needs fixing first.

    Another point is that there were land problems of calibration overlap when moving from thermometer to thermocouple MMTS. Is there a similar errror when going from buckets to engine intakes? Another Q: Is there a TOB correction for SST (say even day versus night) or is the sea condidered homogenous over a day?

    Sorry for the usual problem of more questions than answers.

  5. Leif Svalgaard
    Posted Dec 23, 2007 at 9:21 PM | Permalink

    But what solar input are these temperatures compared to? It is as if the solar input is taken as given [somehow] while it may be just as systematically wrong as the temperature runs.

  6. steven mosher
    Posted Dec 23, 2007 at 9:27 PM | Permalink

    re 3. ya..i think your right, UC thought it interesting so I wasnt inclined to trash heap it.

    If the adjustment is flawed, then one easy way to detect it might be with change point.
    I suspect youre right and the change is not detectable at high levels of significance

  7. Harry Eagar
    Posted Dec 23, 2007 at 10:33 PM | Permalink

    When I first heard of the CO2 forcing idea,in the mid-’80s, it sounded reasonable. But very shortly after that I read an interview with a Harvard researcher (no idea now who) who was explaining how he was going to use SSTs, of which the Challenger data were the biggest single part for early years.

    He also noted that he had applied to the British Museum (Natural History) to borrow Challenger’s bucket in order to determine the convection losses during the period when the bucket was being hoisted aboard.

    There’s such a thing as being too fine. Anyhow, when I tbink of the ‘bucket problem,’ being too fine is what it means to me.

  8. Posted Dec 24, 2007 at 1:32 AM | Permalink

    While it is nice to see a refutation of the bucket correction, it is surprising that there is no attempt to calibrate SSTs against exposed coastal weather stations. The failure of Folland and Parker to match e.g. Valentia’s record is obvious. There must be a chain of such stations, able to indicate trends if not actual temperatures — the Azores, Canaries, Bermuda etc in the Atlantic for example. The trends should be good enough to demonstrate the error in the assumptions behind the correction. Where the calibrating stations are airfields, there should be good cloud data as well — aircrew are rather keen on knowing about low cloud over their runway.

    However, removing F&P not only removes the unexplained fall in temperatures from ’41 onwards, it generates a large hump starting in 1939 (see graphs of bucket correction in this blog). Then we have the problem of explaining that hump. What do the statistics say about an excursion of .5 deg above the steady trend line from 1910 to present? That’s a remarkable el Nino.

    JF

  9. Posted Dec 24, 2007 at 3:13 AM | Permalink

    I’d just thought I’d tell you that I can recall dangling a bucket over the side of a ship as late as the 1980’s in order to obtain SST measurements. None of us who did the chore enjoyed doing it and as a result we did it as quick as possible with little regards to professional standards.

    My experience would question the accuracy of any SST results that has involved obtaing water samples over the side of a ship, particularly in the colder oceans.

    Martin J

  10. Geoff Sherrington
    Posted Dec 24, 2007 at 4:56 AM | Permalink

    Re #9 Martin

    Agreed, but that’s no reason to apply an “adjustment” of a certain sign or magnitude, aspecially when you have a good think about Paltridge and Woodruff (1981); and then to apply other “adjustments” on top of this and so on and so on. The Paltridge and Woodruff trend is real, unless I can be convinced otherwise. Also, logically, the 1941 excursion has to be teated as Natural until a better explanation is contrived. It is too widespread and too similar smell like an accident of instrumentation.

    Re #8 Julian Flood
    There have been attempts to reconcile shore stations with nearby SST. See my quote at 4 from Phil Jones for but one example. But it seems to me that “inland” can be a large mountain range with glaciers (a la New Zealand) or a flat land with periods of hot strong winds (like Perth WA). I suspect that the differences in the interiors (say for stations on the west coast of each of these 2 places) makes the near-shore records difficult to reconcile with each other when there are winds from land, thus the SST also. BTW, whatever happened to data from moored buoys?

  11. Posted Dec 24, 2007 at 7:04 AM | Permalink

    Re#10,4: Your correct, of course; but as you say: “more questions than answers”

  12. Posted Dec 24, 2007 at 9:12 AM | Permalink

    This is an important topic, something I’ve long been concerned about. It will be interesting to see if you can shed light on the sea surface temperatures in the pre satellite era. The sampling prior to 1920 is too sparse to do much with, but the period 1920-1975 is certainly an important period to sort out.

  13. DocMartyn
    Posted Dec 24, 2007 at 9:30 AM | Permalink

    In WWII there was a switch from coal to oil for cargo ships. In the late 60’s and early 70’s the big switch was made to diesel. Oil burners use sea water to make steam, so they need a lot of it. Diesel powered ships use seawater for cooling the engines, and so require only a little,

  14. hans kelp
    Posted Dec 24, 2007 at 9:59 AM | Permalink

    A commentary to seewater measurements.

    To be honest, I think it´s beyond all probability that one will be able to utilize measurements of seewater temperatures in buckets to anything scientificcally in detail. To claim the opposite is, I think , scientific hubris at best, and is only an expression of wishful thinking. I will try to explain.

    If you throw a bucket into the water , the bucket has its own temperature from the beginning which in itself is 1) higher than the temperature of the water when the temperature of the air is higher than the temperature of the water and 2) lover than the temperature of the water when the temperature of the air is lover than the temperature of the water and 3) the temperature of the bucket equals that of the water when both the bucket and the water has a temperature of zero degrees celcius or + 273,15 degrees Kelvin.

    The situation in number 3 is in my opinion rather unlikely and of no importance statisstically anyway. This means that a lot of corrections has to be applied to find the correct temperature ( whatever it is ), and to simplify my illustration I am going to use a bucket made from the metal known to most sailors as zinc.

    When using the “bucket”-method you have to take two situations in consideration. Is it a modern fast moving ship or is it a slow moving ship?
    The ships that I have been sailing myself, both as an able seaman as well as an officer ( navigation officer ), has traveled at speeds where I consider it worthless to use the bucket method. Why?

    If you go by ten knots or more the bucket will start to jump off the water and the speed of the ship will work up a wind which is going to chill or warm the water in the bucket on its way up on deck. To this you have to add or distract the speed of the wind itself to the speed of the ship. On smaller ships with a distance of a few meters or feets from the surface to the top of the bulwark ( the rail ) it doesn´t matter much. But on bigger ships with a distance from the surface of the water till the rail of lets say ten maybe twenty meters , it really matters because you have to pull the bucket on deck by hand, at least on those ships that I sailed myself, and it takes maybe a minute or two. In that interval the wind is working on the bucket and the water in it and thereby influences the temperature of the water, ±.
    Furthermore, the water in the bucket will be mostly from the foam blown off of the crest of the waves and can by no means be compared to the water whose temperature you were supposed to measure. As you can see, it looks very difficult to put any reliance on the “bucket”-method when we talk about big fast moving ships. How do they correct for this? Guesswork I suppose.
    How about the “bucket”-method on smaller ships then? Let´s go through it.

    When you have pulled the bucket out of the sea, the water in it has from the very first “lift-off” been under an influence physically known as “the first clause of thermology” which simply means that the temperature is continually influenced by its surroundings, ±. By surroundings I also mean the surrounding air and the influence from the buckets position while measuring the temperature. Is it put in the shadow or is it put in the sun. The temperature you will get is in the end a representation of the sum of all of the surrounding influences. And this goes also for the influences from the thermometer you put into the water. One could say that what was mentioned under point 1, 2 and 3 on top is at work here too. I´m sure you will agree that time is a factor that is very important for the end result.
    This are just a few of the points that need proper correction before you can use them for anything else but pure entertainment. A list of questions I think you have to answer before you can go on from this is as follows

    1) How much water is in the bucket when you measure the temperature and
    what is the mass of the water compared to the mass of the bucket ?
    ( for use in calculations according to “the first clause of thermology”)

    2) What is the temperature of the bucket when you throw it into the water ?
    ( for use in calculations according to “the first clause of thermology”)

    3) What is the temperature of the framework of the thermometer when you
    put it into the water? ( Some I have seen were rather big!)
    (for use in calculations according to “the first clause of thermology”)

    4) What is the mass of the framework of the thermometer? And the chemical
    composition of the material(s) used herein?
    ( for use in calculations according to “the first clause of thermology”)

    5) Did you give the thermometer enough time to stabilize? (Are people
    patient enough??)

    6) What is the overall quality of the thermometer? ( Are there any stated
    elements of uncertainty (± – corrections to be used ) from the
    manufacturer?)
    7) Are the thermometers used on all ships all of the same kind, made of the
    same material and by the same manufacturer?

    8) Is it exactly the same procedure used on all ships?

    9) What is the percentage of salt in the seawater you are measuring?

    10) What is the humidity and the air pressure during measurements?

    As you can see there´s a lot of uncertainties you have to account for before you can decide what the correct SST is. To this you may also add that the temperatures are not taken on the exact same position and at the exact same time under the exact same weather conditions. Has it been overshadowed for some time, days or weeks, has the sea been dead calm for some time with sunny or not sunny weather, or has it been raining for some time before the measurements were done. All of this influences the SST, and to the best of my knowledge , those people claiming that they will be able to make out the correct SST from such a mess of uncertainties must be heavily departed from reality.

    I don´t know if SST is measured by using the “bucket”-method anymore. From my time as a mate we were informed by the engineers on duty at every change of watch ( every four hours ) what the temperature in the seawater intake was. The seawater ( coolingwater ) intake is positionend just above the bilge keel which is the rounding from the horizontal to the vertical down at the bottom at the ship. This position is sometimes several meters below the sea surface. Maybe ten to tventy meters on bigger ships but six or seven meters is quite normal. Anyhow , what you measure in a ships seawater intake is far from the SST you want to know about within tens of a degree, and it can by no means be used directly unless you know for sure the right corrections for both influences from the hull, the friction of the hull against water and the proper temperature correction for the depth of the water intake.
    To my knowledge no such corrections were known during my time as a mate. 1976-1989.

    Merry Christmas to all.
    H.K.

  15. Posted Dec 24, 2007 at 10:41 AM | Permalink

    I don’t think I worked on the same boats as Hans(re#12) but I agree with everything he says. I vaguely remember using plastic buckets in the North Sea: I don’t know how the material the buckets were made from effected results, but I suspect it did. I remember my hands being numb with the cold because of all the icy water we had to handle. We just did the job as quick as possible.

  16. Posted Dec 24, 2007 at 10:42 AM | Permalink

    #13 Doc Martin,

    Steam was recycled in steam ships. Evaps were only used for make up water. The continuous loss steam engine went out with the steamboat era (for the most part). The reason for this is that “natural” water would foul the boiler tubes. The amount of cooling required is strictly a function of engine efficiency and size.

  17. Posted Dec 24, 2007 at 11:02 AM | Permalink

    The reading of the intake temps on steam ships was to get an idea of what kind of vacuum you could pull in the condensers. That would give you an idea if there were leaks to the atmosphere in the condensers and to determine if the engine gang/captain was wasting fuel. This is all with respect to turbine engines. I have no experience with steam piston engines.

  18. Jon
    Posted Dec 24, 2007 at 11:34 AM | Permalink

    In WWII there was a switch from coal to oil for cargo ships. In the late 60’s and early 70’s the big switch was made to diesel. Oil burners use sea water to make steam, so they need a lot of it. Diesel powered ships use seawater for cooling the engines, and so require only a little,

    I’m a bit confused by your remarks. Most large cargo ships use bunker oil (Bunker C, No. 6 fuel oil) not diesel (No. 2 fuel oil). The latter being much, much too expensive and unnecessary: something akin to building a car to run on jet fuel. The reason for this is that heavy fuel oil is very cheap but usable given a sufficiently large plant and continuous operation. Diesel of course is used by small vessels–but this is ‘oil’ in the sense that it is the same stuff people refer to as ‘home heating oil’.

  19. Dennis Wingo
    Posted Dec 24, 2007 at 11:41 AM | Permalink

    I am impressed with the Paltridge and Woodruff paper for not only presenting the data, but presenting all of the errors and ways that they could be wrong. A vast number climate related papers today read like marketing documents for AGW.

    An interesting modern point is that Christy’s data is showing a pretty dramatic fall off in southern Hemisphere temperatures which would imply an increase in the temperature gradient between the tropics and pole, at least in the southern hemisphere, which would explain the cold temp records down under this past winter.

  20. SteveSadlov
    Posted Dec 24, 2007 at 12:37 PM | Permalink

    http://en.wikipedia.org/wiki/Emma_M%C3%A6rsk

  21. Larry
    Posted Dec 24, 2007 at 1:21 PM | Permalink

    One more time, the climate folk are using a butterknife for a screwdriver. We have weather stations used as climate stations. We have trees used as trend recorders. Now we have ship intakes used as ocean trend recorders. How on earth are we supposed to arrive at error bars for this?

  22. bender
    Posted Dec 24, 2007 at 1:31 PM | Permalink

    Error bars? What do you want those for? You’re just trying to spread fear, uncertainty and doubt.

  23. steven mosher
    Posted Dec 24, 2007 at 3:43 PM | Permalink

    RE 22. Error bars? I thought that was place on the other side of the tracks where
    you chugged mistakes with a water back.

    RE 21. Error bars are +-2C until someone convinces me otherwise. I’m gunnar stubborn on this.

  24. Larry
    Posted Dec 24, 2007 at 3:53 PM | Permalink

    23, If you could only make glue that stubborn…

  25. Larry
    Posted Dec 24, 2007 at 3:58 PM | Permalink

    23 you’re gunnar be sorry you said that…

  26. Posted Dec 24, 2007 at 4:21 PM | Permalink

    Steve S #20

    That is pretty fast traveling for a cargo ship. The wiki claims 25.5 knots. No mention of cruising speed. Probably on the order of 20 knots as power goes up as the cube of speed.

  27. steven mosher
    Posted Dec 24, 2007 at 5:08 PM | Permalink

    RE 25. you are a bad man.

  28. bender
    Posted Dec 24, 2007 at 5:13 PM | Permalink

    error bars is where you gunnar make dating errors.

  29. Larry
    Posted Dec 24, 2007 at 5:21 PM | Permalink

    Only if you jae walk to get to the error bar.

  30. Larry
    Posted Dec 24, 2007 at 5:26 PM | Permalink

    27, bad boys, bad boys, watcha gunnar do? watcha gunnar do when they’re as stubborn as glue?

  31. bender
    Posted Dec 24, 2007 at 5:35 PM | Permalink

    happy holidays, guys. myself, i can’t wait to see jae’s reaction when he opens up all his presents from lucia, judith, andrey. lucky guy. drowning in answers. me, i got nothin but unanswered questions.

  32. hans kelp
    Posted Dec 24, 2007 at 5:56 PM | Permalink

    Re #26.
    The first series of container ships from the Maersk company,
    the so called C-series, were build to a cruising speed of 25 knots.
    I was on board the M/S Clifford Maersk as an able seaman in 1974. At that
    time their cruising speed was only 23 knots because of restrictions
    due to construction failure in the bows which were shortened by heavy seas
    at about 5 feet. Also, it was not allowed to carry containers
    on the number one hatch because they used to be washed overboard.
    They didn´t want to slow down because of heavy seas. When we left Hamburg
    in Germany our first port of call was Singapore. The trip lasted exactly
    three weeks. Every sailor know that that´s a pretty fast trip, and I
    should like you to try it .It´s really something special!

  33. steven mosher
    Posted Dec 24, 2007 at 7:13 PM | Permalink

    bender and larry: you are evil

  34. Andrey Levin
    Posted Dec 25, 2007 at 12:58 AM | Permalink

    Jon:

    It was switch from coal to oil to fuel steam boilers, and later from steam boilers to diesel engines, not diesel fuel (although it took quite a time to teach propulsion marine diesel engines to run happily on such junk as bunker oil).

  35. Andrey Levin
    Posted Dec 25, 2007 at 1:21 AM | Permalink

    Ostap-Suleyman-Berta-Maria-Bender-Bei:

    http://en.wikipedia.org/wiki/Ostap_Bender

    Thanks for unexpected ‘present…andrey’ from#31.

    Merry Christmass and happy New Year to you and your family!

    Andrey Levin,

    proudly wearing ‘POV like A.L., who freely admits to being an agenda parasite’ from CA and ‘tool of carbon-emission interests on payroll of Russian oil tycoons, who will sell the truth and his own future for a few dollars’ from GCC.

    Cheers!

  36. Robert Wood
    Posted Dec 25, 2007 at 2:16 AM | Permalink

    An innocent, but rhetorical, question, not relating just to this thread:

    Why can’t measurements be made and taken for what they are, rather than be adjusted all the time according to some convenient excuse to squeeze a point?

    Yes, discrete measurement series may not be “global” series, but if the trends are “global” then surely they would be evident in the local, discrete measurement series.

    I say there should be no need of adjustments. The global trends should still be evident in the local, or discrete, series. Therefore I am skeptical of “adjusted” data.

    When I was at school, I would have got an “F” for “adjusting” my data, even been called a cheat.

  37. Geoff Sherrington
    Posted Dec 25, 2007 at 2:32 AM | Permalink

    Paltridge and Woodruff (1981), at the core of this thread, note that buckets were used until 1960 in their graphed data. This makes talk of other forms of ship temperature measurement a bit academic for the time being, unless P & W have made an error. In any case, how do the possible error causes discussed above disturb the pattern in graph 4 of P & W other than adding noise? Perhaps the postulate that anything inside +-2 deg C should be disregarded is correct and all hope abandoned.

    Is it possible that the rise in “unknown” methods of measurement post WWII could be because of restriction of information for security on military vessels? There was a lot of work on transmission of communications and enemy craft detection signals to/from submarines starting about then.

    There can be little doubt that the data from Paltridge and Woodruff have some meaning, albeit a lot of noise. My inclination is that some reflection of SST is there and that hope should NOT be abandoned.

    But, there should be a clear exposition as to why SST needs to be reconstructed and then, the size of acceptable errors. A useful definition of SST should be evolved (depth of sample, lack of disturbance from ship, correlation with wind, storm, etc). I’m amused by the the suggestion that wily ships captains tracked currents to shorten voyages, perhaps with help from 4-hourly temp logs. Feedback at work. Put that into your global climate model.

    The main reason I can imagine for getting an accurate, defined, global SST data set is to help settle the question of change in global solar irradiance. It would help along the way if a few data manipulators were outed.

  38. STAFFAN LINDSTROEM
    Posted Dec 25, 2007 at 3:33 AM | Permalink

    NOW it’s Christmas Eve in all 48 contigious
    PREDICTION of Nobel Peace Prize to Al Gore/IPCC..100%
    (as a sidenote we found some Norwegian fullsize flags
    in a container behind Telenor’s office in Stockholm, also one Swedish
    I should say it’s a bigger crime to throw away flags like
    that than to burn them and finders keepers…??!!)PREDICTION of Norwegian NPP secretary speech Dec 10 would be written 908 (CMIIW) by Mark Lynas or you could easily
    believe it at least …YOUR CHOICE

    PREDICTIONS OF WARMEST YEAR: FAILED (rather miserably)

    PREDICTIONS of N Atlantic Hurricane season, well
    numbers not to bad but strength and duration not much, hence LOW ACE
    Bad landscape planning kills many more than nature,
    just like New Orleans etc etc etc …
    PREDICTION of Australian drought to go on “forever”

    YKTA…
    PREDICTION that an UN organisation, Red Cross etc to nominate
    2007 “Worst year ever catastrophic disasterwise” To my
    uttermost astonishment they did NOT do that, not even in
    advance but again, I may have missed some…
    PREDICTION for global lower tropospheric temperatures to fall
    0.5C or so in 9 months Here I NEED YOUR HELP Has anybody seen THAT??
    A second call to John Christy??
    Hmm, after this, I take a break for more VEGETARIAN (READ GERMAN BEER!)Xmas food
    TO BE CONTINUED

  39. STAFFAN LINDSTROEM
    Posted Dec 25, 2007 at 3:40 AM | Permalink

    #38 …As Mosher would have put it??!! STAFFAN
    too many German beers??…”…Christmas eve…”
    OF COURSE it’s Christmas Day…[3:33 halftimedevil...]

  40. Posted Dec 25, 2007 at 3:55 AM | Permalink

    bender says:
    December 24th, 2007 at 5:35 pm

    happy holidays, guys. myself, i can’t wait to see jae’s reaction when he opens up all his presents from lucia, judith, andrey. lucky guy. drowning in answers. me, i got nothin but unanswered questions.

    You have harder questions. :-)

  41. steven mosher
    Posted Dec 25, 2007 at 12:01 PM | Permalink

    RE 36. There is nothing wrong with adjusting data, provided you do the following.

    1. Preserve the raw data.
    2. Provide a clear analysis of why you think the data requires adjusting.
    3. Provide the code/analysis you conducted to determine #2.
    4. Make your adjustement transparent.
    5. Analyze the data with and without the adjustment.
    6. provide each and every file you created/modified in your analysis

    Anytime you do a study you run the risk of Systematic errors creeping in. These
    errors can be identified and compensated for but the process needs utter transparency

  42. Dan White
    Posted Dec 25, 2007 at 3:39 PM | Permalink

    This isn’t really rocket science. Has anybody put a bucket into the water and read the temperature while dangling a thermocouple in the water at the same time for comparison? I admit I haven’t read the papers mentioned in this thread, but judging from all the uncertainty expressed in the comments it makes me wonder if anyone has done an exhaustive study on how various bucket sampling methods REALLY affect the results. It seems to be something very much reproducible. Even if you knew how high above the water and how fast the HMS Beagle sailed over 100 years ago, you could get results very similar to actual measurements at that time.

  43. EW
    Posted Dec 25, 2007 at 3:53 PM | Permalink

    Larry said:

    How on earth are we supposed to arrive at error bars for this?

    Plus all the uncertainties surrounding Waldo, aa, TSI and solar cycles…I’m feeling rather nihilistic concerning both temperature reconstructions and forecasts/modeling. Maybe we should drop it altogether.

  44. Larry
    Posted Dec 25, 2007 at 4:07 PM | Permalink

    42, from a purely mechanical standpoint, that’s probably right. But from an organizational standpoint, there’s really no reason for these crew members to be diligent. Getting these numbers isn’t critical to their mission. You’re depending on the work ethic of a crew from Bumtulips. Excuse me if I don’t think the numbers are going to be accurate under those circumstances.

  45. maksimovich
    Posted Dec 25, 2007 at 4:16 PM | Permalink

    re 41

    Here you go

    http://www.marineclimatology.net/wiki/index.php?title=Main_Page

  46. maksimovich
    Posted Dec 25, 2007 at 4:24 PM | Permalink

    re 42

    Results show that the temperature change in bucket SST resulting from the air–sea temperature difference can be detected. The analysis suggests that bucket SST may be in error by a fraction from 0.12° ± 0.02° to 0.16° ± 0.02°C of the air–sea temperature difference. When this temperature change of the bucket SST is accounted for, a warm bias in engine-intake SST in the mid- to late 1970s and the 1980s was found to be smaller than that suggested by previous studies, ranging between 0.09° ± 0.06° and 0.18° ± 0.05°C. For the early 1990s the model suggests that the engine-intake SSTs may have a cold bias of −0.13° ± 0.07°C.

    http://ams.allenpress.com/perlserv/?request=get-abstract&doi=10.1175%2FJTECH1845.1

  47. Harry Eagar
    Posted Dec 25, 2007 at 4:53 PM | Permalink

    In ‘At Sea with the Scientifics,’ which is a collection of letters home written by a clerk on HMS Challenger, it is reported that it was a hot day in Honolulu, temperature 125 F.

    No doubt the thermometer was left in the sun on a piece of metal or something, but that was the same thermometer used to measure the seawater samples.

    That gives an error bar of plus/minus 30 F, based on the fact that no temperature above 95 has ever been recorded at Honolulu International Airport.

    So when masksimovich posts a link to a report claiming to detect differences to hundredths of a degree, we know how much credit to give it.

  48. Willis Eschenbach
    Posted Dec 25, 2007 at 5:34 PM | Permalink

    Perhaps I came in late or missed something. I had thought that various authors used the sea surface temperature (SST) as a reasonable approximation of the marine air temperature (MAT). Paltridge and Woodruff, for example, say:

    The two types of data are compatible in the present context since the concern is with change rather than absolute value. In any event, Cayan (1980) has shown that SST variability is generally a good indication of surface air temperature variability on time scales of more than a month.

    This makes sense, since the thermal mass of the ocean is hugely larger than that of the air, so if the two are in contact for any amount of time, the air will quickly take up the temperature of the ocean.

    However, this would also imply that on average, the air would be a bit cooler than the ocean about as often as the air would be a bit warmer than the ocean. This in turn would mean that any difference between bucket and intake temperatures would be related, not to air temperature, but to wind speed (evaporative cooling) and sunshine.

    The analysis by Kent and Kaplan (thanks, maximovich) says they found a very small difference (ranging from 0 to 0.2°) between bucket and intake temps … by using a “novel type of linear regression” … why do the hairs stand up on my neck when I read that? However, they used “nighttime observations made at moderate wind speeds, allowing the effects of solar radiation and strong vertical gradients in the upper ocean to be neglected” to make their comparison. As a result, it is not clear what their study means about real-world observations. Buckets of water on deck are cooled by the wind and heated by the sun. Me, I’d be very cautious about any estimates of how much each of those factors affected the real world measurements. While I’d love to ignore the sun, wind, and vertical gradients as the authors did … I don’t think we can do that.

    So while as Dan White says, “it’s not rocket science”, it’s also not simple. Yes, we can say that there’s about a tenth of a degree between night-time moderate wind bucket and intake readings. But how much does that tell us about some swabbie in 1903 hoisting a bucket of water up over the side, leaving it in the sun and wind while looking for the thermometer, and then reading the temperature to the nearest degree? I don’t think we can say, to an accuracy of a tenth of a degree, how much that changed the temperature of that bucket of water, or of the global average of buckets like that.

    So absent some good reason to mess with the data, I think I’d leave it alone. Yes, there may be a bias error due to a change in the measurement method. But without strong evidence, not just “a novel type of linear regression” but actual evidence that the change in measurement methods has a significant effect (not just an effect but a significant effect), I wouldn’t adjust it.

    I say this in part because of a generic reluctance to mess with any data, and partially because in this instance even the sign of the correct adjustment will change depending on the circumstances. This particular error is different, in both magnitude and sign, for different regions, different seasons, and different times of day. Because of this, any global adjustment we might employ will make the error greater in about half the cases … not good.

    w.

  49. maksimovich
    Posted Dec 25, 2007 at 6:03 PM | Permalink

    re 48

    As a result, it is not clear what their study means about real-world observations

    Gulev et al find some interesting results on the o/a flux interchange.

    Sampling uncertainties in the voluntary observing ship (VOS)-based global ocean–atmosphere flux fields were estimated using the NCEP–NCAR reanalysis and ECMWF 40-yr Re-Analysis (ERA-40) as well as seasonal forecasts without data assimilation. Air–sea fluxes were computed from 6-hourly reanalyzed
    individual variables using state-of-the-art bulk formulas. Individual variables and computed fluxes were subsampled to simulate VOS-like sampling density. Random simulation of the number of VOS observations and simulation of the number of observations with contemporaneous sampling allowed for estimation of random and total sampling uncertainties respectively. Although reanalyses are dependent on VOS, constituting an important part of data assimilation input, it is assumed that the reanalysis fields adequately
    reproduce synoptic variability at the sea surface. Sampling errors were quantified by comparison of the regularly sampled (i.e., 6 hourly) and subsampled monthly fields of surface variables and fluxes. In poorly
    sampled regions random sampling errors amount to 2.5°–3°C for air temperature, 3 m s1 for the wind speed, 2–2.5 g kg1 for specific humidity, and 15%–20% of the total cloud cover. The highest random sampling errors in surface fluxes were found for the sensible and latent heat flux and range from 30 to 80 Wm2. Total sampling errors in poorly sampled areas may be higher than random ones by 60%. In poorly sampled subpolar latitudes of the Northern Hemisphere and throughout much of the Southern Ocean the
    total sampling uncertainty in the net heat flux can amount to 80–100 W m2. The highest values of the uncertainties associated with the interpolation/extrapolation into unsampled grid boxes are found in subpolar
    latitudes of both hemispheres for the turbulent fluxes, where they can be comparable with the sampling errors. Simple dependencies of the sampling errors on the number of samples and the magnitude of synoptic variability were derived. Sampling errors estimated from different reanalyses and from seasonal forecasts yield qualitatively comparable spatial patterns, in which the actual values of uncertainties are controlled by the magnitudes of synoptic variability. Finally, estimates of sampling uncertainties are compared with the other errors in air–sea fluxes and the reliability of the estimates obtained is discussed.

    JOC JAN 2007

  50. Posted Dec 25, 2007 at 6:56 PM | Permalink

    Willis Eschenbach says:
    December 25th, 2007 at 5:34 pm,

    Willis,

    We just apply the TOBS adjustment and that will fix every thing. Plus it has the advantage of already having been worked out. How hard can it be?

    \sarc off

  51. Posted Dec 25, 2007 at 7:09 PM | Permalink

    Larry #44,

    I can also tell you that log entries can also be “interpolations” if the crew member was lazy, missed a reading, or failed to record a reading.

    In some cases where readings were required 4 times an hour several hours of data might be interpolated. This was on a
    Navy ship. Commercial ships might not be as diligent. Perhaps one of our commercial sailors can give an insight.

  52. Larry
    Posted Dec 25, 2007 at 7:18 PM | Permalink

    51, I know how plant operators on graveyard shift can be, and I figure that cargo ship crew are probably about three notches below that, so I can only imagine where most of those numbers really came from.

  53. Willis Eschenbach
    Posted Dec 25, 2007 at 8:04 PM | Permalink

    As a preamble, I may have mentioned before that I have spent a good chunk of my life at sea, from the tropical Pacific to the frozen Bering. I have gone there as a commercial fisherman, both before and behind the mast; as a long-distance sailor; as a diver, both commercial and sport; as a fishing guide; doing maritime salvage; as a boat builder; and as a surfer.

    Dipping up a bucket of water from the deck of a moving vessel can be difficult, depending on speed and sea conditions. The usual technique is to toss the bucket in the direction of movement, mouth forwards, so the bucket starts to fill immediately and is filling as the boat is moving towards it. Then you pick the bucket vertically out of the water when the boat’s motion brings you even with it. DONT let the boat pass the bucket without picking it up, because if the bucket gets dragged behind, it can yank you out of the boat. For the same reason, NEVER wrap a bucket rope around your hand or wrist, or tie it to you.

    Yes, I understand that there are insulated buckets, and buckets with extra weight to take them down deep, and Nansen bottles if you want to go that far. But speaking as a swabbie here, I’d say the usual sequence of events, be it 1882 or 1952, was the First Officer passed it on to Ships, who passed it on to the Gunny, who told the Leroy the Cook’s Assistant to grab a bucket of seawater, dammit … whereupon Leroy, overjoyed at being released, went outside into the cold and rain and wind and dipped up a bucket of seawater using the technique above. About then, his brain caught up with his feet and he realized he could light up a smoke, so he sat down the bucket and lit up, but only managed two puffs before the Gunny with eyes everywhere told him to get his !@#$% ass back inside, so he did, and set down the bucket, and peeled potatoes again. In about five or ten minutes the First Officer arrived with the thermometer, took it out of its case and put it in the bucket, waited a minute or less or more, wrote “Sea Temperature 36° ” in the log, cased the thermometer and took it away, the day’s final paperwork done. All that remained was for the Gunny to tell Leroy to not forget to empty the !@#$%^ seawater bucket again, just so he could catch Leroy smoking again … he always allowed him exactly two puffs …

    Compared with the errors inherent in that whole process, I must confess I’d be reluctant to adjust an alleged tenth of a degree in any case, no matter the putative cause.

    Now, the problem here is the merging of two very different datasets. By “very different”, I don’t mean that they are very different measurements of the same variable.

    Buckets and intakes are not even measuring the same thing. Buckets measure the sea surface temperature, SST. Engine intakes measure what might be called “BSST”, for “below sea surface temperature”. The main requirement for the location of the intake water pipe on a boat is that it never suck air, that it is always some distance below the surface, hence BSST.

    So what is the usual difference between SST and BSST? Well, as a surfer and a diver, I’m in a good position to answer that. The answer is, like the atmosphere, it is very variable. Sometimes the surface is warmer than a metre or so down, and sometimes it is colder.

    An oddity of the ocean/atmosphere interaction is that the atmosphere is stable at night and unstable in the day. The ocean, on the other hand, is exactly the reverse – unstable at night, and stable during the day. At night, the top layer of the ocean cools. The top becomes denser and sinks, and is replaced by warmer underlying water.

    During the day, on the other hand, a warm layer of water forms, with the hottest part at the very surface where the sun is brightest. Because the hottest is the least dense, the layering is stable, and there is no thermal overturning.

    Now, there is one curiosity of this system that has not received adequate attention. It functions as insolation against further heating, while encouraging further cooling. This is because the heat does not mix downwards. The thin layer of heated water tends to stay at the surface. This maximizes heat loss through radiation, conduction, and evaporation.

    The net result is that on a clear, calm day, I’ve swum in the ocean in bathtub-warm water. The warm water was in a layer shallow enough that my hands would go down into cold water with each stroke, while my body was warm. If a day like that is exceptionally warm, the resulting layer of warm water may stay on the surface radiating and convecting and evaporating all night. In such a case, there is no nocturnal overturning at all. The layer is still there in the morning.

    If a cool day follows, then the following night that thin layer of warm water may finally cool down enough to be dense enough to sink. Note that almost all of the heat in that water has gone back upwards in radiation or conduction or evaporation. Very little has gone deeper, into the mass of the water.

    Once that top layer cools, however, the cool water will mix downwards into the deeper ocean until it reaches equilibrium. Since the cool water naturally sinks and the warm water does not, the flow of heat into the mass of the ocean is reduced. And at the same time, the flow of heat from ocean to atmosphere is maximized by keeping the hottest, most radiant water at the top.

    (This, of course, is all “ceteris paribus”, or “all other factors being equal”, which of course they never are. Note that this simplified explanation doesn’t include a wide variety of forcings and feedbacks like wind and wave affecting evaporation and cloud nuclei generation … rain affecting temperature and salinity … that kind of thing …)

    What does all of this mean for buckets and inlets, for SST and BSST? Well, we can say the following:

    1) SST greater than BSST is stable.

    2) BSST greater than SST is unstable, and will overturn and mix downwards.

    3) On average, global SST is greater than global BSST.

    How much? Ah, there’s the rub …

    A final curiosity … this suggestion, that on theoretical grounds we should see not a warm, but a cold bias in the intake data versus SST, is borne out by the study cited by maximovich, which found such a bias from 1990 onwards.

    w.

  54. hans kelp
    Posted Dec 25, 2007 at 9:24 PM | Permalink

    Re#51
    I was once a commercial sailor. As an able seaman I have tried a few times to pull on board a bucket made of cinz.The boatswain did the measurements. Later, when I became a mate myself, we reported the observed meteorological data to the World Meteorological Organization. This included the water temperature taken from the sea water intake at just above the bottom of the ship.I think a lot of ships have been doing that. Why they wanted to use buckets in my time as an able seaman might have been because they thought it to be the most correct method. Afterwards,when I think about the difficulties I met in trying to get that damned bucket to get filled with seawater, I can see that the efforts I put into it were not justified, and commercial ships do not stop to fill a bucket with sea water. No way.Please read my commentary in # 14. Before I finish this I can give you an example: I sailed a Norwegian car carrier “Dyvi Oceanic” as an able seaman. On that ship we had ( I had, maybe because I was the only Dane on board )to heave the bucket on board from the poop deck which was the lovest part of the ship ( astern). The ships cruising speed was about 13-14 knots. When you threw the bucket into the water it actually never happened to get into the water. It simply jumped off the sea surface because of the ships speed ahead. What we caught was simply the foam blown off from the crest of the waves. Next, the bucket had to be pulled on board the ship again. The vertical distance from the poop deck to the sea surface was exactly 16 meters.I know that definitely because I once jumped into the water from the poop deck when the ship stayed at the port of Jeddah down in the red sea. Afterwards I spoke to the chief mate and we took a line, dropping it to the surface and measured the height. What I thought to be a height of ten meters proved to be 16 meters ( I did hurt my shoulder badly !). Fully loaded the difference was within a meter less than 16 meters ( ten decks with empty cars ( 2200 pieces of small Herbies ) doesn´t weigh anything at all ). Well, then again, that means that I had to heave the bucket on board from a distance that was much more than 16 meters ( take the square root of 15^2*15^2 ) and very much closer to 21 meters. It´s not an easy job. It took at least a minute or two. In that time the wind worked on the bucket with its chilling effects, whatever it amounted to. To make it short, the temperature measurements from ships that is believed to be SSTs can be anything else than just that. How on earth can they correct for something they don´t know anything about. Unbelievable.
    H.K.

  55. Pat Keating
    Posted Dec 25, 2007 at 9:40 PM | Permalink

    51 Simon

    I have to confess that such sloppy record keeping is also true of the land based readings from those little ‘beehives’ that Anthony Watts is checking out.
    I was a high-school student who had the responsibility of taking wet/dry bulb measurements, rain gaging, etc., during school vacations. I’m afraid that sometimes we forgot or was too lazy, and I have to confess we also made up data, like your AB seamen.

    There was at least one occasion when one of the other students urinated in the rain gage, but I’m sure a correction was applied….

  56. Dan White
    Posted Dec 25, 2007 at 10:37 PM | Permalink

    re 48 and others:
    When I say this isn’t rocket science, I mean that given someone has the time/money and interest to do so, virtually all of the possible scenarios wrt bucket sampling can be reproduced with associated error noted. I understand that there are other variables that cannot be quantified so easily (such as making up numbers, sampling foam instead of water, etc.). However, given that a long period of SST data does exist, it just makes sense to get the most out of it possible before throwing up our hands and concluding that it is useless. If nothing else, such a study would provide a better basis for error bars than a study that intentionally sampled only at night under calm conditions.

    On the other hand, if things are really as bad as Willis says, then forget everything I just said. :) Actually I have to think a lot of this kind of error comes out in the wash with decades worth of data.

  57. Willis Eschenbach
    Posted Dec 26, 2007 at 3:17 AM | Permalink

    Dan W., good to hear from you. I agree that it isn’t rocket science. However, I’m less sanguine than you that the error comes out in the wash. For that to happen, overall the errors would have to be equally above and below the true number, and of equal sizes. Somehow, I suspect that Murphy lives, and there will be an overall bias in the answer.

    That wasn’t my point, though. My point was that we have shifted from measuring sea surface temperature (SST), the part of the ocean that is in contact with the atmosphere, to measuring below the surface temperature (BSST). (As an aside, we’re also increasing the depth of the measurement over time as ships have grown larger and deeper).

    As SST and BSST are two very different things, I do not think that it is possible to combine them into a single record. The surface temperature is the important one, that’s where the action is, where the radiation happens, where the evaporation is going on. The ocean temperature three metres down is a very, very different thing. The ocean surface can be cooking like crazy in the tropical sun, radiating and evaporating at a rate of knots, while three metres down, it’s cold water.

    This is further confused by the addition of satellite based temperature records for the recent periods. The satellite records are once again measuring what the buckets measured … the surface temperature.

    However, I would not conclude anything is useless. I’m just cautious of making systematic adjustments, even small ones, in spliced data that is noisy, full of error, and most importantly, measures two different things.

    w.

    PS – the scientists have chosen the best way to minimize this error. This is to use the SST as a proxy, not for the overall temperature, but for the night-time marine air temperature (NMAT). However, that means for global datasets we’re combining daytime temperatures over land with night-time temperatures over the ocean …

    w.

  58. Geoff Sherrington
    Posted Dec 26, 2007 at 3:18 AM | Permalink

    The way I feel, if nobody has said it to date, I don’t really trust any temperature proxy at all unless the change in temperature exceeds several degrees C on a systematic and comprehensible basis and the proxies are youger than 20 years before present. I want to, because good use can be made of good data.

    I think there are many people who earnestly hope that data reconstruction can be useful, who work hard to show it, but deep down know that they would not gamble their futures on it.

    I suspect that I can state a fatal flaw to just about all temperature reconstructions outside the above limits. Many of us are too polite to use the word “junk”, but sooner or later a lot of past work has to be accepted as such, no matter how honest or earnest the endeavour.

    There are a couple of radioisotope methods, not for temperature, but for age dating, that I can trust, but beyond that there’s a paradise for grant seekers on the back of global warming.

    Now I’ve got that off my back. Anyone else feel this way?

  59. Posted Dec 26, 2007 at 3:29 AM | Permalink

    Dan,

    Yes widen the error bars. How do you determine the error bar on faked data? How do you determine if the data is faked? Find out when the holiday periods were and look for anomalies for land records? How do you handle it for SST records?

    And yes you can test the scenarios and figure out the correction to make. OK. How do you go back through the record and find out which scenario applies? Then of course to get it right you have to go back and get, at least roughly, the distance between the bucket tosser and the sea. Which means you have to go back and look at the ship loading. Records are kept. So now you have to correlate another set of records to the temp record. How do you get a grant for that? And who wants to make their contribution to climate science that way? Will it get you invited to the lecture circuit for that bit of research?

    You also have to go back and try to get some idea of how the job was done on wooden ships. Any contemporary descriptions of that?

    In any case, the records and technique will be a lot better on a research vessel than on a commercial ship. Easier to do a statistical study (per comment #46) and make a declaration.

    For all its problems BSST is probably a more real estimate of “something” (again assuming you know the ship loading) than bucket temps. However, again the numbers can’t be taken in isolation. In heavy seas the depth is not going to be constant.

    Then of course you have the navigation problem. How good was the dead reckoning between navigation fixes? How much time elapsed between the navigation fix and reading the bucket? Not to mention: how good was the navigator?

    Error bars of tenths of a degree (the statistical assumption) don’t seem reasonable.

  60. Larry
    Posted Dec 26, 2007 at 8:19 AM | Permalink

    There was at least one occasion when one of the other students urinated in the rain gage, but I’m sure a correction was applied

    The tinkle factor?

  61. Pat Keating
    Posted Dec 26, 2007 at 8:54 AM | Permalink

    60
    Good one. I was attempting a pun with ‘correction’. He had his backside caned for his trouble.

  62. Kenneth Fritsch
    Posted Dec 26, 2007 at 11:10 AM | Permalink

    Re: #53 and #54

    So Willis and Hans, what should we tell Virginia? Is there or isn’t there and would she accept an “Isn’t there”?

  63. Posted Dec 26, 2007 at 12:42 PM | Permalink

    Another error factor, not included in the above comments:

    As an engineer sailor 1965-1968 (instead of army service) on commercial ships, I had to notice the engine cooling water inlet temperature every shift (6 shifts of 4 hours per day). But I noticed jumps of several degrees C even over minutes, probably due to near surface currents and/or overturning of surface/deep waters…

    How that is incorporated in past and today’s trends is a big question for me…

  64. Willis Eschenbach
    Posted Dec 26, 2007 at 3:40 PM | Permalink

    Ferdinand, thanks for your comment. You say:

    As an engineer sailor 1965-1968 (instead of army service) on commercial ships, I had to notice the engine cooling water inlet temperature every shift (6 shifts of 4 hours per day). But I noticed jumps of several degrees C even over minutes, probably due to near surface currents and/or overturning of surface/deep waters…

    I suspect your explanation of the jumps in temperature are correct. They agree with my experience of the temperature differences between surface water and water only a few feet down. It reinforces my point that buckets and inlets are measuring very different phenomena.

    w.

  65. Dan White
    Posted Dec 26, 2007 at 6:46 PM | Permalink

    Willis #57 and M. Simon #59:
    Thanks for the education, Willis (along with Hans) re SST, BSST, ocean temp gradients etc. It’s all very interesting. My main concern was in not discounting 100 years worth of data as junk without really knowing what the sampling error is. Can I say not to throw the baby out with the bucket water??? Bucket sampling isn’t my thing, so I don’t really know what the maximum expected temp change would be given a “normal” sample. Let’s say you take a hot wooden or metal bucket bucket on the hottest, least windy day of the year and use it to sample the water. By the time all the various thermodynamic effects have worked on the sample, how much does the temp really change relative to the natural SST variability? Other than gross negligence in sampling, I would consider this a probable upper bound for error. If it is large relative to the SST trend, then no need to worry about the gross negligence — the data would probably be suspect enough already.

    I’m probably just stating the obvious here anyway. :)

  66. steven mosher
    Posted Dec 26, 2007 at 7:59 PM | Permalink

    the anomaly method cures all the error issues, biases will all be centered and normal,
    the trend is not impacted by measurement vagaries or changes in method

    Did I just say all that?

  67. Willis Eschenbach
    Posted Dec 27, 2007 at 2:27 AM | Permalink

    Again, Dan and others, the issue to me is not how much a bucket cools or heats up after being hoisted out of the water.

    The issue is that SST and BSST are measuring different things. And not just different things, but things that are not linearly related. We can’t tell whats happening at the surface by looking at what’s happening three metres down.

    The ocean surface is exposed to the atmosphere. It radiates, it absorbs, it convects, it evaporates, it condenses. As a result, it can get much warmer (but not much colder) than the below sea surface waters.

    The temperature swings in the below sea surface waters are much smaller than at the surface. It gains and loses heat much more slowly than the surface.

    Now, remember the purpose of all of this. We are trying to estimate the oceanic equivalent of the terrestrial 2-m above the ground air temperature measurement. Now. To best estimate marine air temperature, should we use

    1) sea surface temperature or

    2) the temperature 3 metres down.

    Of course, that’s a trick question. We should use surface, but we have to use what we’ve got.

    Having to use the below sea surface temperature, however, does not mean we can blithely splice them onto bucket temperatures at one end and onto satellite derived temperatures at the other end. That’s a bridge too far.

    w.

  68. Posted Dec 27, 2007 at 6:16 AM | Permalink

    I was just wondering how convoys handled SST?

    If every ship in a convoy did a measurement (at around the same time) some idea of the real error bars in bucket measurements might be derived. At least we might get an idea of how technique affected the measurements. The other errors would remain.

  69. Posted Dec 27, 2007 at 6:19 AM | Permalink

    Also bucket temps. vs. BSST could be looked at in the period of measurement overlap.

  70. Willem Kernkamp
    Posted Dec 27, 2007 at 7:46 PM | Permalink

    I found the following link on the Met Office Site:

    A comparison of bucket and non-bucket measurements of sea surface temperature, D.E. Parker, Met O 13 Branch Memorandum No. 157, February 1985

    This is unpublished work, but it performs a separation in the SST dataset between “Bucket” measurements nd “Non-Bucket” measurements in the same geographic area. The results are:

    i) Buckets on average read nearly 0.1°C colder. Area-weighted averages are -0.11°C, -0.06°C, -0.08°C, -0.08°C for 60°N to 40°S for Dec-Feb, Mar-May, June-Aug, Sept-Nov, respectively.

    ii) The relative coldness of bucket data is greatest in lower-mid latitude winter.

    iii) Buckets read warmer than non-buckets in upper-mid latitudes, particularly in summer but even in winter to some extent. This result is based on the Northern Hemisphere, there being few data south of 40°S

    iv) In the deep tropics buckets read on average about 0.05°C colder than non-buckets.

    v) Buckets tend to read warmer than non-buckets in the eastern half of the Pacific.

    The outlying points for location and time of year range between plus and minus 0.3 deg C, while the majority of points is in-between -0.2 and 0.0 deg C; i.e. buckets slightly colder.

  71. Willis Eschenbach
    Posted Dec 28, 2007 at 4:18 PM | Permalink

    Willem, thanks very much for the most interesting link. Unfortunately, it suffers from two very large problems:

    1) There are no error estimates, and

    2) There is (AFAICT) no adjustment for autocorrelation.

    As a result, it is not clear how much reliance we can put on the results.

    One thing is clear from the study, however. This is that, as I mentioned, the size and even the sign of the results are dependent on the location, time of day, and time of year. Because of this, I would say that we cannot make any blanket adjustments.

    w.

  72. steven mosher
    Posted Dec 28, 2007 at 6:39 PM | Permalink

    re 71.. also interesting the first 15 years of data ( 1960-1975) are not used…

  73. Evan Jones
    Posted Dec 31, 2007 at 10:23 PM | Permalink

    I have an odd notion that the flattening (even decrease) of the Eenie-Weenie ice Age (1940-1970) may have been due to a mixture of

    a.) sudden and dramatic Full War production (probably the least of it),

    b.) 100 cities either blasted or firebombed,

    c.) A good half-dozen-plus trosophere-puncturing, blast-furnace, wind-sucking firestorms (Dresden, Osaka, Hamburg, Kassel, Tokyo, etc.). Oh, yeah, plus the Atomic Bombs as an afterthought.

    d.) Around a thousand aboveground H-bomb tests (many being mult-megaton), usually over the Pacific Ocean or on top of a nice thick layer of arctic ice up Novaya Zemlya way.

    Do you think that might have had a bit of an effect?

  74. Phil.
    Posted Dec 31, 2007 at 11:29 PM | Permalink

    Re #73

    Add to the that the contrails from thousand+ bomber raids and the associated flak over Europe. Some bomber raids ran into problems because they encountered contrails from the previous day’s raids, lots of high altitude aerosols!

  75. Evan Jones
    Posted Dec 31, 2007 at 11:58 PM | Permalink

    Hmm. Yes. 40,000 feet or more up, some of those raids, IIRC.

    Any other WWII effects?

    Wars are smoky things. Of course it has to get up where it won’t all just rain out in a few days or weeks, which is why the firestorms loom large (they’d punch right up through the troposphere and put up lots and lots of gunk very high up).

    But The UK, USSR, US,(and finally) Germany (in that order) all kicking full-out must have had some sort of continual particulate “brown cloud”. OTOH, there might have been a bit of a “dirty snow” warming effect, too.

    And of course the whole post-WWII airline industry was clicking in, too.

    And the “failsafe” patrols, etc.

    I wonder how it all adds up? Or how one would go about doing so.

  76. _Jim
    Posted Jan 1, 2008 at 12:29 AM | Permalink

    Hmm. Yes. 40,000 feet or more up, some of those raids, IIRC.

    Positive displacement piston-power power plants, super-charged even?

    B-25 service ceiling: Sources agree on numbers around the 24,000 foot figure. The B-17 shows service ceiling figures of 35,000 feet. B-29, a service ceiling of 31,850 feet (but was introduced later in the war).

    the Eenie-Weenie ice Age (1940-1970) may have been due to a mixture of

    a.) sudden and dramatic Full War production (probably the least of it),

    b.) 100 cities either blasted or firebombed,

    c.) A good half-dozen-plus trosophere-puncturing, blast-furnace, wind-sucking firestorms (Dresden, Osaka, Hamburg, Kassel, Tokyo, etc.). Oh, yeah, plus the Atomic Bombs as an afterthought.

    Something else that was happening during/prior to this period: The Dust Bowl, a series of dust storms causing major ecology and agricultural damage to American and Canadian prairie lands from 1933 to 1939 … TONS of soil/particulate matter boosted into the atmosphere …

  77. Evan Jones
    Posted Jan 1, 2008 at 5:56 PM | Permalink

    Hmmm. Well, the dust bowl lasted from ’33 to ’39, and that is generally considered a time of warming. Do we know what effect it had?

  78. Raven
    Posted Jan 1, 2008 at 6:22 PM | Permalink

    Evan Jones says:

    I have an odd notion that the flattening (even decrease) of the Eenie-Weenie ice Age (1940-1970) may have been due to a mixture of…

    If that was the case we would expect to see a sudden drop in temperatures followed by a recovery within a few years (much like the response after a volcano). In the northern hemisphere the temps bottomed out around 1975: http://www.cru.uea.ac.uk/cru/data/temperature/nhshgl.gif

  79. Severian
    Posted Jan 1, 2008 at 7:29 PM | Permalink

    With respect to above ground nuclear testing, I’m surprised I’ve not seen that mentioned often. I think the only real treatment of it I’ve read is on that young lady’s “Paunder the Maunder” page.

    Of course, maybe it’s because it could go either way. Either it lofted enough particulate matter up to have an effect, or maybe it put enough radioactively charged small dust particles in the air to influence cloud formation, ala Svensmark.

    If it’s the latter, that definitely would not be something the AGW alarmists would mention, as it would tend to validate cooling due to radiation influenced cloudiness and increase the magnitude of the solar effect.

  80. Geoff Sherrington
    Posted Jan 1, 2008 at 8:54 PM | Permalink

    The energy output of nuclear weapons was miniscule compared to natural effects – so much so it can be left out of the equations. #76 _Jim has a better approach via the dustbowl.

  81. BarryW
    Posted Jan 1, 2008 at 9:13 PM | Permalink

    Re #80

    But what about particulate matter from the above ground nukes? The temp data referenced in #78 shows peaks about late 50’s early 60’s which may correlate with those tests?

  82. Posted Jan 1, 2008 at 10:24 PM | Permalink

    Re #82

    Try here: http://www.johnstonsarchive.net/nuclear/atest00.html

  83. Mike Davis
    Posted Jan 1, 2008 at 11:12 PM | Permalink

    Steve: They were discussing Above ground testing and I suggested tooking at the history of the Nevada Test site as that was where the tests were conducted. other than first in NM and the ones in South PAC and ther is a lot of info at that site.

  84. Evan Jones
    Posted Jan 2, 2008 at 3:04 PM | Permalink

    Thanks, all. I’ll look into this further when i get the chance.

  85. Willem Kernkamp
    Posted Jan 16, 2008 at 11:54 PM | Permalink

    1. I think the dust bowl/firestorm effects could have been extended if it caused increased albedo in the polar regions.

    2. The number of aircraft actively deployed may have peaked at 15,000 mostly operating at low altitudes. The impact of modern civil aviation could well be far larger. Under the right conditions con trails can turn a clear sky overcast. Adoption of polar routes may have reduced sea ice thickness by reducing winter radiation losses.

    A hypothesis based on 1 and 2 would read that cause 1 initiated a cool period which was terminated by the advent of the jet age (cause 2).

    I think we have gone of topic here, sorry,

    Will

  86. Steve McIntyre
    Posted May 29, 2008 at 11:22 AM | Permalink

    Bump. This thread was a more recent consideration of bucket adjustments than the March 2007 thread, including a useful discussion of what people thought pre-Folland bucket adjustments. I recommend re-reading it.

7 Trackbacks

  1. [...] in the recent record is likely from the 1941 sea surface temperature (SST) adjustment. See here, here, and here for discussion of this adjustment. There is an abrupt jump [...]

  2. [...] December 2007, still long before Thompson et al, I discussed bucket adjustments again, noting the impact of the Folland adjustments on earlier work (referring [...]

  3. [...] Well, folks, the discontinuity may have been overlooked by Hadley Center, CRU, NOAA and NASA and by the stadiums of IPCC peer reviewers, but it wasn’t overlooked here at Climate Audit. The absurdity of Team bucket adjustments had been discussed in two early CA posts (here, here, here ). In March 2007, after publication of Kent et al 2007 showed the prevalence of buckets as late as 1970 (discussed here), I showed in a post entitled The Team and Pearl Harbour that this directly contradicted the Team’s Pearl Harbor adjustment and even showed the impact of a more plausible phasing in of bucket adjustments (see below). The issue was re-visited in Dec 2007 here. [...]

  4. By The Team and Pearl Harbor « Climate Audit on Jul 12, 2011 at 7:46 AM

    [...] (Note – May 2008): see current discussion here and Dec 2007 post here [...]

  5. [...] new paper takes on the well-known buckets-vs-inlets issue (Steve McIntyre also visited the issue several times) related to ship based sea surface temperature measurements and as a result, produces [...]

  6. [...] new paper in GRL takes on the well-known buckets-vs-inlets issue (Steve McIntyre also visited the issue several times) related to ship based sea surface temperature measurements and as a result, produces [...]

  7. […] and Engines, The Team and Pearl Harbor, Bucket Adjustments: More Bilge from RealClimate, Rasmus, the Chevalier and Bucket Adjustments, Did Canada switch from Engine Inlets in 1926 Back to […]

Follow

Get every new post delivered to your Inbox.

Join 3,332 other followers

%d bloggers like this: