Briffa embraces tree ring uncertainty in new study, gives 'message to the paleoclimate community'

Alternatively, if a nearby tree falls, then the increased sunlight will cause an increase in growth.
Without knowing the history of a particular tree, all reconstructions are little more than guess work.

“all reconstructions are little more than guess work.”
Isn’t most science little more than guess work? Followed by tedious testing of your guesses? As I noted above, this paper looks like it fits into the testing part of the process.

” I don’t agree that climatology has a general problem with over-certainty.”
You’re the only one.

Chic Bowdrie,

At the outset, I’m going to admit no expertise in uncertainty measurements. But I’m happy to learn.

This is why I like talking to you. I’m not expert either, hence I doubt I have much to teach you. I think I’m much better at pointing out flaws in arguments, and unfortunately that’s what I mostly do in this post.

How certain are you that GISS and Hadcrut accurately measure an average global temperatures? How certain are you that satellites don’t more accurately reflect global temperature based on troposphere measurements?

The short answer is that I have more confidence in the surface record than the satellites. Long answer is long, but contains why I think so.
I have no other way of quantifying this than by appealing to what the data providers themselves have to say about their own products. This is a little difficult because both RSS and UAH quantify uncertainty in terms of the linear trend over their entire recordset, whereas Hadley CRU and GISS only give it in terms of mean monthly or annual uncertainty. Hadley provides an error estimate for every month and year, whereas AFAIK GISS does not — they give us a range from past to present, with the older data being less uncertain than recent data.
Both Hadley and RSS do something nobody else does; they generate multiple realizations of their temperature estimates and then average them to produce an ensemble mean estimate as a final product. Such an ensemble method provides ways of estimating uncertainty and comparing them to each other in an apples to apples way that can’t be done with data from any other provider. Kevin Cowtan took the ensembles from RSS TLT and HADCRUT4 and did such an analysis. He estimated that the RSS TLT trend uncertainty between 1979-2012 was +/-0.015 K/decade, whereas the for HADCRUT4 it’s +/-0.003 K/decade at the 95% confidence level, or about 5 times less uncertain.
I woudn’t say that I’m 5 times more confident in HADCRUT4 than RSS TLT. I can say that since the central estimate for HADCRUT4 lies within the 95% CI of RSS TLT, and that HADCRUT4 shows the higher trend that the RSS TLT central trend estimate is likely lower than it should be with HADCRUT4 only possibly being higher than it should be … instead of the other way ’round.
References for Dr. Cowtan’s analysis begin with my comments on this recent thread here at WUWT: http://wattsupwiththat.com/2016/01/18/monday-mirth-old-reliable/#comment-2122673
My confidence is very low that any atmospheric temperature record tells us much about how much absorbed solar energy is being retained/lost by the system as the oceans are by far the largest heat sink in system. If I had to pick one single indicator already available to us, it would be the upper two kilometers of the oceans:
http://climexp.knmi.nl/data/itemp2000_global.png
Question then becomes how uncertain are those estimates?
Something to keep in mind about me here, I don’t constrain myself to using the single “best” metric from the single “best” data source. No non-trivial scientific finding I’m aware of (think the Standard Model of physics, theory of plate tectonics, etc.) has ever relied on a single line of evidence to establish its claims. Climate science is no different, and relies on consilience of evidence to make stronger conclusions than could be made by appealing to only the “best” one.

How can sparse measurements of the vast ocean accurately predict a warming of 0.2K and how is 90% of it claimed to be man-made with a straight face?

Sparse measurements have been the bane of human knowledge since we developed the ability to actually take them. Obviously such measurements cannot be taken with zero uncertainty, so we developed statistical methods for estimating it. The simplest rule anyone who’s taken basic stats is that standard error of the mean varies as the ratio of the sample standard deviation over the square root of the total number of observations. It gets more complex when we’re making observations across both time and two- or three-dimensional space, but again these are not problems unique to climate. The same difficulties apply to mobile satellites, radiosondes and ocean bouys used for 3D estimates over time as they do to surface-based thermometers which mostly sit in the same spot for years on end which are used for 2D estimates over time.
You’re right to be skeptical, but to be credibly skeptical you need to be consistent in your skepticism. I frankly don’t think you are, because basically you’ve just called into question ANY scientific conclusion based on sparse estimates and/or wonky instrumentation — which all field instruments are to some degree.

Example #2 of climate science over-certainty comes from an older paper by GL Stephens et al. 2013 where they estimate energy imbalance as 0.6 +/- 0.4 W/m2. How do they get that accuracy and/or precision based on numbers for outgoing radiation reported as 100.0 +/- 2 for SW and 239.7 +/- 3.3 for LW? This is to balance the incoming 340.2 +/- 0.1 W/m2 radiation coming in.

What does the paper itself say?

I call that being over certain.

Your opinion is noted. Now tell me why I should hold your opinion in higher esteem than that of domain experts who have actually done the work?
For me the credibility for the research comes from the very fact that they list the identifiable sources of uncertainty, the various assumptions they make to derive their estimates, and allow for possible other sources of uncertainty they haven’t been able to identify. I can scan the list of references and/or read some of them and note that other methods done by other teams arrived at similar conclusions. I can do my own analysis. For instance I get an imbalance of 0.69 W/m2 from 1948-present using the NCEP/NCAR reanalysis data for latent heat, sensible heat, long- and short-wave fluxes at the surface, which is well inside their +/- 0.4 W/m2 uncertainty estimate.
Here you’re also making the implicit argument that 0.4 W/m2 is a teensy number compared to the magnitudes of the fluxes involved, 340.2 +/- 0.1 W/m2 for incoming SW, 100.0 +/- 2 for outgoing SW and 239.7 +/- 3.3 for outgoing LW. I don’t see why that should be such an unreasonable argument on the face of it, why shouldn’t a well-designed modern detector be able to return an accurate reading from 0 to several thousands of W/m2 at a near-constant precision across that entire range? For sake of consistency, are you not also claiming that orbiting microwave sounding units have enough absolute accuracy and precision to derive tropospheric temperature to their stated level of precision? How is it that RSS and UAH uncertainty claims are credible, but Stephens et al. 2013 claims which rely on orbiting instruments are not?
Finally let’s look at something here: 0.6 +/- 0.4 W/m2
Note that the estimated uncertainty is 66% of the central estimate. I don’t know where the point of ridiculously over-certain begins by this somewhat dubious method, but I daresay it’s a smaller number than 2/3rds.

The main reason I think climatology over does certainty comes from IPCC language like 95% certain, etc. and estimates of climate sensitivity which are all over the place vis-a-vis all the claims humans are primarily responsible for global warming without definitive evidence to back them up.

What do the IPCC say they’re 95% certain of?
Isn’t the fact that IPCC-published estimates of climate sensitivity to CO2 range from 1.5-4.5 K/2xCO2 an contra-indicator of over-certainty? Measurement error propagates, right? Models don’t agree with each other. Many climate sensitivity estimates come from paleo evidence, huge error bars, and as this very OP points out, some of which is likely underestimated. Isn’t the it the honest thing for the IPCC to do to publish a range large enough to drive a city bus through sideways if that’s what the sum total of the original research indicates as range of estimates?
Isn’t “definitive” evidence the same as saying evidence with an estimated uncertainty of zero? Isn’t such a small error estimate something you are claiming is ridiculous?

Brandon,
“Hadley provides an error estimate for every month and year, whereas AFAIK GISS does not — they give us a range from past to present, with the older data being less uncertain than recent data.”
Older data being less uncertain than recent data? That says to me there’s something terribly wrong with GISS data. Are there less measurements today? Is the instrumentation less accurate today?
“Kevin Cowtan … estimated that the RSS TLT trend uncertainty between 1979-2012 was +/-0.015 K/decade, whereas the for HADCRUT4 it’s +/-0.003 K/decade at the 95% confidence level, or about 5 times less uncertain.”
Trend uncertainty? Seems to me trend analysis is a function of all the data points, not the accuracy of the individual points. IOW, if Hadcrut measured less extreme values compared to the RSS values relative to the true values, wouldn’t the trend stats show less error for the Hadcrut trend even though the RSS trend was more accurate? The fact that one trend is within the confidence limits of the other says nothing about the certainty of the individual data points. However, this does illustrate how to overdo certainty statements.
Your justification for the validity of sparse measurements is weak. Not all scientific conclusions are based on sparse estimates and/or wonky instrumentation. You can’t go to the moon using sparse measurements and wonky instrumentation. The certainty of a scientific conclusion is inversely proportional to the scarcity of estimates and wonky instrumentation.
“What does the [Stephens et al.] paper itself say?”
How did you calculate your 0.69 W/m2 imbalance? Their incoming radiation was very precise: +/- 0.1 W/m2. But their outgoing radiation is the sum of two amounts with errors summing greater than 10 times the error reported for the total imbalance. How do you get +/- 0.4 from numbers known to only +/- 2 and 3? Based on their numbers, the imbalance could have been -5 W/m2.
“How is it that RSS and UAH uncertainty claims are credible, but Stephens et al. 2013 claims which rely on orbiting instruments are not?”
I did not make any evaluation of RSS/UAH uncertainty claims. Nevertheless, I expect variability associated with a specific location and short time interval to be more precise than the variability associated with data points averaged over wide areas and longer time intervals.
“What do the IPCC say they’re 95% certain of?”
Each IPCC report increased the level of confidence that humans are responsible for more than half of global warming. Yet there has been no refinement on the estimates of CO2 sensitivity and climate models overestimate global temperatures. If sensitivity is low, how can humans be blamed for the warming? That’s over-certainty.
“Isn’t “definitive” evidence the same as saying evidence with an estimated uncertainty of zero?”
No. Definitive evidence is data that shows an incremental increase in CO2 has a certain influence on global temperature and that humans have a specific contribution to that CO2 increase.

Brandon,

Both Hadley and RSS do something nobody else does; they generate multiple realizations of their temperature estimates and then average them to produce an ensemble mean estimate as a final product. Such an ensemble method provides ways of estimating uncertainty and comparing them to each other in an apples to apples way that can’t be done with data from any other provider.

I took a look at your SKS link to the post by Kevin Cowtan. He does a good job of explaining the difference between statistical and structural uncertainty. Unfortunately, in his comparison of the RSS and HadCRUT4 ensembles, he doesn’t control for statistical uncertainty. It looks like RSS data has more statistical uncertainty, or in Cowtan’s terms, more wiggles. Do the wiggles reflect less accurate data? Not necessarily.
I would be more impressed with an analysis that not only controlled for the statistical uncertainty, but used ensembles produced using the same paradigm. Otherwise, the analysis is still comparing the proverbial apples and oranges. Obviously it’s problematic when the target measurements are different, but the criteria Mears et al. used to produce ensembles may have exaggerated the degree of wiggling compared to that used by Morice et al.
An explanation of the ensemble process may have tempered to your debate with richardscourtney on the Monday mirth WUWT post you referred to earlier. Even Dr. Cowtan pointed out in Figure 3 that the HadCRUT4 ensembles omit some sources of uncertainty. It is a bit disingenuous to leave out that last sentence of the caption on the figures you posted.

Brandon,

That wasn’t clear to me at all, perhaps you can explain further [the effect of ensemble production on trend uncertainty].

Mears et al. and Morice et al. produced ensembles representative of a set of data points for a given day of global temperatures. This may or may not be a correct statement as I’m not familiar with this methodology. Presumably they didn’t produce the ensembles specifically for the purpose of comparing surface vs. satellite measurements. Had that been the case, could criteria for producing ensembles have been standardized to prevent the statistical uncertainty from either measurement biasing the trend’s structural uncertainty? Again, I don’t know the methodology so I don’t even know if that makes sense.

Where the 50% attribution line falls on the distribution depends on the shape of the distribution, not just the upper and lower bounds. All that’s required for that 50% line to move is for the bulk of estimates to move away from the tails.

I don’t see what this has to do with the fact that there has been no zeroing in on how much CO2 affects global temperatures or the degree to which humans have to do with it.

Something else to keep in mind; CO2 sensitivity is a function of climate sensitivity to ANY external forcing, natural or not. Error propagates, and any uncertainty in that estimate flows through everything.

I don’t get this either, but it seems to have something to do with uncertainty being a lot more than climate scientists want to admit.

Chic Bowdrie,

Mears et al. and Morice et al. produced ensembles representative of a set of data points for a given day of global temperatures. This may or may not be a correct statement as I’m not familiar with this methodology.

I think it’s probably close enough to true for purposes of discussing multi-decadal trends.

Presumably they didn’t produce the ensembles specifically for the purpose of comparing surface vs. satellite measurements.

Probably also a true enough assumption for our discussion.

Had that been the case, could criteria for producing ensembles have been standardized to prevent the statistical uncertainty from either measurement biasing the trend’s structural uncertainty? Again, I don’t know the methodology so I don’t even know if that makes sense.

It makes sense. Since I don’t have expertise in these methodologies, it’s beyond my present capability to answer it definitively. Way I see it, I have only two objective choices:
1) accept them both as reasonable estimates of uncertainty and consider them comparable
2) reject both of them as reasonable estimates of uncertainty and not make the comparison

I don’t see what this has to do with the fact that there has been no zeroing in on how much CO2 affects global temperatures or the degree to which humans have to do with it.

Some review is in order, my fault for not doing it sooner. In AR4, the IPCC only considered anthropogenic GHG forcings in their attribution statement:
Most of the observed increase in global average temperatures since the mid-20th century is very likely due to the observed increase in anthropogenic greenhouse gas concentrations.
In AR5, the IPCC added “other anthropogenic forcings”:
It is extremely likely that more than half of the observed increase in global average surface temperature from 1951 to 2010 was caused by the anthropogenic increase in greenhouse gas concentrations and other anthropogenic forcings together. The best estimate of the human-induced contribution to warming is similar to the observed warming over this period.
So we have an apples and oranges problem here.

I don’t get this either, but it seems to have something to do with uncertainty being a lot more than climate scientists want to admit.

Odd point to make since I’m getting it directly from them, Knutti and Hegerl (2008), The equilibrium sensitivity of the Earth’s temperature to radiation changes: http://www.image.ucar.edu/idag/Papers/Knutti_nature08.pdf
They even share your lament right in the abstract:
Various observations favour a [CO2] climate sensitivity value of about 3°C, with a likely range of about 2–4.5°C. However, the physics of the response and uncertainties in forcing lead to fundamental difficulties in ruling out higher values. The quest to determine climate sensitivity has now been going on for decades, with disturbingly little progress in narrowing the large uncertainty range.
An excerpt from the first paragraph:
Climate sensitivity cannot be measured directly, but it can be estimated from comprehensive climate models. It can also be estimated from climate change over the twentieth century or from short-term climate variations such as volcanic eruptions, both of which were observed instrumentally, and from climate changes over the Earth’s history that have been reconstructed from palaeoclimatic data. Many model-simulated aspects of climate change scale approximately linearly with climate sensitivity, which is therefore sometimes seen as the ‘magic number’ of a model. This view is too simplistic and misses many important spatial and temporal aspects of climate change. Nevertheless, climate sensitivity is the largest source of uncertainty in projections of climate change beyond a few decades1–3 and is therefore an important diagnostic in climate modelling4,5.

There is a problem with that sentence’s syntax.

Big time. Try: We expect more variability in the troposphere than at the surface. Therefore, it would be reasonable to expect tropospheric measurements to “look” more uncertain due to their higher variability.

Also, who is we and why do you expect the troposphere to vary more than the surface?

We being humanity by way of scientific research. Why is complex, and my non-expertise is such that a short answer citing the primary causes would not be very good. Tropical convection, lapse rate feedback and latent heat transfer are the things I most commonly see mentioned.

Please define what variability you are referring to.

Monthly, seasonal, annual, interannual, decadal … even interdecadal. Long-term trends are expected to be amplified as well.

For example, I would expect a similar diurnal temperature swing at the surface compared to the lower troposphere and both to have a larger swing compared to the mid tropopause at a specific location.

If you say so. I would actually expect air to change temperature more rapidly than ground or bodies of water for any given change in net energy flux. Nighttime and/or wintertime inversions can create exceptions to those rules. Not that understanding daily cycles isn’t important, but we’ve gotten to the point that we’re talking way more about weather than climate now.

Another source of variability is from the number of measurements at a specific location. For the average temperature at a specific location on any given day, I would expect the standard deviation of the measurements to decrease with an increasing number of measurements. Do satellites take more measurements than surface thermometers?

Dunno. It’s also apples and oranges again. The sats have to combine nadir and limb readings (straight down and sideways) to derive the lower troposphere, which means they’re not looking at exactly the same place at the same time. They’re reading microwave emissions from oxygen atoms through the entire thickness of the atmosphere when they do this, so there’s a volume component to consider as well. It’s not very “pinpoint” in spatial terms. And there is some ground-clutter to remove as well.
OTOH, in temporal terms, a single typical weather station gives us only a daily min/max. They don’t move around as a normal course of operation, though we don’t always know when they’ve moved nor when some idiot put one next to an air-conditioning unit or decided to build a parking lot and not tell someone about it.
We can puzzle through these pros and cons together anecdotally, but I for one cannot weight them quantitatively in my head. For that, I prevail on experts to tell me the estimated uncertainties.

Brandon,

Odd point to make since I’m getting it directly from them, Knutti and Hegerl (2008)

It’s clear to me that the whole sensitivity issue is fraught with inconsistencies, bad physics, and specious exaggeration of global warming. If you want to discuss this further, start by reading comments by AlecM, William Astley, and Dr. Happer at this link: https://tallbloke.wordpress.com/2016/01/21/can-a-gaseous-atmosphere-really-radiate-as-black-body/ and this paper: http://www.john-daly.com/forcing/hug-barrett.htm

If you say so. I would actually expect air to change temperature more rapidly than ground or bodies of water for any given change in net energy flux.

This addresses the issue of what variability we’re talking about. As you point out, the daily surface measurement is a high/low average. Right there you’ve introduced a bias. The greater the difference between the high and low (amplitude), the greater the potential for bias. If I interpreted this paper correctly, it confirms my expectations: http://www.arl.noaa.gov/documents/JournalPDFs/SeidelFreeWang.JGR2005.pdf

Not that understanding daily cycles isn’t important, but we’ve gotten to the point that we’re talking way more about weather than climate now.

We’re talking about uncertainty in data measurements and how they are processed. I think it’s relevant, but there’s no need to continue.

We can puzzle through these pros and cons together anecdotally, but I for one cannot weight them quantitatively in my head. For that, I prevail on experts to tell me the estimated uncertainties.

Being a skeptic, I can’t do that. But I know a lot more about what I didn’t know, so thanks for puzzling.

Brandon,

Where I recall you and I leaving the radiative physics question was with the elevator analogy for surface to tropopause convection. My final comment to you suggested that you consider for every up elevator in the system is one going down.

Yes, I remember having trouble answering this question of yours from the Lindzen post:

161 W/m^2 at ground level is the global average for absorbed solar according to Trenberth and Kiehl. 396 W/m^2 is the global average upwelling LW from the surface according to same. Please explain the difference?

The difference is 235 W/m2 which is the baseline emission that is reinforced every day by incoming insolation. If the sun never came up again, the 235 W/m2 at the surface and at the top of the atmosphere would gradually drop until a new equilibrium was established due to cooling of the earth’s core.

When I think of bias in this context, I think of a non-random error which affects the trend of a central estimate.

Correct.

A min/max reading guarantees that the calculated mean and median value are the same on any given day for any given single instrument, which is not a desirable feature.

Why do you say not a desirable feature? The mean calculated from a minimum and maximum is not the same as a mean calculated from an average of temperatures taken every hour over a 24 hour period. Check out this data for Ames, Iowa showing the min/max average is usually greater than the average of the hourly measurements: http://mesonet.agron.iastate.edu/onsite/features/cat.php?day=2009-07-31
With regard to the other variability, monthly, seasonal, etc. I think each has to be considered in the context of the measurements and the experimental objective. Santer et al. paper mentions month-to-month tropical data comparing surface to troposphere. But the surface is thermometers and the troposphere is measured by satellites. So, is the month-to-month variability due to the locations or to the instrument differences? The long term trends show the troposphere warms less than the surface, so I don’t really get what the point of the paper is. It certainly doesn’t indicate a preference for either one of the measurement techniques.

You’re not the only skeptic in this conversation, Chic. And you do apparently trust expert opinion when it confirms your expectations, just as you wrote when you cited Wang (2005) above. In that respect, I don’t think I work any differently than you do.

Forgive me for assuming you support the AGW position, as opposed to a Skeptic like me, rather than just being genuinely skeptical. What defines expert? What if an expert is wrong? I don’t consider scientists, especially climate scientists, experts just because they publish. Therefore I don’t know how expert Seidel, Free, and Wang (2005) are. DJ Seidel was a co-author on the Santer et al. paper which I don’t think much of. But I could understand the Seidel et al. paper well enough to know that, right or wrong, it supported the point that I was making.
As long as you’re pursuing scientific truth, then I agree you don’t work any differently than I do.

Forgive me for assuming you support the AGW position, as opposed to a Skeptic like me, rather than just being genuinely skeptical.

I meant “generally” skeptical.

Brandon,

I strongly suggest that this is the reason you’re having trouble answering that first question.

I answered it, but not well enough apparently. The net loss from the surface is zero. If there was a non-zero net loss, the surface temperature wouldn’t average 288K anymore. The equation is 161 + 235 = 396. The 161 is actually 322/2, because it comes in only during day. The point is, on average, every day, 235 W/m2 is being emitted from the surface in the morning and an additional 322/2 adds to it during the day. Meanwhile at the TOA there is an average 235 W/m2 emitting to space, which provides energy balance. This is heat transfer 101, but Kiehl-Trenberth diagrams obfuscate it by inserting the recycled backradiation nonsense. And yes, it is an accounting problem, but the LWIR cancels on the bottom floors. The bulk of the energy is going up the elevator to the top floor by convection. Then radiation takes over getting the energy out to space.

Oh, to follow up, I did look at both links you provided last post.

I’m impressed. I didn’t read Schmithüsen et al. (2015), because it was peripheral to what I was interested in. Now I’ll have to take a look at it. There is also further discussion of it on tallbloke’s which you may have already followed.

I find nothing in [the Kiehl-Trenberth energy budget diagram] to support the contention: This law specifies that a good absorber is a good emitter and the IPCC supporters have interpreted this to indicate that they should in equal measure emit the terrestrial radiation absorbed by the greenhouse gases.

The alternative is to acknowledge, as you have, that collisions predominate when the atmosphere is dense. Therefore in the lower troposphere, IR active gases absorb more than they emit.

That’s why a daily min/max for any given fixed location is not a desirable feature. The sats don’t even give you that. What is your response?

I’m glad you were aware of that. I wasn’t sure.

We tend to trust statements about things we already believe or expect to be true, no matter who says them. Confirmation bias. I’m not immune to it, nobody is.

The best immunity is simply to seek the truth, scientific or otherwise.

Brandon,

IIRC, I’ve shown you previously by tallying up ALL the values, each of the three levels in the diagram net roughly to zero.

I don’t recall us discussing the Kiehl-Trenberth numbers at all levels, but there is no question they sum to zero at all levels. The point is the numbers are unphysical, because they are on average approximately correct only twice a day. At no time does 161 W/m2 hit the whole surface area of the planet at the same time. When evaluating the budgeting process, you also have to recognize that there was an initial balance that accumulated previously. From that point on there are deposits and withdrawals that average close to zero, but there is never a zero balance. This is crucial to realize and if you can’t fathom it, there is no point in us continuing to discuss the Kiehl-Trenberth diagram. Plus it prevents you from accepting what little role backradiation plays in the atmosphere’s heat transfer process.
Assuming the physical unreality of the energy budget and just working with the math, all levels balance and temperatures will adjust to changes in the long (or short) term average NET flux at ANY level. That’s not an issue for me.

You keep skipping over the middle part.

No, convection handles the middle part.

Emission/absorption is constantly occurring. What happens when a photon jumps off an up elevator on a middle floor and catches a down elevator?

Yes that is the point! For every potential photon (it’s not a photon until it’s emitted) going up or down, there is a 50:50 chance it goes up or down. So there is statistically very little difference in any net radiation going up or down except via the window where surface radiation takes the express to the top and at the top where thin air prevents collisions from absorbing upward bound photons.

More CO2, more net absorbers, more thermalization.

Another key point. CO2, or any other IR active gas, can only absorb the available radiation. Apparently, this amount never exceeds that which can be absorbed relatively close to the surface other than that which goes through the window directly to space. So the degree of thermalization is not affected by increasing CO2 or water vapor as long as they are present in sufficient concentration. What is affected is the rate of thermalization which is where convection comes in. The more radiation, the more absorption, the more thermalization, the more convection.

Exact opposite occurs in the thinner stratosphere closer to space — more radiators closer to the exit, more cooling.

Although it isn’t exactly the opposite phenomenon, I agree that more CO2 in the upper atmosphere should increase cooling, all else being equal. BTW, some of us believe that increasing CO2 should have a net cooling effect because of its contribution to enhancing convective cooling below and radiative cooling above.

I’m asking you if you realize that the sats don’t even give you a daily min/max for any given location?

I don’t know the satellite methodology. What is it and how is it worse than a high/low average taken of an incomplete set of non-homogeneous measurements?

Chic Bowdrie,

I don’t recall us discussing the Kiehl-Trenberth numbers at all levels, but there is no question they sum to zero at all levels.

Probably was directed at someone else on a different thread then, my apologies for my bad memory. Salient point is that they should sum to zero at all levels in a theoretically steady-state equilibrium.

The point is the numbers are unphysical, because they are on average approximately correct only twice a day.

If the intent of the diagram were to represent fluxes for every clock-tick of the diurnal cycle for every cubic meter of the climate system, I would agree with you.

At no time does 161 W/m2 hit the whole surface area of the planet at the same time.

The fact that a globally averaged figure is what’s reported in the budget diagram does not mean that the supporting observations and calculations were done at such a highly-summarized resolution.

When evaluating the budgeting process, you also have to recognize that there was an initial balance that accumulated previously.

We’re never going to be able to trace every energy flux back to the back to the Big Bang. Holocene max temps were higher than the LIA which has some implications for residual heat in the deep oceans. If I know this, Trenberth and Co. surely do, plus more.

From that point on there are deposits and withdrawals that average close to zero, but there is never a zero balance.

Of course not, and nobody I’m aware of who does the actual research and writes the literature says otherwise. “Steady-state equilibrium” in this context does not mean an isothermal system with constant net input/output flux of exactly zero.

This is crucial to realize and if you can’t fathom it, there is no point in us continuing to discuss the Kiehl-Trenberth diagram.

I believe I have fathomed it as I have both read and thought a lot about it. I think what you have said above are justifiable arguments for a healthy amount of uncertainty in the globally averaged estimates. However, in my admittedly lay opinion, I don’t that see your arguments justifiably negate the physicality of those estimates.

Plus it prevents you from accepting what little role backradiation plays in the atmosphere’s heat transfer process.

I think you falsely presume a lack of understanding on my part, which is fine: you can’t possibly know everything I think I understand until I write about it.

Assuming the physical unreality of the energy budget and just working with the math, all levels balance and temperatures will adjust to changes in the long (or short) term average NET flux at ANY level. That’s not an issue for me.

Since you’ve noted above that all levels of the budget cartoon net to zero, doesn’t that at least suggest your assumption of bad physics is incorrect? IOW, how can both of your above statements be simultaneously true?

No, convection handles the middle part.

Convection goes both ways.

Yes that is the point! For every potential photon (it’s not a photon until it’s emitted) going up or down, there is a 50:50 chance it goes up or down.

That’s the 1D model. I’ve mostly been thinking and writing using a 2D model, but on review maybe this argument works better in 1D, so I’ll go with that for now.

“Emission/absorption is constantly occurring. What happens when a photon jumps off an up elevator on a middle floor and catches a down elevator?”
So there is statistically very little difference in any net radiation going up or down except via the window where surface radiation takes the express to the top and at the top where thin air prevents collisions from absorbing upward bound photons.

The only things riding the elevator all the way to the tropopause are sensible and latent heat. All LW radiation not in the window regions does NOT; it’s constantly hopping floors up or down, and because convection goes both ways, its net effect on radiative flux outside the window region is ZERO.

Another key point. CO2, or any other IR active gas, can only absorb the available radiation.

Agreed, and again, I know of nobody in “consensus” literature arguing otherwise.

Apparently, this amount never exceeds that which can be absorbed relatively close to the surface other than that which goes through the window directly to space.

Sure. I mean, how could it without violating any known laws of thermodynamics?

So the degree of thermalization is not affected by increasing CO2 or water vapor as long as they are present in sufficient concentration. What is affected is the rate of thermalization which is where convection comes in. The more radiation, the more absorption, the more thermalization, the more convection.

The implication then is that the system is doing more work. Think Carnot cycle. There are two ways we can increase the work done by a heat engine:
1) increase its efficiency
2) increase the temperature differential between the hot and cold reservoirs
… or we can do both. Your argument implies that the temperature differential remains constant, meaning that increasing CO2 leads to increased efficiency, option (1). If you agree with that, please explain how.

Although it isn’t exactly the opposite phenomenon, I agree that more CO2 in the upper atmosphere should increase cooling, all else being equal. BTW, some of us believe that increasing CO2 should have a net cooling effect because of its contribution to enhancing convective cooling below and radiative cooling above.

Yes I’m aware of the cooling at all levels argument. My standard response: explain Venus.

I don’t know the satellite methodology.

From my previous comments in this thread: January 30, 2016 at 5:22 pm
Compare the satellites. They’re in sun-synchronous polar orbits, designed such that for any one complete orbit they get readings at approximate local noon on the daylight side of the planet, and readings at approximate local midnight on the nighttime side of the planet. NOAA-15 has an orbital period of 101.01 minutes, or about 14.25 orbits per mean solar day. The orbits of the other sats are not such that covering the same track at roughly solar noon or midnight for any one location is guaranteed. Their orbits are all decaying and precessing at different rates.
Not only do the sats not give us a daily min/max for a single location on a daily basis, they don’t give us the same set of observations for any location on a daily basis.
What is your response? I’m most interested in the min/max part of the question here.

What is it and how is it worse than a high/low average taken of an incomplete set of non-homogeneous measurements?

I’ve been asking for your comment on the fact that sats don’t even give us a daily min/max for any given location for a couple of posts now. I consider it premature of you to be asking me about surface station homogenization while your response on satellite daily/min max is outstanding.

Chic Bowdrie,

I am enjoying the clarity of your comments, albeit lengthy.

Thank you. That last post is on the order of 1/2 the size of the original draft, concision is admittedly not my strong-point. I appreciate you sticking to the science and challenging my arguments on their merits (or lack thereof), a rare occurrence in my experience and much to your credit.

In case comments get closed here, there is a recent post about sats and Christy’s Senate testimony where we might rendezvous.

I’ve been avoiding it (too many political threads make me cranky), but may poke my head in as I referenced it in another comment to you elsewhere.

Whatever the circumstances causing us to arrive at present day conditions, there is a net amount of energy previously accumulated in the planet that allows us to monitor changes to the present day amount whatever it is.

I assume we’re talking about a pseudo-steady state (but never isothermal) where the net input/output flux fluctuates around zero on decadal time scales.

I think that’s perfectly suitable for this discussion.

Pardon my mistake forgetting that not all AGW proponents have the same understanding about backradiation.

No worries, it happens, I do it too.

My point was to reiterate that 161 W/m2 is the diurnal average of roughly 0 W/m2 at night and a daily average solar insolation of 322 W/m2 during the day or some similar paradigm accounting for latitude and longitudinal variations. […] Did I say bad physics? By physically unreal, I mean the tacit assumption that there is a constant source of solar energy radiating on all surfaces of the Earth at the same time.

Bad physics was my translation. Here’s Trenberth, Fasullo and Kiehl (2008): http://journals.ametsoc.org/doi/pdf/10.1175/2008BAMS2634.1
There’s a box on pg 5 of the pdf called “Spatial and Temporal Sampling”. Looks like KT97 used a globally averaged 15 C temperature plugged into Stefan-Boltzmann with surface emissivity set to unity to compute outgoing LW. Further down is this:
To compute these effects more exactly, we have taken the surface skin temperature from the NRA at T62 resolution and 6-h sampling and computed the correct global mean surface radiation from (1) as 396.4 W m−2. If we instead take the daily average values, thereby removing the diurnal cycle effects, the value drops to 396.1 W m−2, or a small negative bias. However, large changes occur if we first take the global mean temperature. In that case the answer is the same for 6-hourly, daily, or climatological means at 389.2 W m−2. Hence, the lack of resolution of the spatial structure leads to a low bias of about 7.2 W m−2. Indeed, when we compare the surface upward radiation from reanalyses that resolve the full spatial structure the values range from 393.4 to 396.0 W m−2
Elsewhere they speak of datasets with 3 h temporal resolution. One specific solar reference comes from a discussion about clouds:
The chronic problems in correctly emulating the distribution and radiative properties of clouds real-istically in the reanalyses preclude those as useful guides for our purpose of determining a new global mean value. Accordingly, the most realistic published computations to date appear to be those of Zhang et al. (2004) in the ISCCP-FD dataset in which ob-served clouds were used every 3 h, at least for the solar components where the TOA view of clouds is most pertinent.
So while the latest cartoon might imply that daily means are driving all the subsequent calcs it’s pretty clear that they’ve made an attempt to take the diurnal cycle into account.

But what is more egregious in the cartoon is the long arrows with large numbers of LWIR going up and down. This leads to misinterpretation of the physics dealing with mean path lengths between collisions and the preponderance of absorption compared to emissions in the dense troposphere.

I agree with you that the arrows make it look like upwelling LW shoots straight to the tropopause and some of it is re-emitted all the way back to the surface. I also agree with you that we know that isn’t what happens because the path length at the surface is on the order of tens of meters or less. I don’t know what else to say to you other than this highlights the importance of actually reading the paper and not just looking at the pretty pictures.

It also causes misunderstandings of the 300+ W/m2 DWLR, commonly known as backradiation. The cartoon obscures the fact that the average 396 W/m2 radiating from the surface, corresponding to an average global temperature of 288K, is primarily due to the accumulated total energy of the planet. If the sun did not come up tomorrow, radiation from the surface would gradually drop as the planet cools, not go immediately to zero.

Ok, I think I’m finally beginning to wrap my head around this argument, and I agree with you in part: the upwelling LW is indeed to previously accumulated solar energy. I also agree that if the Sun blinked out that the planet would gradually cool, not immediately drop to the 2.725 K temperature of the cosmic background radiation. After that, we part ways because I don’t think the cartoon obscures that, but rather presumes the audience has enough experience with heat retention and everyday objects to not jump to that conclusion. I don’t typically appeal to common sense unless I’m trying to be a jerk, but I’m kind of at a loss for anything else to say.
I’d rather discuss the descriptions of the physics as written rather than what may or may not be implied by the diagrams, which I hold are necessarily simplified due to the complexity of the system.

“The only things riding the elevator all the way to the tropopause are sensible and latent heat. All LW radiation not in the window regions does NOT; it’s constantly hopping floors up or down, and because convection goes both ways, its net effect on radiative flux outside the window region is ZERO.”
I’m quite surprised you agree with that. Same for your next two statements.

I’m quite surprised you agree with that, I didn’t expect you to. I fear one or both of us has misunderstood the others’ previous arguments on this particular point.

By saying CO2 increases the rate of thermalization I wasn’t implying that the system is doing more work, and I’m not sure that is correct. Convection is globally isochoric. What’s expanding here is contracting elsewhere. There might be some non-pdV work to consider, but again it would be similarly compensated for. I’m way outside my comfort zone with this. What I meant to suggest is that more CO2 means more energy absorbed per unit time per unit volume of air and unit of available radiation. Therefore that unit parcel of air stimulates greater convection which is equivalent to faster cooling. I’ll have to learn how to say that in more thermodynamically explicit terms.

You may not have meant to imply more work is done, but when you say convection increases it does imply to me that more work is being done. It’s actually what I’d naively expect to happen if the surface is heated by any means, because that would increase the temperature differential between it and outer space. I say “naively” expect, because there was an interesting paper published last year and covered here which might call that into question.
Thanks for “isochoric”. Explain “non-pdV” to me, not a term with which I’m familiar.

“My standard response: explain Venus.”
No can do. Not prepared to do that now, maybe after looking into it which will be awhile.

No problem, it’s sufficient that you’ve at least considered it.

Sorry, but I can’t tell you what I don’t know. If the point is to expose my ignorance, mission accomplished.

But I explained it once already before you asked me to explain it. Aaack!!!! Not at all my intent to expose your ignorance, more just nudging you to revisit what I’ve already written — which is admittedly a bunch.

I presume that sats measure more frequently than twice a day and, between the 12 or so of them, cover a greater area more homogeneously.

Each sat is measuring almost continuously, I don’t know how many observations per orbit. Your mention of area is spot on as it were because each observation covers a pretty broad area:
http://www.remss.com/measurements/upper-air-temperature
http://images.remss.com/figures/measurements/upper-air-temperature/footprint_map_for_web.png
Figure 2. Two example scans for the MSU instrument. The satellite is traveling in the South to North direction, and scanning (roughly) West to East, making 11 discrete measurements in each scan. Footprints used to construct the near-nadir MSU products (TMT, TTS, TLS) from the top scan are shown in green. The numbers in each footprint are the weights assigned to the footprint when constructing the average for a given scan. The footprints used to construct TLT are shown in red and blue in the lower scan, with red denoting negative weight.
With multiple sats, there are multiple daily overlaps, what they do is grid everything up and take means, not unlike what the surface folks do. Differences are that the surface stations are fixed in place, and theoretically give an “exact” diurnal min/max, but not necessarily b/c different equipment has different response times, and its been shown that co-located instruments of different types give statistically different min/max readings over time. The sats only get an approximate local noon and midnight observation for any given spot, each sat slightly different deviation away from local solar time, requiring diurnal adjustments.

However, if sats only measure a given point twice a day at the same times, how is that better or worse than a high/low average?

Diurnal temperature range (DTR) is one key metric in the quiver of AGW predictions — it’s expected to decrease, i.e., minima are expected to increase at a slightly higher rates than maxima. I would expect that to show up better in the surface record than the satellite record based on methodology alone, but that may not be the case. My understanding is that estimating DTR is actually pretty tricky.
I don’t have a problem with how satellite estimates are being done per se. Multiple daily overlapping observations, gridded means — seems fine for purposes of estimating long-term climate trends. What I don’t buy is that homogeneity isn’t a problem, as is often alleged in this forum, because the literature on the sats much of which is published by the providers themselves clearly document the known homogenization issues that they’ve adjusted for. They also allow for the real possibility that some exist which they haven’t anticipated or yet detected. Surface observation literature contains all the same language.
Which is more reliable? You know my answer based on the trend uncertainty estimates. Which is better? Depends on the application. That I believe the satellite uncertainty to be higher doesn’t mean I think the data are useless.
The relatively flatter satellite trends in the “pause” era raises my eyebrows because it doesn’t match my expectations. One thing I’m keen on is how the current El Nino will play out. The next six to eight months will be interesting to me on that score.

Chic Bowdrie,

I realize that. What I didn’t realize is that taking the rotation of the planet into account doesn’t make the cartoon represent the pertinent physics any better. IOW, convection vs. radiation, etc.

I was surprised by that as well. Way I’m reading it, what they found out since ’97 is that the regional errors mostly cancel out.
At this point it’s not clear to me that you have any objections about the physics behind the cartoon, only its presentation.

I’m not the only one where that presumption has caused confusion. The cartoon implies that the additional heat to make up the 396 W/m2 comes solely from the atmosphere.

I just don’t understand that argument, sorry. It’s literally clear as day to me that insolation is the only input considered and that net LW flux at the surface is negative (a net loss). IOW, exactly what my understanding of “correct” physics would suggest.

Why would anyone want to estimate DTR?

Just as I said last post. Nighttime temps are expected to increase faster than daytime temps under GHG-forced warming.

We need actual temperature measurements as accurate as can be, not estimates of DTR from estimates of high and low temperatures.

I don’t see those as mutually exclusive in principle. In practice, for most historical surface data, min/max is all we’ve got. As technology progressed, we’ve been obtaining more and more hourly and sub-hourly data: https://www.ncdc.noaa.gov/data-access/land-based-station-data
Click on the “Integrated Surface Hourly Data Base (3505)” link, goes to an FTP folder with a bunch of data. I’ve no idea how extensive it is, or how it’s being used. I think it would be ideal if all surface data were hourly, and reported a min/max. We’d still be stuck with the legacy data however. We can only use what we have.

Trends depend on temperature measurements. Before a trend uncertainty can be used to compare measurement methodology, the accuracy (not necessarily precision?) of the methodologies must be known.

For long-term climate, I think precision is more important. Most important is lack of bias. If I have a thermometer that consistently reads 1 degree cold +/- 0.1 degrees, and is “known” to have not drifted from its initial bad calibration in terms of accuracy, then I don’t care about its absolute value for purposes of creating an anomaly time series.
Different story for modelling, we need to know the accurate absolute temperature.

I think I would trust trends from accurate satellite measurements with greater uncertainty than more certain trends from less accurate surface measurements.

So would I, but how do I know which is more accurate? How do you? You keep saying satellites are, how do you know?

I think, because I would have to do some kind of error propagation analysis to be sure.

I’m pretty sure you’re correct. Autoregression and co-variance matrices pop up a lot in literature on homogenization.

Which is why eyebrows of skeptics are raised when they see the adjustments in the surface data making the past colder instead of eliminating the urban heat effects of the present. Doesn’t that trouble you, as well?

Only until I did my homework. There are two punchlines: BEST, and the fact that SST adjustments are net-cooling. Busting “The Pause that never was” exercise of Karl et al. (2015) failed to put a dent in the long-term net adjustment. Going by magnitude of change, even Karl (2015) is less eyebrow-raising than what UAH TLT v5.6 to v6.0beta did post-2000, and they still haven’t passed peer-review OR released code yet. If it weren’t for RSS, that would be sufficient to throw UAH under the bus by way of the raised-eyebrow criterion.

Chic Bowdrie,

How can the LW flux at the surface be a net loss? I would expect you to claim a slight gain with the excess going into the ocean or being the cause of the alledged surface warming that replaced the alleged pause that wasn’t.

I do claim a slight gain, but not so much that net LW goes positive. For that to happen, the atmosphere would need to be warmer on average than the surface, which it clearly is not, implying that CO2 is a source of energy, which it also clearly is not.
Let’s look at the cartoon again:
Net LW flux at the surface, clearly negative, 40 W/m^2 alone due to the atmospheric window. Tally up all the other fluxes except the Sun, and you get -160 W/m^2. Absorbed solar is 161 W/m^2, net balance = +1 W/m^2.
Sun heats the surface, GHGs reduce the rate of loss. Temperature rises until the radiative balance is restored. I do believe most of the excess is going into the oceans:
http://climexp.knmi.nl/data/iheat2000_global.png

5.47E21 J/yr / 3.16E07 s/yr = 1.73E14 J/s (W)
1.73E14 W / 5.10E14 m^2 = 0.34 W/m^2

As you cited earlier, Stephens (2013) gives 0.6 +/- 0.4 W/m^2, so with just over half of the mass of the oceans considered, we’re already inside the error bounds and over halfway to the central estimate.
Compare HADCRUT4 over 1957-2015 as a proxy for net atmospheric temperature change:

0.0130 K/yr * 5.18E+18 kg * 1.005 kJ/kg K * 1,000 J/kJ = 6.72E19 J/yr
6.72E19 J/yr / 3.16E07 s/yr = 2.13E12 J/s (W)
2.13E12 W / 5.10E14 m^2 = 0.004 W/m^2

That’s a rounding error no matter whether we’re estimating air temperature change with surface thermometers, balloon-borne thermometers or orbiting microwave sounding units. But noting that 0.004 / 0.6 = 0.7%, it’s within reach of what the IPCC say we should expect (I do this from memory), and multiplying 0.6 W/m^2 by these percentages:

0.93 oceans 0.558 W/m^2
0.05 land        0.030 W/m^2
0.01 latent heat 0.006 W/m^2
0.01 atmosphere  0.006 W/m^2
----
1.00 total       0.600 W/m^2

For sake of argument, I’ll assume all my missing heat is going into the oceans …

5.47E21 J/yr * 0.558 W/m^2 / 0.34 W/m^2 = 8.98E21 J/yr
8.98E21 J/yr / 1,000 J/kJ / 4.006 kJ/kg K / 1.35E21 kg  = 0.0017 K/yr
0.0130 K/yr / 0.0017 K/yr = 7.8

… implying that surface temperatures are rising about 8 times faster than the vertical average ocean temperature.
Now, NOT assuming all the missing heat is going into the oceans, I get a ratio of 7.6:1 surface/ocean temperature.
From Bintanja (2008) I expect the ratio to be 5:1. The higher ratios of 7.6:1 observed (or 8:1 calculated) combined with an estimated imbalance of 0.6 W/m^2 plus the well-constrained physical parameters of ocean water and atmosphere strongly suggests to me that oceans are in fact absorbing most of the heat and are the main reason for the lag between increased forcing and final (pseudo-)equilibrium temperature.
In short, as is often said, there’s more warming “in the pipeline” even if CO2 levels stabilized tomorrow …
… but only, of course, if the observations above are reasonably correct, I didn’t badly screw up any math and long-standing theories of radiative physics aren’t completely wrong.

“Nighttime temps are expected to increase faster than daytime temps under GHG-forced warming.”
That may be the case, but that doesn’t mean that average daily temperature is actually increasing.

Minima and maxima are both increasing. It’s difficult to imagine the mean not going up as well.

A high-low average is biased towards warming.

Why?

The only way to know if the actual temperature is rising is to make frequent measurements, ideally infinitely many.

Never going to happen. Already NOT happening with satellite data, which you do trust.

Precision is desirable, but accuracy is king.

Depends on intended application as I’ve explained previously. If you need absolute temperature, accuracy is absolutely king.

Bias is lack of accuracy.

I though we agreed that bias in this application is a drift in the absolute reading over time. If the initial calibration was wrong, so long as the thing doesn’t drift from that wrong calibration, precision is what I’d most care about for purposes of trend calculation.

The whole urban heat island issue involves artificially high temperature readings, aka bias, aka lack of accuracy.

It’s the increased urbanization which is the problem in UHI — change in surroundings over time. You can have UHI in Minneapolis, which is a damn sight cooler on average than a remote atoll in the tropical Pacific which has never so much as had an outhouse built on it.

It doesn’t matter how consistently wrong a reading is, if it’s wrong, it biases the results.

Again, that’s only true if we care, or only care, about absolute temperature. The slope of two parallel lines is exactly the same no matter what their y-intercept is. Temperature anomaly products are not averages of mean absolute temperatures, but mean anomaly from a baseline average for each series of observations.

If I jumped to the conclusion satellites are more accurate, I shouldn’t have. I haven’t looked into the methodology that closely.

Spencer et al. (1990) may be a good place to start: http://journals.ametsoc.org/doi/pdf/10.1175/1520-0442%281990%29003%3C1111%3AGATMWS%3E2.0.CO%3B2
They report precision of +/-0.03 K for two-day hemispheric averages, and +/-0.01 K for one month means. If they report what the typical precision is for a single grid, I did not see it. As you dig in, you’ll find that much depends what the satellite is flying over, implying that not all grids are equal from the standpoint of precision.

The way to test is to take data from locations known to be free of urban heat effects and make sufficient measurements to eliminate diurnal bias.

I agree. However, going back to 1850 and doing it over proper by any definition of “proper” isn’t an option.

Then the satellite measurements must cover those same areas.

Clearly not an option prior to 1979. And we’re still humped when it comes to the poles with satellites, IIRC because not enough is known about surface emissivity in those areas to correct for surface returns as they do everywhere else.

Of course this is still comparing apples and oranges.

Yes, that’s what I’ve been saying.

Radiosome data at the same locations might be able to help synchronize the data somehow. It’s possible that there is a real, ie accurate, actual difference between these measurements that changes from place to place. For example, land vs. sea, flatland vs. mountains. This all makes evaluating trends complicated.

I very much agree.

But without this baseline info, how can trend comparison be definitive?

That all depends on what your standard of definitive is. If you need > 6-sigma certainty, which I understand to be the standard in high-energy physics, I’m pretty sure you’re not gonna get it any time soon. I go by the error estimates and say, this is what we think is true within that envelope … same as I would any other complex science.

Brandon,
Now I think I understand why the high/low average is usually biased towards warming. The rate of temperature rise in the morning will generally be faster than cooling later in the day. But the cooling rate decelerates at night causing a simple high/low temperature average to be greater than an average of hourly readings. The greater the deceleration, the greater the resulting high/low average bias.
If this bias remained the same year to year, then the long term trends wouldn’t be affected. However, let’s say that increasing CO2 or water vapor causes long term diurnal warming or cooling rates to change. These changes could bias the trends, because the high/low average daily temperatures would not reflect the truer daily temperatures calculated from the average of hourly readings.