Guest Post by Willis Eschenbach [see update at the end]
How much is a “Whole Little”? Well, it’s like a whole lot, only much, much smaller.
There’s a new paper out. As usual, it has a whole bunch of authors, fourteen to be precise. My rule of thumb is that “The quality of research varies inversely with the square of the number of authors” … but I digress.
In this case, they’re mostly Chinese, plus some familiar western hemisphere names like Kevin Trenberth and Michael Mann. Not sure why they’re along for the ride, but it’s all good. The paper is “Record-Setting Ocean Warmth Continued in 2019“. Here’s their money graph:
Now, that would be fairly informative … except that it’s in zettajoules. I renew my protest against the use of zettajoules for displaying or communicating this kind of ocean analysis. It’s not that they are not accurate, they are. It’s that nobody has any idea what that actually means.
So I went to get the data. In the paper, they say:
The data are available at http://159.226.119.60/cheng/ and www.mecp.org.cn/
The second link is in Chinese, and despite translating it, I couldn’t find the data. At the first link, Dr. Cheng’s web page, as far as I could see the data is not there either, but it says:
When I went to that link, it says “Get Data (external)” … which leads to another page, which in turn has a link … back to Dr. Cheng’s web page where I started.
Ouroborous wept.
At that point, I tossed up my hands and decided to just digitize Figure 1 above. The data may certainly be available somewhere between those three sites, but digitizing is incredibly accurate. Figure 2 below is my emulation of their Figure 1. However, I’ve converted it to degrees of temperature change, rather than zettajoules, because it’s a unit we’re all familiar with.
So here’s the hot news. According to these folks, over the last sixty years, the ocean has warmed a little over a tenth of one measly degree … now you can understand why they put it in zettajoules—it’s far more alarming that way.
Next, I’m sorry, but the idea that we can measure the temperature of the top two kilometers of the ocean with an uncertainty of ±0.003°C (three-thousandths of one degree) is simply not believable. For a discussion of their uncertainty calculations, they refer us to an earlier paper here, which says:
When the global ocean is divided into a monthly 1°-by-1° grid, the monthly data coverage is <10% before 1960, <20% from 1960 to 2003, and <30% from 2004 to 2015 (see Materials and Methods for data information and Fig. 1). Coverage is still <30% during the Argo period for a 1°-by-1° grid because the original design specification of the Argo network was to achieve 3°-by-3° near-global coverage (42).
The “Argo” floating buoy system for measuring ocean temperatures was put into operation in 2005. It’s the most widespread and accurate source of ocean temperature data. The floats sleep for nine days down at 1,000 metres, and then wake up, sink down to 2,000 metres, float to the surface measuring temperature and salinity along the way, call home to report the data, and sink back down to 1,000 metres again. The cycle is shown below.
It’s a marvelous system, and there are currently just under 4,000 Argo floats actively measuring the ocean … but the ocean is huge beyond imagining, so despite the Argo floats, more than two-thirds of their global ocean gridded monthly data contains exactly zero observations.
And based on that scanty amount of data, which is missing two-thirds of the monthly temperature data from the surface down, we’re supposed to believe that they can measure the top 651,000,000,000,000,000 cubic metres of the ocean to within ±0.003°C … yeah, that’s totally legit.
Here’s one way to look at it. In general, if we increase the number of measurements we reduce the uncertainty of their average. But the reduction only goes by the square root of the number of measurements. This means that if we want to reduce our uncertainty by one decimal point, say from ±0.03°C to ±0.003°C, we need a hundred times the number of measurements.
And this works in reverse as well. If we have an uncertainty of ±0.003°C and we only want an uncertainty of ±0.03°C, we can use one-hundredth of the number of measurements.
This means that IF we can measure the ocean temperature with an uncertainty of ±0.003°C with 4,000 Argo floats, we could measure it to one decimal less uncertainty, ±0.03°C, with a hundredth of that number, forty floats.
Does anyone think that’s possible? Just forty Argo floats, that’s about one for each area the size of the United States … measuring the ocean temperature of that area down 2,000 metres to within plus or minus three-hundredths of one degree C? Really?
Heck, even with 4,000 floats, that’s one for each area the size of Portugal and two kilometers deep. And call me crazy, but I’m not seeing one thermometer in Portugal telling us a whole lot about the temperature of the entire country … and this is much more complex than just measuring the surface temperature, because the temperature varies vertically in an unpredictable manner as you go down into the ocean.
Perhaps there are some process engineers out there who’ve been tasked with keeping a large water bath at some given temperature, and how many thermometers it would take to measure the average bath temperature to ±0.03°C.
Let me close by saying that with a warming of a bit more than a tenth of a degree Celsius over sixty years it will take about five centuries to warm the upper ocean by one degree C …
Now to be conservative, we could note that the warming seems to have sped up since 1985. But even using that higher recent rate of warming, it will still take three centuries to warm the ocean by one degree Celsius.
So despite the alarmist study title about “RECORD-SETTING OCEAN WARMTH”, we can relax. Thermageddon isn’t around the corner.
Finally, to return to the theme of a “whole little”, I’ve written before about how to me, the amazing thing about the climate is not how much it changes. What has always impressed me is the amazing stability of the climate despite the huge annual energy flows. In this case, the ocean absorbs about 2,015 zettajoules (10^21 joules) of energy per year. That’s an almost unimaginably immense amount of energy—by comparison, the entire human energy usage from all sources, fossil and nuclear and hydro and all the rest, is about 0.6 zettajoules per year …
And of course, the ocean loses almost exactly that much energy as well—if it didn’t, soon we’d either boil or freeze.
So how large is the imbalance between the energy entering and leaving the ocean? Well, over the period of record, the average annual change in ocean heat content per Cheng et al. is 5.5 zettajoules per year … which is about one-third of one percent (0.3%) of the energy entering and leaving the ocean. As I said … amazing stability.
And as a result, the curiously hubristic claim that such a trivial imbalance somehow perforce has to be due to human activities, rather than being a tenth of a percent change due to variations in cloud numbers or timing, or in El Nino frequency, or in the number of thunderstorms, or a tiny change in anything else in the immensely complex climate system, simply cannot be sustained.
Regards to everyone,
w.
h/t to Steve Milloy for giving me a preprint embargoed copy of the paper.
PS: As is my habit, I politely ask that when you comment you quote the exact words you are discussing. Misunderstanding is easy on the intarwebs, but by being specific we can avoid much of it.
[UPDATE] An alert reader in the comments pointed out that the Cheng annual data is here, and the monthly data is here. This, inter alia, is why I do love writing for the web.
This has given me the opportunity to demonstrate how accurate hand digitization actually is. Here’s a scatterplot of the Cheng actual data versus my hand digitized version.
The RMS error of the hand digitized version is 1.13 ZJ, and the mean error is 0.1 ZJ.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Willis, thanks for calingl BS on this paper. Your comments and observations re. the inherent impossibility of measuring what they think they’re measuring are spot on.
Argo floats are nifty, but methinks their utility has been over sold. Not sure what the purpose is other than to provide endless amounts of data to be molested by serial data molesters.
Nick Stokes, my friendly email guy with connections to Australia’s CSIRO, has made many useful and perceptive comments about accuracy here on WUWT.
I used to own a laboratory, one of the first with NATA certification in NATAs formative years. We had expensive thermometers traceable to international reference gear and we had constant temperature water baths. There was and still is, great difficulty in achieving stability better than 0.1 degrees C.
I took several visits to the National Measurement Laboratories to see how it was done with other peoples’ money. They had a constant temperature room that could be adjusted for each person entering the room, maximum 4 folk. They got to 0.01 degrees C.
My neighbour worked elsewhere on accurate measurement and standardisation procedures and we chatted about relevant problems.
Nick, I do not know your personal experience in any detail. However, this matter of true accuracy of Argo floats cries out for a comment from top research bodies. Maybe you have already donned your Lone Ranger mask and are on the trail to a Nick Stokes WUWT comment.
Looking forward to reading it. Cheers, Geoff.
Willis,
Your concerns about this data and its presentation are spot on. However, here is something worth quibbling about.
This is only strictly true if the measurements are independent and identically distributed, or IID. Unless this is so one cannot factor out a constant variance in the propagation of error, which leads to a factor (1/n). It would take a lot of effort to convince me this is true of the Argo data set. This is only one instance of the many ways I hate how climate science uses statistics.
True, Kevin. I didn’t want to open that whole can of worms, in particular since it can only make things MORE uncertain. I settled for looking at the minimum error, since that can’t be argued with.
w.
“In general, if we increase the number of measurements we reduce the uncertainty of their average.”
In all the statistics classes I ever took – as an engineer – it was ALWAYS argued that the reduction in uncertainty is ONLY achieved if the measurements are made using the same equipment, in the same environment (the same piece of water), at virtually the same time. Clearly a practical impossibility with temperature of seawater measurements at ANY depth. Thus, adding and averaging a multitude of readings taken at different places at different times does NOTHING to improve the uncertainty.
I the sailing days, the midshipman dipped a bucket in the ocean and measured it with a thermometer that could perhaps be read to fractions of a degree but how close that was to the ‘real’ temperature was probably not much better than 2 deg.
It took me about 300 million nanoseconds to read this article. Does that make me an extremely slow reader? — according to fourteen unnamed authors, YES.
Yes, but the large number of separate readings taken by each of your Mark 1 eyeballs means that you read the article with an extreme degree of precision and accuracy!
Yay for you!
w. ==> I quite agree that the amazing thing about Earth’s climate is its long term stability. The stability of the Climate System [the whole shooting match taken all together] is, in my opinion which is shared by a few others, due to the stability inherent in chaotic non-linear dynamical systems [ see Chaos Theory]. See my much earlier essay “Chaos & Climate – Part 2: Chaos = Stability” .
Of course, the Earth climate also exhibits a two-pole “strange attractor-like” character, shifting between Ice Ages and Interglacials.
The claim to any knowledge about the “average temperature” or “heat content” of the Earth’s oceans [taken as a whole] is silly buggers scientific hubris writ large. The zigs and zags in the early parts of the paper’s heat content graph are “proof” that the metric is non-scientific and does not trepresent any kind of physical, real-world, reality.
Short and sweet and well said, Kip!
I wonder if the 0.03 degree uncertainty is more related to the 0.02 degree resolution of the argo instrumentation.
Since argo sensors began to be deployed in 2000, what was the source of data from pre-2000 measurements and what assurance is there that those measurements are accurate?
John, pre-Argo we had scientific expeditions using Nansen bottles. Note that they say that pre-1960, less than 10% of their monthly gridcells had any data at all …
w.
Have there been any changes to the design of the ARGO probes over time?
Yes, Mark, they have been modified and improved over the years…and also made and programmed to go deeper.
Initially they only went to 1000 meters, for one thing.
In the just below this one are links to the ARGO website and there is a lot of info there and at various other sources that can be found with a web search.
I am sure one of the improvements was giving them better batteries.
When deployment first began around 2001-2002 or so, and floats were gradually added after that…lithium ion batteries were not nearly as good as the best ones available today, IIRC.
BTW the way everybody…just in case anyone is unaware of it…the ARGO buoy project was not even conceived of until the late 1990’s (1999 to be exact), and the first float deployed several years after that.
The number of buoys deployed only reached what was deemed to be an operationally meaningful(3000 floats were deployed as of 2007) number of units around 2009…IIRC, and for many of the years they have operated, they only went down to 1000 meters, not the 2000 meters they were only recently reprogrammed to dive to.
In 2012, the one millionth measurement was taken…so if one assumes 4000 floats, that would be 250 measurements as of 2012 per float…each of which measures some 90,000 square kilometers of ocean, and only does so every days tens (36 measurements per year) at best.
One might wonder where all the rest of the data came from?
What about before the first float was launched in the early part of the 2000’s?
How about between then and when there was enough to be considered even marginally operational in 2007?
What is going on with mixing up numbers from when we used to only measure the surface with buckets and ship intakes at random places and intervals, with measurements taken prior to 2009 when the ARGO buoys only went to 1000 meters, and then since then when they were gradually reprogrammed to go down to 2000 meters?
The truth is, all of this information (and it is a lot of information, do not get me wrong) is being reported as if everything is known to chiseled-in-stone certainty, exactly as reported in the papers and relayed in graphs and such.
It aint!
To be scientific, information must be reported as measured, and all uncertainties and shortcomings revealed and accounted for…at a bare minimum. Even then, conclusions and measured results can still easily be wrong.
But without meeting those bare minimum standards, the results can in no way be considered scientific.
It barely qualifies as informed speculation.
Some more random bits of info and the sources of what I am opining on here:
– As of today, January 14th of 2020, the official ARGO site says they deploy 800 new units per year, and there are 3,858 in service at present. Hmm…that sounds like even the huge amount of coverage per unit reported is overstated.
– The official ARGO site reports that the accuracy (the word they use…wrongly) of the temperatures reported is + or – 0.002° C, as quoted here from the FAQ page:
“How accurate is the Argo data?
The temperatures in the Argo profiles are accurate to ± 0.002°C and pressures are accurate to ± 2.4dbar. For salinity,there are two answers. The data delivered in real time are sometimes affected by sensor drift. For many floats this drift is small, and the uncorrected salinities are accurate to ± .01 psu. At a later stage, salinities are corrected by expert examination, comparing older floats with newly deployed instruments and with ship-based data. Corrections are made both for identified sensor drift and for a thermal lag error, which can result when the float ascends through a region of strong temperature gradients”
– Each float lasts for about 5-6 years, as they report, and other info on their site puts the actual number of units gathering data as 3000 at any given time, gathering about 100,000 measurements every year. 4000 units with one reading every ten days would give far more…144,000 readings…so…yeah. (Also from the FAQ page)
– There are large gaps in the spacing of the units, and entire regions with none, and none of them are near coastlines, and none in the part of the ocean with ice part of the year. Ditto for the entire region between southeast Asia, Sumatra, and the Philippines…clear north to Japan.
I could go on all day with criticisms, all from their own source page…but I gotta sometimes.
ARGO site and page with current map:
http://www.argo.ucsd.edu/About_Argo.html
FAQ page:
http://www.argo.ucsd.edu/FAQ.html
Lots more info and a bunch of references here:
https://en.wikipedia.org/wiki/Argo_(oceanography)
Willis
I have been following your thermostat theory, and offer the following for your consideration.
IMO your theory is stage two of the thermostat. The first consideration should be – what percentage of the gross energy presented at the ocean / atmospheric interface is actually transported away. Therefor the first stage is the release capacity into the atmosphere.
Considerations could include
1 – wind speeds have reduced by about 15% over the modern warming period.
2 – Tropical cyclones have decreased over the same period due to such things as a weaker Arctic
Your charts identify a significant increase at 26C surface temperature. But what percentage of the energy presented at the surface at that temperature and higher, is actually transported away, given that extremely high relative saturation exist at the ocean / atmosphere interface.
Why do Tropical Cycles exist –
They exist to transport area’s of very high humidity away from areas of high thermal release as the two natural transports of vertical and horizontal are insufficient to accommodate. They step in where the primary mechanisms of transport lack capacity.
What do Tropical Cyclone do –
They transport energy from the tropics both vertically and horizontally.
This in turn allows retention of ocean heat for mixing then raising the average however small. The release of which occurs on much longer time scales.
Ocean heat content increase is not the outcome of CO2 etc, it simply can’t escape during certain climate states due to lack of transport capacity.
With regards
Martin
Thanks, Martin, interesting ideas. I’ll need to ponder on that.
w.
Climate science is the only field in which you can take one temperature measurement in one place, then use a second thermometer to take a reading 100 miles away, and then claim that the existence of the second measurement makes both measurements more accurate.
Yes indeed Mark.
Anyone using statistical techniques to improve the reliability of measurements needs to know this.
This method is only considered to be valid if the measurements were each a separate measurement of the same thing!
Measuring different parcels of water with different instruments can never increase the precision and accuracy of the averaged result.
The water temp is different in ever location and depth.
The temperature in the same location and depth is different at different times.
Everything is always changing, and yet they use techniques that are only valid in a particular set of circumstances and conditions as if it was a general property of measuring things!
And that is only one of the many ways what they are doing does not stand up to even mild scrutiny.
Nicholas McGinley January 14, 2020 at 6:35 pm
Mmm … not true.
Consider a swimming pool. You want to know the average temperature of the water. Which will give you a more accurate answer:
• One thermometer in the middle of the pool.
• A dozen thermometers scattered around the pool.
Obviously, the second one is better, despite the fact that no two of them are measuring the same parcel of water.
See my post on “The Limits Of Uncertainty” for further discussion of this important question.
w.
I think two different things are being talked about here.
You can not increase the precision, nor the accuracy of one thermometer’s measurements by using measurements from a different one in the group of twelve. You can not adjust the reading from one thermometer by the reading of another thermometer in a different location.
If you make multiple, independent measurements of the same thing and you are assured that the “errors” are random, i.e. you have a normal distribution of “true value + errors”, then the mean of the readings will provide a “true value”. Please note it may not be accurate, nor will it have better precision than the actual measurements.
Just in case, the Central Limit Theory DOES NOT allow one to increase the precision of measurements.
No matter how many readings taken, you can never improve your uncertainty beyond the limits of your thermometers.
If you managed to measure every single molecule of water in a pool, with thermometers that are accurate to 0.1C, you will know the temperature of the whole pool, with an accuracy of 0.1C. As you reduce the total number of thermometers you ADD uncertainty as you increase the amount of water that isn’t measured.
The accuracy of individual probes is the base for your uncertainty. You can only go up from there, you can never go down.
If you had 100 probes measuring the same molecule of water at the same time, then you could use your equation to calculate the reduction in uncertainty.
However, if you take 100 probes to measure 100 molecules of water, then your equation does not apply.
There are two types of uncertainty.
There is uncertainty in the accuracy of the reading of an individual thermometer.
There is uncertainty in whether the readings taken, regardless of how many, accurately reflect the actual temperature of the entire pool.
Adding more thermometers can reduce the second uncertainty, it can never reduce the first uncertainty.
Mark,
Adding more thermometers will only improve a result under certain conditions.
For one thing, they must all be accurate and precise, that is, have sufficient resolution and be properly calibrated…and then they have to be read by a competent observer.
IOW…if all of the thermometers are mis-calibrated, it will not matter who reads them or how many one has…the true temp of the pool will not be measured.
I made an unstated assumption that all the thermometers were identical and read identically.
Willis, In your swimming pool example, which I recall from the last time we discussed this several years ago…are you assuming the pool has a uniform temperature from top to bottom and that this is known to be true?
There are several separate things being asserted and discussed here, and conflating them all into one thing, in my opinion, is muddling the various issues.
How about if we make the swimming pool more like the ocean by making it a really big one, Olympic sized…50 meters long. And at one end someone is dumping truckloads of ice into it, and at the other end giant heaters are heating it, and at various places in between, cold air and hot air are being blown over the surface.
So no one knows what the actual average of the pool is.
And the heaters are being turned off and cranked to high over a period of years, randomly, and the trucks full of ice are of unknown size and temperature and frequency…but ongoing at various times, also over many years.
Ten thermometers will give one more information about what the average temp might be at a given instant, if they are all taken at once.
But suppose they are floating around randomly, and each one, on a different schedule, gives a reading every ten days of the top part of the pool only. Also instead of a regular pool it is a pool with steps and ledges of random shapes and sizes and depths…but none of the thermometers is in these shallower parts, and none of them can go where the ice is being dumped…ever.
So…will having ten instead of one give more information?
Of course.
Will ten readings on ten separate days let one determine the accuracy and precision of the measurement at other places and other days with a different instrument?
Can these readings by many instruments at many places but specifically not at certain types of other places, over many years, be used to determine more accurately the total heat content of the pool at any given time, let alone all the time…and how it is changing over time?
I am not disagreeing with you, I am saying that you have not delineated the question about the pool clearly enough for an answer that is, IMO, meaningful.
A swimming pool in a backyard is known to be roughly the same temp from one end to the other and top to bottom.
And one might assume that the ten thermometers would logically be read at the same instant in time…or at least close to that. But one on a cloudy day after a cold night, one on a day prior to that when it is sunny and had not been cold for months on end, and yet another at the surface while it is pouring rain?
No one knows the “true value” of the heat content of the ocean at an instant in time, so how exactly does one know how much uncertainty resides in a reported value such as a change in ocean heat content over time?
I have been reading and discussing this morass here for years, and I know you have been writing about it a lot longer than that.
I spent a bunch of years in college science classes and in labs learning the proper methodology for measuring things, calculating things, and reporting things based on what is and what is not known.
Then a lifetime of real world experience after that, much of which time I have spent ding my best to understand what we know and how that is different from what me might only think we know.
There are entire textbooks on the subjects of accuracy vs precision, but one can read several Wikipedia articles to get a good overview of the concepts.
Reading about it and keeping it all straight however…that is the tricky part.
I am gonna do something which may be annoying but I think is warranted…quote a page from an authoritative source on the interrelated topics of error, uncertainty, precision, and accuracy:
“All measurements of physical quantities are subject to uncertainties in the measurements. Variability in the results of repeated measurements arises because variables that can affect the measurement result are impossible to hold constant. Even if the “circumstances,” could be precisely controlled, the result would still have an error associated with it. This is because the scale was manufactured with a certain level of quality, it is often difficult to read the scale perfectly, fractional estimations between scale marking may be made and etc. Of course, steps can be taken to limit the amount of uncertainty but it is always there.
In order to interpret data correctly and draw valid conclusions the uncertainty must be indicated and dealt with properly. For the result of a measurement to have clear meaning, the value cannot consist of the measured value alone. An indication of how precise and accurate the result is must also be included. Thus, the result of any physical measurement has two essential components: (1) A numerical value (in a specified system of units) giving the best estimate possible of the quantity measured, and (2) the degree of uncertainty associated with this estimated value. Uncertainty is a parameter characterizing the range of values within which the value of the measurand can be said to lie within a specified level of confidence. For example, a measurement of the width of a table might yield a result such as 95.3 +/- 0.1 cm. This result is basically communicating that the person making the measurement believe the value to be closest to 95.3cm but it could have been 95.2 or 95.4cm. The uncertainty is a quantitative indication of the quality of the result. It gives an answer to the question, “how well does the result represent the value of the quantity being measured?”
The full formal process of determining the uncertainty of a measurement is an extensive process involving identifying all of the major process and environmental variables and evaluating their effect on the measurement. This process is beyond the scope of this material but is detailed in the ISO Guide to the Expression of Uncertainty in Measurement (GUM) and the corresponding American National Standard ANSI/NCSL Z540-2. However, there are measures for estimating uncertainty, such as standard deviation, that are based entirely on the analysis of experimental data when all of the major sources of variability were sampled in the collection of the data set.
The first step in communicating the results of a measurement or group of measurements is to understand the terminology related to measurement quality. It can be confusing, which is partly due to some of the terminology having subtle differences and partly due to the terminology being used wrongly and inconsistently. For example, the term “accuracy” is often used when “trueness” should be used. Using the proper terminology is key to ensuring that results are properly communicated.”
I think we all have trouble making sure our commentary is semantically perfect while discussing these things…because in everyday usage many of the words and phrases are interchangeable.
So…how well do the people writing up the ARGO data do at measuring the true value of the heat content of the ocean?
No one knows, of course.
But we would never be aware of that from reading only what they have to say about what they do and have done.
How many significant figures are appropriate, knowing that it is only correct to report a result in terms of the least accurate data used in the calculation…when large areas of the ocean are not even being sampled?
And the different floats are descending to different depths (I came across this eye-opening tidbit of info on the ARGO site just today)?
So, more quoted text:
“True Value
Since the true value cannot be absolutely determined, in practice an accepted reference value is used. The accepted reference value is usually established by repeatedly measuring some NIST or ISO traceable reference standard. This value is not the reference value that is found published in a reference book. Such reference values are not “right” answers; they are measurements that have errors associated with them as well and may not be totally representative of the specific sample being measured.”
“Accuracy and Error
Accuracy is the closeness of agreement between a measured value and the true value. Error is the difference between a measurement and the true value of the measurand (the quantity being measured). Error does not include mistakes. Values that result from reading the wrong value or making some other mistake should be explained and excluded from the data set. Error is what causes values to differ when a measurement is repeated and none of the results can be preferred over the others. Although it is not possible to completely eliminate error in a measurement, it can be controlled and characterized. Often, more effort goes into determining the error or uncertainty in a measurement than into performing the measurement itself.
The total error is usually a combination of systematic error and random error. Many times results are quoted with two errors. The first error quoted is usually the random error, and the second is the systematic error. If only one error is quoted it is the combined error.
Systematic error tends to shift all measurements in a systematic way so that in the course of a number of measurements the mean value is constantly displaced or varies in a predictable way. The causes may be known or unknown but should always be corrected for when present. For instance, no instrument can ever be calibrated perfectly so when a group of measurements systematically differ from the value of a standard reference specimen, an adjustment in the values should be made. Systematic error can be corrected for only when the “true value” (such as the value assigned to a calibration or reference specimen) is known.
Random error is a component of the total error which, in the course of a number of measurements, varies in an unpredictable way. It is not possible to correct for random error. Random errors can occur for a variety of reasons such as:
Lack of equipment sensitivity. An instrument may not be able to respond to or indicate a change in some quantity that is too small or the observer may not be able to discern the change.
Noise in the measurement. Noise is extraneous disturbances that are unpredictable or random and cannot be completely accounted for.
Imprecise definition. It is difficult to exactly define the dimensions of a object. For example, it is difficult to determine the ends of a crack with measuring its length. Two people may likely pick two different starting and ending points.”
“Precision, Repeatability and Reproducibility
Precision is the closeness of agreement between independent measurements of a quantity under the same conditions. It is a measure of how well a measurement can be made without reference to a theoretical or true value. The number of divisions on the scale of the measuring device generally affects the consistency of repeated measurements and, therefore, the precision. Since precision is not based on a true value there is no bias or systematic error in the value, but instead it depends only on the distribution of random errors. The precision of a measurement is usually indicated by the uncertainty or fractional relative uncertainty of a value.
Repeatability is simply the precision determined under conditions where the same methods and equipment are used by the same operator to make measurements on identical specimens. Reproducibility is simply the precision determined under conditions where the same methods but different equipment are used by different operator to make measurements on identical specimens.”
Right!
Now, which of us can keep all of this in mind…and it is only part of a single technical brief on the subject…while we read and comment on such things?
Who thinks anyone in the world of government funded climate science spends any time concerning themselves with repeatability and reproducibility, let alone the distinction between the two concepts?
Here is a link to the brief I quoted:
https://www.nde-ed.org/GeneralResources/ErrorAnalysis/UncertaintyTerms.htm
Now then, if you are still reading, I just realized I did not specifically answer your question about the pool.
I asked some questions back.
If the pool is not well mixed and the readings simultaneous, the ten thermometers are not reading the same thing, but a different thing…water in another part of the pool.
From the Wikipedia article Precision and Accuracy:
“The precision of a measurement system, related to reproducibility and repeatability, is the degree to which repeated measurements under unchanged conditions show the same results.”
It is impossible to measure using ten different instruments in ten places to say anything about the precision of the average of them, or how that compares to one reading.
Note that “repeatability” and “reproducibility” are distinct and separate concepts and both relate to precision and accuracy.
I am gonna skip the links to each of these articles or this comment will go into moderation. I’ll include them in a separate comment after.
Nicholas –> One point. A lot of folks make the mistake that with random error and a sufficient number of measurements, one can make the assumption that the “true value + random error” develops into a normal distribution. This lets you take the average and use the assumption that the random errors cancel out.
This doesn’t mean three different measurements. This means a lot of measurements. It doesn’t mean measurements of different things at different times combined into a population of data (like temperatures). It means the same thing. with the same device.
This means you must be sure that the random errors are random and form a normal distribution so they cancel out.
Overall, one must be cognizant of uncertainty when combining non-repeatable measurements, i.e., temperature measurements versus measurements of the same thing with the same device. Temperature measurements at different times, different locations, and with different devices combine uncertainty a whole lot differently than multiple measurements of the same thing with the same device.
Hi Willis,
I first want to thank you for your reply, which I neglected to do in my first go at responding.
Then I wanted to answer again after rereading your comment, because I think I replied the first time with what was on my mind at the time I read your comment.
So, you asked:
“Consider a swimming pool. You want to know the average temperature of the water. Which will give you a more accurate answer:
• One thermometer in the middle of the pool.
• A dozen thermometers scattered around the pool.”
I agree that the second choice is “better”, all else being equal.
But how about this choice:
Which is better, measuring a swimming pool in ten places by one person at the same time by walking around the pool and using the same thermometer, or having ten people read ten different thermometers at ten random times over a one week interval?
(I have another question for anyone who would like to consider it: How long would it take to read the Wikipedia article on accuracy and precision, and then read all of the reference material, and then read each of articles for the hyperlinked words within the article, and read it all enough times that you have it all clear in your mind?)
Thanks again for the response, Willis.
P.S.
Have you reread the comment section of the article you linked to?
I used to post under the name Menicholas back then, when I was working for a private company and had to worry about getting fired for being a d-word guy.
Hi again Willis,
I am glad you linked to that article, “The Limits Of Uncertainty”, for several reasons, and one of them is because I never got a chance to clear up something regarding a question I asked you, and you answered, here in this comment:
https://wattsupwiththat.com/2018/12/20/the-limits-of-uncertainty/#comment-2563304
You missed that I was quoting that guy Brian!
I never said that, he did.
It was in his first paragraph.
He said all sorts of stuff that made no sense, and several things that were flat out wrong, and I just wanted to make sure I was not the one who was not thinking correctly that night.
You thought I believed that, and I never got a chance to clear it up…and I hate it when that happens, so…
That was a particularly fun discussion, for me anyway.
Thanks for clarifying that, Nicholas, appreciated. And I agree with you that getting tight uncertainty bounds on even a swimming pool is tough, much less the real ocean.
Regards,
w.
You are partially wrong there Willy. More measurements will very likely increase accuracy in your example, but it will not increase as the root of the number of measurements. This only applies to the decrease of random measurement errors of independent measurements of the same quantity.
In all other cases the increase will be less, often much less.
If the ocean is heating up, then one should see an increase in water/water vapor circulation. One would likely measure this as rainfall – I am not sure how cloud cover would correlate. So unless average rainfall has increased to match this additional heat, I would remain highly skeptical of their study.
The problem is, of course, how does one come up with a worldwide average rainfall accurate to within 0.003%? One doesn’t, so their study is safely tucked away from being disproved (at least through this route).
I think I looked up the accuracy of the Argo temperature data once before… and there is no way it can provide an accuracy of 0.003%. If I remember right, they use salinity as a proxy for temperature? Or maybe to correct the temperature measurement…can’t remember.
In any case, the Argo floats do not work under ice nor where the ocean is shallow – they require 2000m depth. This means even if you have a lot of floats, you will not measure a significant amount of the ocean area. The floats are “free ranging” and so one cannot expect them to keep a regular dispersal – there will be clumps and voids over time
Yup!
here is the map, supposedly updated in real time.
There are huge voids and dense clumps.
Large areas regions have zero floats.
Look at the area north of Australia, all the way up to Japan.
Nothing!
Look at the area West of Japan.
Jammed with floats.
There are numerous dense clumps and many areas, some nearby to these clumps, that have none.
And it can be seen that the Arctic has few…although it deos appear there are some under the ice north of the Bering Straits. I am thinking it may be hard to get a reading from those ones!
Arabian Sea…jammed up with them.
http://www.argo.ucsd.edu/About_Argo.html
Actually there are probably more floats in the arctic, but only those that happen to surface in a polynya manage to transmit data.
The “jam” west of Japan is due to a separate Japanese research program, parallell to ARGO proper.
Unfortunately this does not extent to the Sea of Okhotsk, north of Japan, which is completely unsampled.
And unfortunately this uneven sampling is not random. There are huge areas that have never been sampled:
Yep. 1.5mm pa increase in rainfall over the last 60 years, as the planet increases its cooling cycle.
Thanks Willis from a climate layman.
“….warmed a little over a tenth of one measly degree.”
But a long writeup to say “meh”. (smile)
I am not a scientist but an expert in brute force logistics with a scientific mind. I understand data and statistics and appreciate the power of good analysis. BUT….I’ve had to listen to a host of “experts” expounding on suggested improvements that will not make a hill of beans. Their suggestions, assertions and supporting arguments typically fail when I ask penetrating questions about terms of references, assumptions, data, analytical methods…. and what the hell is the marginal improvement, the necessary investment, and the payoff. I suspect it is the same in all endeavors, including as I have seen, in my admittedly imperfect understanding of climate.
Hi Willis,
Great work, and very illuminating to this retired IT guy.
I understand that seawater contains lots of dissolved CO2, and releases it to the atmosphere as the water temperature rises (and vice-versa).
Can you calculate (or even estimate) how much CO2 may have been released by the oceans if the claimed temperature rise had in fact happened, and of course how that compares with claimed’man-made’ CO2 emissions over the same period?
Cheers from smoky Oz.
But if oceans emit CO2 when they get warmer how can they in the same moment absorb all that man made CO2 ?
https://www.businessinsider.es/oceans-absorb-carbon-emissions-climate-change-2018-10?r=US&IR=T
And how can we have Ocean Acidification in a warmer CO2 outgassing ocean?
https://www.ucsusa.org/resources/co2-and-ocean-acidification
BoyfromTottenham January 14, 2020 at 4:25 pm
I understand that seawater contains lots of dissolved CO2, and releases it to the atmosphere as the water temperature rises (and vice-versa).
Only if there is no other source of CO2 in the atmosphere (Henry’s Law)
Can you calculate (or even estimate) how much CO2 may have been released by the oceans if the claimed temperature rise had in fact happened, and of course how that compares with claimed ’man-made’ CO2 emissions over the same period?
As I recall an increase in SST of 1ºC leads to an increase in pCO2 of 16ppm, since 1960 the pCO2 has increased by ~100ppm.
Sounds like a scientific audit of NOAAs climate data is needed
I searched the PDF of the study for the words “solar” and “sun” but found none, they did not even bother to say “Pay no attention to that bright object in the sky”.
Off topic but of great interest. James Delingpole at Breitbart.
Delingpole: Greta Thunberg’s Dad Writes Her Facebook Posts
“Greta Thunberg doesn’t write her own Facebook posts. They are largely written for her by grown-up environmental activists including her father Svante Thunberg and an Indian delegate to the U.N. Climate Secretariat called Adarsh Pratap.
The truth emerged as a result of a Facebook glitch revealed by Wired. A bug made it briefly possible to see who was really running the accounts of celebrity puppets like Greta.”
https://www.breitbart.com/politics/2020/01/14/greta-thunbergs-dad-writes-her-facebook-posts/
Who’da thunk it?
This is like week old news now. You’re at least the fourth person to post it OT to various threads. I don’t really find it a big deal.
No, anybody with a functional forebrain must have realized this long ago.
“Perhaps there are some process engineers out there who’ve been tasked with keeping a large water bath at some given temperature, and how many thermometers it would take to measure the average bath temperature to ±0.03°C.” Been there, done that with relatively small baths of 30 liters. This is quite challenging and expensive. The platinum resistance thermometers and associated electronics (fancy ohm-meters) adequate to do this job are about $3500 per set. Then you’ll need one standard platinum resistance thermometer (SPRT) to check all the others. It’s $4000 to get a good one and another $4000 to get a top metrology lab to calibrate it using fixed point standards. Then a $5000+ ohmmeter to read it. The idea that they have this kind of precision and accuracy is laughable.
Willis – to test their precision, you could use the densest grid of Argos floats, calculate the heat content and temperature of the ocean in that grid, drop 99% of them, and recalculate the temperature. It should not vary by more than ±0.03 K from the denser grid.
Fantastic article,
I see it mentions _<ohc anomaly on the y axis,
why does everyone use 'anomaly'
I refer to the super 'Philosophical Investigations' videos
https://youtu.be/S50_juP5S5U
using just the real data points would make the increase even les dramatic!
Willis,
For the sanity of the world, thank you for another common sense zinger.
I’ve often wondered how the argo bouys will end up being distributed over time
GIven your life experiences I’m sure you have also sat on the banks of a stream or river and watched the flotsam collects in eddies and stagnant points.
Now imagine the same situation for the Argo bouys and what It might mean for their data , and yes, that is a challenge to your inquisitive nature.
And the end result is very far from evenly distributed, or even random:
“PS: As is my habit, I politely ask that when you comment you quote the exact words you are discussing. Misunderstanding is easy on the intarwebs, but by being specific we can avoid much of it.”
“Next, I’m sorry, but the idea that we can measure the temperature of the top two kilometers of the ocean with an uncertainty of ±0.003°C (three-thousandths of one degree) is simply not believable.
https://yourlogicalfallacyis.com/personal-incredulity
Disbelief can arise as a result of knowledge, Steve. As in the case of disbelieving that the uncertainty global average ocean temperature is ±0.003 C.
Steve, first, if you go to your doctor and tell him you think you have copronemia, and he says “Based on my experience, I find it unbelievable that you have copronemia”, do you bust him for personal incredulity? Probably not. Experience is sometimes the finest guide that we have to the truth of some claim.
Next, if I tell you “I can high jump 4.6 metres”, you’ll tell me you don’t believe me for one minute. Why? Personal incredulity. Is that a fallacy? Hell, no. It’s good judgment based on your experience.
Me, I have what I call a “bad number detector”. When it starts ringing, I pay attention. I often have no idea why it’s ringing, but I trust it.
Why do I trust it? Because with very few exceptions, it has turned out to be right in the long run. How do you think I can so quickly identify flaws in published work? I know where to look because I trust my bad number detector.
However, I don’t depend just on that. As you point out, that would be foolish. So next, I went on to show exactly why it is not believable. I used a form of “reductio ad absurdum”, I’m sure you’re familiar with that.
Note that this is a valid form of argument, and at the end of the day, it relies on personal incredulity that something that is totally absurd could be true.
Go figure.
I demonstrated that if their claim is true that you can get an uncertainty of 0.003°C from 4,000 Argo floats, then an uncertainty of 0.03°C could be gotten from forty Argo floats. I assume you know enough statistics to know that that is a true conclusion from their 0.003°C claim.
And if you believe that absurd conclusion, then you desperately need to get your bad number detector checked and re-calibrated.
Finally, Pat Frank pointed out that Argo measurements don’t agree with in-situ measurements …
Error of plus or minus six tenths of a degree … doesn’t bode well for 0.003°C …
All the best,
w.
I think Mosher just likes having his behind handed to him, in shreds, by Willis.
You know Willis, the Argo project was designed and dimensioned by very smart guys, professional experienced oceanographers with thorough training and experience in math, physics , and oceanography.
I works fine, as it was planned to do. There were pressure sensor problems the first few years, but since 2007 (when the Argo array reached target deployment) everything is OK
These guys are also much smarter with data than you. ( They don’t whine and claim “we can’t do this and we can’t do that”) They remove all “known” variance that stems from season, location and depth, which greatly reduces the uncertainty about large scale temperature or heat content changes.
You know, Olof, when you can point out some actual scientific error I made instead of writing a meaningless scientific hagiography of the Argo designers and bitching about how dumb I am by comparison with them, come on back and we’ll talk about it.
In the meantime, you might do well to ponder the comment by one of our most brilliant scientific minds, Richard Feynman, who said:
Wake up and smell the coffee. Recent studies have shown that depending on the field, up to half of peer-reviewed papers in the scientific journals can’t even be replicated … and the amount of crap scientific claims in climate is stunning.
Finally, the folks writing this paper are NOT the designers of the Argo system you refer to, they’re nowhere to be seen. These authors include the noted fabulist Michael Man, inventor of the bogus Hockey Stick and data-hiding Climategate unindicted co-conspirator … smarter than me? He’s not as smart as a bag of ball bearings.
Best regards,
w.
Well, You say that the 0.003 C uncertainty for an annual global average is somehow ridiculous, which should mean that the difference between 2019 and 2018 (~0.004 C, or 25 zettajoules) isn’t statistically significant.
Try to prove the alleged statistical insignificance with a simple nonparametric approach: Compare 2019 vs 2018 month by month, data here:
http://159.226.119.60/cheng/images_files/OHC2000m_monthly_timeseries.txt
What is the outcome of the 12 comparisons? Oops, 12 out of 12 indicate that 2019 is warmer. That is very significant according to Chi-square, binomial tests, etc
Thanks, Olof. There are two issues with that.
The first is that the Cheng data is the most autocorrelated dataset I’ve ever seen. It has a Hurst Exponent of 0.97. This means that you can’t use normal statistical tests on the series. They assume an IID distribution of the data, and this is far, far from IID.
The second issue is that your nonparametric test result would be the same if the uncertainty were twice the claimed amount or half the claimed amount, or if the overall trend were twice or half that of the data.
So you can’t use your test to say anything more than that in general the ocean is warming … but then we knew that …
My best regards to you,
w.
First, sorry for the 0.004 C, the difference between 2018 and 2019 is more like 0.010 C (I think my memory switched the conversion figures from 260 to 620)
I also found that IAP has a depth averaged temperature dataset so conversion between OHC and temperature is not necessary
http://159.226.119.60/cheng/images_files/Temperature0_2000m_monthly_timeseries.txt
http://159.226.119.60/cheng/images_files/Temperature0_2000m_annual_timeseries.txt
Anyway, I don’t think oceans warm by autocorrelation, but rather by physics (heating). Actual temperatures, with a pronounced seasonal signal, are of course autocorrelated. Anomalies does not always remove the seasonal signal, because seasons may have drifted from that of the base period. I think this is true for OHC etc, where the seasonal variation in the southern hemisphere has become more prominent in the recent 10-15 years, compared to the base period.
Hence, the statistically most powerful way to compare years, is to do it pairwise, for example rather a pairwise t-test than the normal t-test.
Regarding the IAP dataset, I believe that it is a little bit special, more like a reanalysis. It’s a observational dataset infilled by CMIP5 model patterns. I don’t know how this affect autocorrelation, but IAP diverge from other datasets during the Argo era when oceans are well sampled.
The problem is that they have no training in metrology, laboratory science, statistics, trending/forecasting, and quality control. It’s not a matter of smart, it is a matter of ignorance. I have seen PhD’s divide numbers with one decimal point and simply copy down the calculator answer with 9 decimal places. Ok for counting numbers, but not for physical measurements.
Willis has done a great job using logic. It doesn’t agree with the paper because the paper is based on Mannstistics – a new realm of mathematical discovery that is difficult for traditionally educated people to understand but which is a very powerful in realizing a new understanding of how the universe operates. Mannstictics explains why, contrary to modern science, CO2 will bring Armageddon at 4:45 June 17, 2030. Only socialists and barely functional academics will survive.
Not sure averaging helps your tolerance. I was always taught that the instrument has a default tolerance, that all measurements will have some error based on the instrument. Averaging multiple measurements together will yield a higher accuracy of the measurement but will NOT decrease the tolerance of the measurement. So you make be able to go from 10.2 +/- 0.5 deg C to 10.1855 +/- 0.5 deg C – the accuracy of the average measurement is improved but the tolerance is not.
The precision is improved, but the accuracy (±0.5 C) is not.
You’ve put your finger on the problem of limited instrumental resolution, Shanghai Dan.
That concept is evidently beyond everyone at Berkeley BEST, UKMet, UEA Climatic Research Unit, and NASA GISS. But every freshman undergraduate in Physics, Chemistry, and Engineering is expected to come to grips with it, and does so.
Actually averaging multiple measurements will not result in a higher precision, i.e. more decimal places. This is what significant digits is all about.
Averaging will provide a “true value” (actually best guess or estimate), without random measuring error if the errors are random and enough measurements are taken of the same thing. You can’t say it provides better accuracy because that is systemic and ALL measurements will off by the systemic accuracy error value.
Tolerance is more generally used as an allowed variation in a product. Tolerance can be affected by a number of measuring uncertainties, both systemic and random.
Willis,
Reading Cheng et al 2020 and your excellent critique of it took me back to the Wong-Fielding ‘three questions’ in Australia of June 2009.
This unique exchange of Questions and Answers between Senator Fielding’s four Scientists, Robert Carter, Stewart Franks,William Kinninmonth and David Evans with Chief Minister Penny Wong and Chief Scientist Penny Sackett, Will Steffen and others was the first occasion to my knowledge when air temperature measurements were essentially discarded in favour of OHC measurements in considerations of global warming.
See http://members.iinet.au/~g lrmc/2009%2008-10%20Fielding%2ODDR%20v.2%20on%20Wong-Steffen%20.pdf
See also David Evans’ post on Jo Nova of his personal views of the meeting.
http://joannenova.com.au/2009/06/the-wong-fielding-meeting-on-global-warming
Now look at the comments on the Argo buoys and the lack of warming shown.
Ever since I have been intrigued to learn what is the actual warming in degrees C shown by the Argo buoys since 2003-04 but like you I ran into zetajoules and such at Argo.net
Trying to get the answer at say NASA.Giss has been equally fruitless.
Recently some climate scientists have claimed Argo readings have swung from negative to positive.
Thanks again for your expose.
February 25, 2013 Your old comment.
“to convert the change in zeta-joules to the corresponding change in degrees C. The first number I need is the volume of the top 700 metres of the ocean. WE has a spreadsheet for this. Interpolated, it says 237,029,703 cubic kilometres. multiply that by 62/60 to adjust for the density of salt vs. fresh water, and multiply by 10^9 to convert to tonnes. multiply that by 4.186 mega-joules per tonne per degree C. it takes about a thousand zeta-joules to raise the upper ocean temperature by 1°C. ”
–
That I believe was for the first 700 meters.
–
I guess you have done similar work here and obviously the 2000 meters requires 2 x more energy than 700 meters so 3000 zeta joules would be needed raise the upper 2000 M by about 1°C.
–
I jut thought that having these figures out for the top 700 meters and 2000 meters makes your explanation clearer when we are trying to convert zeta joules to degrees C.
–
The point should be made that the heat is regulated by the whole of the ocean so there may be a few zetajoules lower down that they missed in this study
“… obviously the 2000 meters requires 2 x more energy than 700 meters…”
Is that obvious?
The radius gets smaller with depth. So the same linear depth as from the surface, at depth, includes less water.
The first 700 m of the ocean contains about 3.2E8 cubic km. The first 2000 m contains about 9.2E8 cubic km.
The 1300 m difference contains about 6E8 cubic km, and so requires about twice the energy of the first 700 m.
Thank you Pat.
I did not mean to imply that I dispute the assertion, only questioning how obvious it is, particularly to anyone who has not had a close look at the numbers for the volumes of the various slices of ocean depth.
I have not had a careful look at them myself, but just from a general knowledge of ocean bathymetry it is readily apparent that much of the ocean is not very deep, and the deeper one goes, the smaller the volume of water in each, for example, 1000 meter layer is.
Descending downwards, first one leaves behind all of the areas that are shallow banks, such as around the Bahamas, Southeast Asia, and around Great Britain, to name a few. Then one leaves behind the continental slopes, shrinking what is left of the ocean basins still further.
Islands and small land areas are all wider at the bases than at the surface, as well.
And before one gets to the bottom of the continental shelves, there are various features protruding up from the ocean bottom…seamounts, and large areas of ridges, of which the spreading center ridges are the highest and the widest.
Below about 6000 meters, only the trenches remain.
I am not so sure the radius of the planet is a big factor…it is difficult for me to visualize the scaling of the actual planet compared to the depth of the ocean, but I do know that the radius of Earth at the equator is ~6380 kilometers, while the deepest trenches are about 11 or 12 km deep.
( In the past, I once found myself checking on the assertion I had once heard that, if an exact model of the Earth was scaled to the size of a billiard ball, and one held it in one’s hand, it would feel smoother than an actual brand new polished billiard ball!
One person who has done the calculation found that, to scale, the crust of the Earth is about as thick as a postage stamp on a soccer ball.
I think I concluded that on a two inch billiard ball, the Marianas trench properly scaled would be two one thousandths of an inch deep scratch. A human hair is about two to two and a half times this distance. So I think that would be easily feelable if you have sensitive hands, especially with Mount Everest so close by with a bump as thick as half a sheet of copy paper.
Graphic of this:
http://i.imgur.com/1oLog.png )
Beyond that…I would not want to be nitpicking…but I would be kind of surprised if the actual ratio was in quite so round of numbers as noted by Angtech.
Thank you for the reply Pat…and thank you so much for the link to your article re your 2015 talk in Sicily and those calibration experiment findings!
Head spinning for sure.
For many of us here, I am sure it confirms what we have always suspected, and some of us have had some knowledge of.
I for one had noted all that way back over 20 years ago that a lot of warming seemed to have appeared when the LIG thermometers inside of Stevenson Screens were replaced with the new units such as the MMTS’s. IMO they should have added the new units and kept the old ones in place for a bunch of years before even thinking about using the data the new units collected.
If anyone think Argo can tell us in within a degree 1 C what going on the ocean is fool, you cannot measure the ocean with a bunch of random measure in less than small percentage of the ocean. The majorty of surface measurements of the earth is less that 3% of the earth, the ocean measurements are less. That not science, the reality is a multiple drops of dice might tells us as much.