Peter Webster started an interesting discussion over at Judith Curry’s blog about the interesting mid-century blip in the temperature record where (relatively) rapid warming transitioned into a few decades of stagnant or cooling temperatures. He provides a critique of Kevin Trenberth’s statement that:
The warming in the first phase was not global but focused in the
 North Atlantic. It was related to changes in the ocean. The warming in
the southern hemisphere is more steadily upwards: no steadying off or
 down phase. So the patterns of change also matter.
Dr. Webster examines two data sets: an arctic temperature chart from Polyakov et al 2003 showing warming from 1920-1940 and a decline post 1960, and a figure from an upcoming paper by Jones showing warming from 1930 to 1940 and cooling from 1940 to 1950 (Dr. Webster argues that the figure shows cooling “through the 50s and 60s”, but that seems not to be shown in this particular image). He raises two questions:
- Even though the IPPC noted the mid-20th century warming , why was it not highlighted as something worth investigating? Various unsatisfying explanations (to me at least) for this are given in Ch 9 on attribution..
- Why is the southern hemisphere drop in temperature from 1945+ much the same as the northern hemisphere? The standard argument for the drop is the increase in aerosols with enhanced industrial activity after WW2. But this seems strange as even now the ration of aerosols between hemispheres is 2:1 NH: SH.
To assess both of these points, its useful to examine the global land and ocean temperature records separately for three distinct regions: the northern hemisphere (> 23.6 latitude), the tropics (23.6 to -23.6 latitude), and the southern hemisphere (< -23.6 latitude). I’ll be using GHCNv3 raw data for land temperatures, and ERSST for SSTs, as I have them handy.
Lets start with the global land record (shown above). Here we see 5-year averages centered on the midpoint of the period averaged. If we zoom in on the warming period from 1900 to 1940 we get:
Here we see that while warming was most rapid in the N. Hemisphere during this period, it was also present in the tropics in southern hemisphere. The contradicts Trenberth’s assertion that the early warming period was not globally distributed.
If we look at the 1940-1975 cooling period, we get this:
In this case, we see rapid decreases in land temperatures in the Northern Hemisphere, with more gradual decreases in the tropics and southern hemisphere. This appears to be consistent with the position that cooling was primarily due to N. Hemisphere aerosol forcing, and contradicts Webster’s argument that the rates of cooling are consistent across hemispheres during this period.
Now lets turn to Sea Surface Temperatures:
Here we see a much more significant difference between the Northern Hemispheric (> 25 lat) record and the other records, though it is worth noting that there is less ocean area in that latitude band than in all of the others.
When we examine the 1900 to 1940 period we get:
The northern hemispheric and tropical areas are both trending strongly upward, but in the southern hemisphere we see little positive trend during this period. This would tend to somewhat support Trenberth’s claim, inasmuch as the southern hemispheric temperature is particularly dominated by the ocean component. That said, this data should be taken with a strong caveat that SST readings during this period were scarce, particularly in the southern hemisphere. Also, the warming in the tropics seems to contradict the claim that it was primarily in the Northern Atlantic.
If we look at the 1940-1975 cooling period, we get:
Here we see considerably less SST cooling than we saw land cooling. There is no cooling at all in the southern hemispheric oceans or the tropical oceans, and only a small negative trend in the northern hemispheric oceans.
So, to sum up:
- Land temps show strong early 20th century warming, somewhat counter to Trenberth’s claim that early 20th century warming was dominated by the Northern Atlantic (though he was referring to global temps, not just the land component).
- Land temps show much stronger post-1940 cooling in the northern hemisphere (> 25 lat) than in the tropics of southern hemisphere (< 25 lat), consistent with the hypothesis of aerosol-forced cooling.
- Ocean temps show much stronger early warming in the northern hemisphere and the tropics than in the southern hemisphere, where there is barely any warming at all. However, sparse coverage makes this conclusion somewhat tentative. Strong tropical warming seems to also run counter to Trenberth’s claim.
- Ocean temps show no post-1940 cooling in the tropics and southern hemisphere, and a small amount in the northern hemisphere. This again seems consistent with the aerosol hypothesis.
In the end, it would seem that Trenberth is likely incorrect about the geographic distribution of early 20th century warming, but that the data seems not to seriously contradict the popular hypothesis that post-1940 cooling was largely driven by rapidly increasing aerosol emissions in the northern hemisphere.
Zeke,
This is interesting because there is a notion in some circles that the apparent long-period oscillations in the global temperature record are due to the PDO (or the AMO, or some combination thereof). However, there is little or no evidence of such cycles in the SST records you have shown here!
Julio,
The N. Hemispheric ocean temps look somewhat cyclical, though the relatively short period examined relative to the duration of the oscillation makes it difficult to say for sure. Tropical and S. Hemispheric SST show less obvious signs.
.
A small correction: the global land temps via GHCNv3 shown in the first figure shows 1880-2009 with a 5-year smooth, and does not include 2010 data.
The climate operates on daily, monthly and 3 month timescales. It does not operate on 5 year timescales.
Using annual, 5 year moving averages or 100 year moving averages hides all the important detail like what is actually driving the temperature changes.
The monthly Hadcrut3 hemispheric charts demonstrates this fairly adequately.
http://hadobs.metoffice.com/hadcrut3/diagnostics/regional/30-30/monthly.png
http://hadobs.metoffice.com/hadcrut3/diagnostics/hemispheric/southern/monthly.png
http://hadobs.metoffice.com/hadcrut3/diagnostics/hemispheric/northern/monthly.png
Julio, Zeke,
All of the data presented by Zeke in this post is entirely consistent with PDO.
I don’t know Bill, the thing that operates on daily, monthly and 3 month timescales sounds more like weather 😛
A small difference in ocean temperature is still a large difference in heat energy. Changing poleward heat transfer in the ocean may well result in much larger changes in land temperature as it probably affects air circulation as well. Ron Broberg has demonstrated a pretty good fit for global temperature with a combination of an exponential and a sine wave.
I favor a parabola and a sine wave, myself 🙂
But does the sine wave correspond to a real, physical oscillation? I see what Zeke means about the NH SST “oscillation”, but it seems way out of phase with the land…
“I see what Zeke means about the NH SST “oscillationâ€, but it seems way out of phase with the land…”
That’s because the main cooling component of PDO is ekman transport along the west coast of N. America causing cool onshore flows on a large scale… is what Liza has been talking about it being so cold in S. California at the moment
“post-1940 cooling was largely driven by rapidly increasing aerosol emissions in the northern hemisphere”
Not necessairly from aerosols. That is testable, what was Siberia during that period? Or Northern Africa? Little aerosols emitted there.
This paper seems to reject aerosols:
https://pantherfile.uwm.edu/kswanson/www/publications/2008GL037022_all.pdf
Bill Illis:
I’d have to disagree with that. The data say something very different:
See for example this.
Bottom line is the human eye isn’t a very good Fourier analyzer.
Zeke (Comment#66903) January 21st, 2011 at 8:46 pm
Two thumbs up to that.
There is no ENSO impacts in the 5 year moving averages. The Tropics variation, in particular, is driven by the ENSO. There is a direct +/-0.5C impact which is completely missing in the five-year averages.
In 1916-17, there was a significant La Nina event. Tropics temps fell to -0.88C. From this period until 1941 there were a series of two and three-year El Nino events. Tropics temperatures peaked at +0.50C at the end of 1941 and it didn’t really reach those temperatures again until the 1997-98 El Nino.
http://img199.imageshack.us/img199/7296/enso19161942.png
The 1877-78 Tropics temperatures are higher than they are today. Yes, super-El Nino in 1877-78.
Then the Northern Hemisphere which has a close correlation to the Gulf Stream. Unlike the AMO, the Gulf Stream does not need to be detrended, there is little trend in it. Just like there is no trend in the long-term ENSO data.
Not hard to see the above cycles in the Gulf Stream numbers.
http://img20.imageshack.us/img20/7135/gulfstream.png
Is it possible a] given that the earth’s ocean to land ratio is approx 70:30 and b] that the Pacific Ocean -which straddles both hemisphere roughly equally and is larger than the earth’s total land mass- has any bearing on this?
Aerosols appear to be trotted out when other usual explanations fail, when in fact we have evidence that temperature changes specifically in the Pacific [e.g ENSO, PDO] have profound short and long terms effects globally.
Bill Illis (Comment#66919) January 22nd, 2011 at 5:23 am
ENSO has been around for a lot longer than a century or so. Why hasn’t it created a hotter earth long ago?
bugs (Comment#66921)
January 22nd, 2011 at 6:07 am
ENSO has been around for a lot longer than a century or so. Why hasn’t it created a hotter earth long ago?
———————-
Obviously, it did.
Whenever there were a series of multi-year El Nino events, the climate cycled up.
In the Permian, the Pacific was twice as big as it is now. The El Nino and La Nina events of the time would have been massive events.
Of course, another way to look at this is to argue that this (NH, SH, tropics) is just not the right way to partition the data. Teleconnections may be important.
A few weeks ago, SteveF showed us something which I think Bob Tisdale has also shown somewhere, namely, that the integrated El Niño 3.4 index looks remarkably like the global temperature anomaly time-series including the 1940 “bump”. If I were a climate scientist trying to understand the reason behind the “mid century warming” (and subsequent cooling), I would definitely focus on that.
Bill Illis:
If that’s true, it has no impact on the long-term global mean average: According to this claim of yours, It would only modulate temperature over time scales shorter than 5-years.
“If I were a climate scientist trying to understand the reason behind the “mid century warming†(and subsequent cooling), I would definitely focus on that.”
That would require the worlds leading climate scientists to admit that their climate computers are all programmed wrong and that most of the evidence for AGW went up in a puff of hot air
Julio, from my point of view, processes like ENSO are primarily responsible for redistributing heat energy between the surface and the interior of the ocean. That’s why you see often see step-like functions in the ocean temperature after an ENSO event (like the strong one in the mid 1940s).
I believe there is a secondary effect on climate, which can be summarized thus: ENSO’s influence the net albedo of the Earth due, for example, to a northern shift of the Westerlies (and associated cloud patterns() during el Niño events.
So they modulate the net climate feedback. f(t) = f0 + fENSO(t).
However,
< G(t) > = < 1/(1-f0 – fENSO(t))> ≠< 1/(1-f0)>
even if < fENSO(t) > = 0.
(Of course this just produces an offset to global mean temperature. To get it to increase global mean temperature, you need the magnitude of < fENSO(t)^2 > to increase over time.
Carrick #66927,
.
I agree that ENSO represents a redistribution of ocean heat, but I suspect there is a good deal more than this happening. For example, the strength of easterly trade winds depends on the strength of the Hadley cell circulation. The strength of the Hadley cell circualtion depends in large part on the temperature of the ocean along the ITCZ (and the resulting degree of moist convection) .. which is partly controlled, with considerable lag, by the easterly trades via ENSO. This kind of system, with considerable time lag in response to a deviation, is typical of systems that have a tendency to oscillate.
.
It seems to me reasonable that the pseduo-oscillatory behavior of ENSO may be in large part a ping-ponging of warm surface water in the Pacific, with the pseudo-oscillation driven by the delayed/lagged feed-back of the Hadley driven easterly trades, pushing the water from east to west with varying force. Some independent measure of the volume of Hadley flow versus ENSO may be a reasonable place to confirm if this is in fact the case.
.
I think the important question is how much the ENSO (along with other identified oscillations like the PDO) might influence global temperatures over longer periods than the typical ENSO under conditions of constant forcing. If ENSO (and perhaps other pseduo-cyclical patterns) represents a random-walk component in the average temperature, superimposed on a deterministic trend, then would it not be reasonable to expect fairly long term (20 years? 30 years? longer?) random-walk type deviations from the underlying average temperature, even if there is always a tendency to ‘decay’ back to the deterministic trend?
.
The integration of the detrended ENSO history appears to explain the warming from ~1910 to ~1945, and the subsequent lack of warming (slight fall in temperature) from ~1945 to ~1975. I posted these graphs the other day; I don’t know if you saw them:
.
http://i53.tinypic.com/2pzbbed.jpg
http://i52.tinypic.com/i77yfd.jpg
.
That second graph sure looks a lot like a GHG forcing curve since 1900. The unexplained warming up to 1945 is no longer there. Note also how the effect of Pinatubo on average temperature becomes more evident in the second graph.
Carrick,
I forgot to add: the title on the first graph has a typo: it is supposed to read “0.065” not “0.06”
Do you assume a linear deterministic trend?
=============
Hi Lucia,
This is a little off topic, but do you have a T min only instrumental temperature reconstruction?
dallas
Re: Carrick (Jan 21 23:34),
Note the redness of all those power spectra. Power increases as the wavelength increases. There is no bright line at 30 years.
Captdallas2–
I don’t. I don’t know if Zeke does either. (I think Zeke and I may need little avatars in addition to the “author” label to indicate who wrote a particular post.)
It is good to see the global warming broken down by regions.
Another interesting analysis is to not assume trends as being linear over long time periods but better broken into shorter periods (of linearity) with statistically significant break points. I know that Steve M claims that breakpoints are controversial in the staistics world, but to look at graphs and assume there are long term linear trends is a bit of stretch and tends to ignore the shorter term trends and possible causes of them.
A more detailed analysis would be to look at breakpoints in regional and zonal temperature trends.
Detrending ENSO does not work unless you also consider all the secondary effects such as PDO which is caused by both ENSO in the south and weather noise from northern patterns. It’s a difficult task given that many of the effects felt outside of the ENSO regions also come with a lag.
Well the ‘linear deterministic trend’ which is the recovery from the LIA, certainly need not be linear. Whatever it represents seems also to have varied on the centennial to millenial scale.
Livingston and Penn coming on in the backstretch. Koutsoyiannis keeping up with the leaders. If I’d bet on Ol’ CO2Ball…
===================
Zeke, you say “In this case, we see rapid decreases in land temperatures in the Northern Hemisphere, with more gradual decreases in the tropics and southern hemisphere. This appears to be consistent with the position that cooling was primarily due to N. Hemisphere aerosol forcing, and contradicts Webster’s argument that the rates of cooling are consistent across hemispheres during this period.”
But it is also “consistent with” (which only says that it does not disprove but doesn’t prove) any theory based on a difference between Northern and Southern hemispheres. Such as % land area, population, degree of industralization, types of vegetation, amount of snow cover, % of land above a given latitude,average wind speeds and directions and sources, etc…
One can think of several of these that would have disparate effects on over land cooling rates as well as warming rates for the Northern and Southern hemispheres. For example with larger land areas the proportion of land near the oceans compared to total land area will result in less of the temperature being moderated by the oceans more stable temps. One can make arguments for any of the other factors alone or in combination explaining the disparate rates of cooling without needing aerosols. This of course does not mean that aerosols are not a part of the reason. Just that we don’t know and “consistent with” is an insufficient argument.
So being “consistent with” is pretty much not very useful in supporting the aerosols kept us from burning up idea.
John Knapp,
Good point on the “consistent with.” Indeed, any warming will always be “consistent with” AGW hoax. When Climate Scientists can show me some warming that isn’t “consistent with” AGW, that would be something. But that would require actual science, so we know that’s not going to happen.
Andrew
… not to mention that observations of optical depth do not add up with the mid 20th century dimming fantasy… but hey, you need to explain all climate change, both cool and warm, on man.
Aerosols -all else failing- explain it all. Aren’t we kidding ourselves?
Folks, as demosntrated elsewhere on this site, we are debating the possible “why” of purported changes that in terms of long term stats are off by at least an order of magnitude of what is verifiable. {0.009C ???]
How about we all have an appropriate beverage [depending on where we happen to be on the face of the planet right now] and take at deep “thinker” about what the point of the argument really is.
0.00whatever?
DeWitt:
Of course, power to (nearly) 1/f spectral relations is pretty much ubiquitous in nature. For example wind noise generated by convectively forced turbulence goes as 1/f^(5/6), as do the thermal fluctuations (Kolmogorov spectra).
IMO there’s nothing really “magical” or really about the appearance of the approximate 1/f relationship here. It probably has something to do with overall dimensionless scaling factors (one way to derive the Kolmogorov scaling law is via dimensionless analysis).
I’ve never seen evidence for a 30-year component, in any of the data I’ve looked at. What’s the significance of a 30-year component in any case?
Zeke,
I am not at all sure you can point at aerosols as the explanation.
The land temperatures worldwide historically show trends that are >1.5 times the ocean trend. The northern hemisphere has much more land than the southern hemisphere. Does this not suggest:
1) The northern hemisphere will always show trends (positive and negative) of greater magnitude than the southern, independent of aerosol effects.
2) Warming/cooling on land in the northern hemisphere will be in part transferred to the northern hemisphere oceans to a greater extent than the corresponding transfer in the southern hemisphere… more ocean warming and cooling in the north than in the south, independent of aerosol effects.
3) The tropics should lie somewhere between.
SteveF:
I think ENSO represents a genuine forced oscillation involving a westerly moving Rossby wave and an easterly moving Kelvin wave.
These sort of spectral peaks are probably akin to seeing a resonance peak resulting from driving an oscillator with random noise.
Carrick,
According to Wikipedia, 1/f -like noise spectra “generally correspond to a wide range of non-equilibrium driven dynamical systems.” That seems a good description of climate.
Regarding north versus southern hemisphere,
There is almost no latitudinal amplification in the Southern hemisphere. Even for land-only.
Part of that Northern-only amplification might be due to the fact that on average land in the Northern hemisphere has a higher altitude (so you get a larger positive feedback from loss of glacial ice).
See this.
If glacial ice loss is an important mechanism for the rapid heating seen in northern climates, there is a “saturation point” beyond which this mechanism shuts off (when most of the permanent ice is lost of course). That has interesting implications itself for potential future warming.
SteveF:
Yep I think so too. Throw in the presence of modes and pseudo-modes driven by 1/f noise and I think that pretty well describes long-term climate variability.
It is also interesting to use the GISS temp map generator to generate trend maps for 1900 to 1940 and comparing to trends from 1980-2010. http://data.giss.nasa.gov/gistemp/maps/
A map of trends from 1900 to 1940 shows a pattern of warming PDO (but neutral to cooling ENSO), and also a warming Atlantic is very obvious. In comparison a map of trends from 1980 to 2010 shows a clear trend towards cooler PDO and cooler ENSO values, and the warming in the Atlantic appears somewhat less significant compared to the rest of the globe.
From this it seems clear to me that if anything PDO/ENSO has reduced the warming trend over the last 30 years, and it is not clear whether an AMO type factor has contributed to global warming, or whether the Atlantic has warmed in reponse to the general global trend.
Not sure this has been discussed in this context before, but worth people being aware of:
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6VHB-5207B59-2&_user=10&_coverDate=01/19/2011&_rdoc=1&_fmt=high&_orig=search&_origin=search&_sort=d&_docanchor=&view=c&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=60ee908b2204f368177ccb057bf379b7&searchtype=a
“The Role of Atmospheric Nuclear Explosions on the Stagnation of Global Warming in the Mid 20th Century”
Yoshiaki Fujii
Rock Mechanics Laboratory, Hokkaido University, N13W8, Sapporo, 060-8628, Japan
Journal of Atmospheric and Solar-Terrestrial Physics
doi:10.1016/j.jastp.2011.01.005
Arthur, Dunno if its been discussed here but it has been researched. The timing and type of nukes along with the material ejected into the atmosphere did not jive with the changes in temps. However, natural variation does explain those changes… and of course, the author of the study you linked used a GCM which has little accuracy unless you tune the numbers.
Re: Carrick (Jan 22 15:15),
Instrument drift is also ~1/f and limits the precision that can be obtained by integrating an initially shot noise limited process like photon counting for longer and longer periods.
As far as the AMO affecting the rate of global temperature increase, the next decade should tell the tale, provided the AMO follows the script. A sine wave plus exponential fit says the rate of increase should be below the rate for 1970-2000 for about the next twenty years. After that, the exponential increase dominates. A pure exponential, though, probably exaggerates the rate in the late twenty-first century compared to the A1B scenario. Then there’s the question of whether all the IPCC scenarios exaggerate the rate of increase of ghg’s in the atmosphere for 2050-2100 by overestimating the consumption rate for fossil fuels.
Re: MikeC (Jan 22 18:06),
Not to mention being from Japan where they are, not surprisingly, obsessed with the effects of nuclear weapons (Godzilla, e.g.).
If you look at the daily UAH numbers back to 1978, I don’t really see the different hemispheres on different trends or different cycles. Some short-term differentials but they are all heading the same direction all the time.
Daily UAH back to 1978.
http://img109.imageshack.us/img109/7522/dailyuah1978.png
Daily UAH over the last 5 years.
http://img692.imageshack.us/img692/138/dailyuah2006.png
Daily UAH in 2010.
http://img831.imageshack.us/img831/4550/dailyuahtemps2010.png
The active areas where the ocean is exchanging energy with the atmosphere (the ENSO, the Kuroshio, the Gulf Stream, the Atlantic Equatorial, the Brazil Confluence and the Agulhas) all seem to be on different cycles, some have no upward global warming trend at all while others have a strong upward trend).
But it seems they must be acting in concert together for the UAH signal to be so coherent amongst the different regions.
Perhaps the differences prior to 1978 are, in fact, artifacts in the data collection and adjustment systems.
Perhaps all the differential signal in Hadcrut is just Arctic polar amplification (which UAH doesn’t cover that well) and then the additional plus or minus influence that the ENSO has on the Tropics.
MikeC (Comment#66954) January 22nd, 2011 at 6:06 pm
Natural variation does not explain those changes. You could say that it is within the bounds of natural variation, but you cannot say it explains those changes.
Question:
If you have a non-linear fit to a data set, what’s the best way to estimate the prediction intervals for the data and confidence intervals for the curve by a Monte Carlo approach? Do you take the best fit curve, add noise and fit more curves or do you add noise to the coefficents and generate a lot of curves? I lean towards the first approach because I don’t know if the coefficients are independent or orthogonal or whatever is the correct terminology.
Sure it does bugs, all you have to do is quit mis-programming the good ol’ GCM and they’ll be right there in front of your face. Of course that also means putting the bong down and knowing how the climate system works instead of trying to blame everything on people who will not give you sponge money. It’s how I beat the computers every year on ENSO forecasts.
DeWitt:
Actually, there are instrumentation amplifiers that get around this limit (they chop the signal). See e.g., this. But, yeah, I’m well aware of 1/f noise in amplifiers. It’s something I battle in my own measurements.
Weather noise is a bit difference because it’s usually not “true” 1/f noise, just “nearly”.
Re: DeWitt Payne (Comment#66959) January 22nd, 2011 at 6:44 pm
Do you mean (a) add noise to the data set and compute a new fit, (b) add random noise to the best fit curve assuming that it represents the “true” values and compute a new fit, or (c) something else entirely?
Re: Arthur Smith (Jan 22 17:02),
nice arthur.
Re: oliver (Jan 22 20:14),
That would be choice b. For a linear fit, the confidence limits for the line are narrow in the center where the error is only due to the error in the intercept and widen at the ends to accommodate error in the estimate of the slope. The prediction limits for the data are wider. The predict.nls function in R is supposed to calculate confidence or prediction intervals for non-linear functions, but doesn’t at the moment.
Carrick, DeWitt,
The 1/f variation in any system is only correct over a limited range of frequencies on the low end. At zero frequency, the 1/f noise goes to infinity, which of course can’t really happen. What I see with Instrument 1/f noise is that it becomes less and less important, and deterministic causes for variation (temperature coefficients, component aging/degradation, etc.) more and more important, below a certain frequency; that transition frequency I guess depends on the characteristics of the system being measured.
At Wikipedia, they say (If I understand correctly) that noise will be 1/f for the superposition of Poisson noise sources processes, each with a characteristic exponential time decay constant L(n) (units 1/time), and with said L(n) values continuously distributed over a range of values Lmin to Lmax. For f falling outside the range of Lmin to Lmax, the noise is no longer 1/f. I guess this makes sense; below Lmin, the signal becomes independent of f, above Lmax, the noise becomes white (1/f^2).
.
The climate sure would seem to be a superposition of noise processes with different decay rates, with an underlying frequency-independent trend. Finding the appropriate Lmin would seem to be the challenge to understanding the frequency range of natural variation. I guess that means understanding enough of the important stochastic processes to identify Lmin.
Sorry that was from Scholarpedia (http://www.scholarpedia.org/article/1/f_noise), not Wikipedia.
SteveF did not you mean the 1/f^2 is red noise and 1/f^0 is white noise?
Kenneth Fritsch,
Yes, that is correct; I had them flipped. 1/f^2 is ‘brown/red’ noise (Brownian Noise).
When, as Zeke has done here and Webster in his post, we take what might be considered arbitrary time periods and calculate trends are not we doing something similar to what is considered weak statistically when we look at breakpoints in a long time series.
Zeke get points, I think, for attempting to associate those time periods with some historical event like heavy aerosol production and its regionally concentrated effects and thus the time periods are perhaps not necessarily arbitrary -although one must search and note the historical records to avoid hand waving here.
Within those time periods we see what might turn out to be statistically significant breakpoints (I have not made the calculations). I think those breakpoints, if significant, would need to be explained.
I think breakpoint calculations get points for being statistically rigorous given the assumption of segmented linear trends. If one could associate a mechanism with these breakpoints then you might have some better insights into what is actually occurring.
Re: SteveF (Jan 23 10:03),
Which is my argument that no climate variable actually is a unit root process. A true unit root process is unbounded and the noise power does increase without limit as the frequency goes to zero.
Re: #66973
One possible reason for a breakpoint in the 40s is station moves from city centers to ex-city locations. MikeC has pointed this out and Warwick Hughes did a piece about this happening at Sydney. Have the NCDC datasets been built with this type of problem removed?
DeWitt Payne (Comment#66975),
.
I completely agree. It is a perception issue. The same process viewed at different time scales (and and different sampling frequencies) appears completely different, even though it is the same process.
.
The exact same concept applies to limitations placed on deterministic reality by time and physical scale. We think of the macroscopic world as ‘deterministic’, but in reality that determinism is restricted to a very limited range of time scales and associated size scales. The motion of a very small particle bouncing around under the influence of Brownian motion is just as ‘deterministic’ as is the pitcher’s throw, or the motion of Mars around the sun, so long as the time scale considered is short enough. if viewed on a time scale shorter than molecular collisions, Brownian motion is completely deterministic; on any longer scale, it is perfectly random, and so perfectly ‘red’ in terms of noise. The location of Mars relative to Earth is ‘deterministic’ on a much longer time scale than a particle moving under Brownian motion, but beyond some X millions of years, that determinism fails, and the position of the orbi of Mars relative to the orbital position of Earth is indeterminate.
.
I believe that the ‘color’ of noise (white, pink, red) is intimately related to the ranges of time and size scales where processes are deterministic in nature. On very short time scales, climate variation (AKA weather) is ‘red’. On very long time scales, climate variation is ‘white’; in between (the world that we humans actually live in), it must be considered ‘pink’, but depending on the processes and times scales being considered, covering everything from almost completely white to almost completely red.
D’ooooh Bears. 😉
Andrew
torn8to,
Beats me, I stopped looking at GHCN long before Peterson became the new knife wielding ninja climatologist… but if they do have an urbanization adjustment, it would be nice if all the analyzers would start looking for urbanization where the literature says it is… Tmin
torn8o (Comment#66976) January 23rd, 2011 at 11:32 am | Reply w/ Link
Re: #66973
One possible reason for a breakpoint in the 40s is station moves from city centers to ex-city locations. MikeC has pointed this out and Warwick Hughes did a piece about this happening at Sydney. Have the NCDC datasets been built with this type of problem removed?
############
Station histories are hard to come by. I can say this.
if you look at stations that were rural in 1900 and stayed rural to 1940
AND
stations that were rural in 1900 and become urban in 1940.
you will not find any UHI signal. quite the opposite. you will find rural
showing more warming.
SteveF:
Yes, of course.
For atmospheric turbulence, the picture is you have a “source region” which is shedding vortices that follow a 1/f^(5/3) scaling law. If you want to use a more realistic model, something like von Karmen’s model of turbulence gets used. (E.g., See this)
In the case of atmospheric turbulence in the boundary layer, the cut off frequency corresponds to the “overturn” time of the boundary which is typically 20-30 minutes. (However, there are still plenty of caveats even on this statement.)
Re: Steve Mosher
Rural shows more warming over what period? 1900-1940 or 1900-present?
Specificly Mosh, it would be nice to see a comparison between rural and urban Tmin
Re: torn8o (Jan 23 21:40),
1900-1940. Rural shows more warming. ( 1940-1980 urban shows more, rural
cools) haha but I’m letting some cats out of the bag
So possible explanations.
1. random. ( not significant)
2. land changes
3. geographical bias
4. urban aersols ( increased albedo)
5. bogus adjustment.
6. bad UHI proxy.
7. insensitive method.
So basically if you expected an easy answer ( rural warms the same or less) the data won’t give that to you on first pass.
means a whole lot more work.
eliminating each of these is a PITA. troyca also has some neat work that is pointing at some other things.
Re: MikeC (Jan 23 22:17),
rural Tmin versus urban tmin.
I might get around to that.
AGW predicts a narrowing of DTR, specifically it predicts that Tmin will show the impact of global warming more than Tmax. So just from a signal detection
standpoint I’m not sure the uhi signal will be any more detectable if you focus on TMin, but it might be worth a look. in due course. I work on what interests me. The data is open, the code is easy. Learn R and make fewer comments.
It seems *strange*. Is that a code word for fraud or incompetence. Why didn’t scientists investigate all the science at once, and come up with all the correct answers instantly. Why aren’t the models perfect by now, come to think of it?
For example, the number of full time staff at CRU at UEA, including Jones, comes to a the grand total of three.
http://www.nature.com/news/2010/101115/full/468362a.html
Re: bugs (Jan 24 06:11),
“For example, the number of full time staff at CRU at UEA, including Jones, comes to a the grand total of three.”
Bugs, wrong as usual.
what Bugs meant to say
“Outsiders are often surprised at how small the unit is, with just three full-time staff scientists.”
which is also Misleading as usual:
http://www.cru.uea.ac.uk/cru/people/
So when bugs is wrong and misleading, may we say he’s right?
==================
Yeah Mosh, But global warming will have an equal impact on Tmin, so if there is an urbanization signal then it’ll show. And try to get R on a netbook, doesnt happen
It’s kind of odd to to hear the period from the earlier 1910’s to the mid 1940’s termed “mid-century”…
What I find most interesting is that said period had, globally, a very similar (nigh-identical) over all trend to the similar length period of the last thirty or so years. There have been MANY people who have suggestions as to what that could mean. However, what it suggests to me is nothing more and nothing less than that nature is capable of having climate trends of magnitude comparable to the recent warming occur all on their own with no obvious “cause”. This is damaging to the strength of the “attribution argument” I think, though not a complete refutation…
I’m pretty confused and lost about the whole aerosol forcing thing. When I hear this I think of a person spraying a can of hair spray? If that is part of the aerosol forcing that is a joke.
Re: MikeC (Jan 23 22:17),
Tmin exaggerates any trend, according to Pielke,Sr., because on calm nights with clear sky and low humidity you get a temperature inversion close to the ground and the 2 m height of the measurement stations is inside that inversion. The possibly different surface roughness for rural and urban may have an effect on the development of temperature inversions. If it were simple, somebody would have done it already.
Steve, when they say full time staff scientists I suppose they mean “hard-money” scientists, i.e., “chaired positions” or “fellows”. Three is actually a lot for that category of job. Clearly it’s a good sized organization:
Ouch, bugs! I think I think you just got squished.
DeWitt, Youu’ve cracked the nutshell… there is a whole section in the literature about UHI and it’s effects. UHI is well known and it’s effects on Tmin are well known. So my question for the analyzers is why are they looking for UHI in Tmean when it’s effects are in Tmin?
DeWitt:
I agree.
But it’s easier to look at the distribution of warming, as Mosh points out, and you see a pattern that is not consistent with primarily urban-heat driven warming. My own claim is you can account for about 0.03°C/century of the trend from urban heating for a period where 1.6°C/century was observed… this is land + ocean… land only it’s closer to 0.1°C/century for the contribution.
Getting urban heating to be an important effect is very much a steep uphill climb: You have to account for the fact that 70% of the surface is ocean (no direct urban effect at all), and further that the strongest non-uniforimity in the global-warming trend is the latitudinal effect.
Urbanization looks attractive as a candidate for an important systematic in global temperature measurments… as long as you never actually run real numbers on it, and see what would be needed to actually account for a significant fraction of the total observed warming.
MikeC:
If there’s an effect in Tmin, then there’s an effect on Tmean. The only thing you could really criticize is if they were looking at Tmax, and no effect were expected.
No carrick, they’re not running the real numbers on it. Tmin and Tmax are affected by different things, ex: Tmax will be affected more by changes in hydrology. So when they avoid running the numbers in the right place, it sets off red flags.
… and since you’ve read all of the literature (except Oki as I recall), you should know this
I’m a little bit upset with Lucia choosing a Trenbirth quote because I think he has been “uncivil” and is ruining the spirit of the balanced debate. He is an idiot and he trys to cheat along with Michael Mann. Obviously, nobody here knows anything about Penn State and Boulder. You can’t find 2 worse, more far left wing schools around. They actually had a class at Boulder on hurricane Katrina. Lucia, in an effort to make the debate more civil can you please quit sourcing biased, whack job clowns?
MikeC:
Since you are well versed in the literature yourself, you are probably aware that:
Tmean = (Tmax+Tmin)/2.
So a signal in Tmin shows up in Tmean also.
This dig about reading literature is just an excuse on your part not to think.
No Carrick, if you properly adjust for UHI in Tmin and study any possible influences on Tmax such as changes in hydrology in rural areas, in farming areas for example, then you get closer to the truth. But some folks have an agenda and do not want to do the work because the results may be adverse to what they want to find.
MikeC, if you think that just looking at Tmin is going to make an astounding difference, then why aren’t you doing it?
I guess that’s my point. But it missing this comment of mine:
Carrick, I lack the hardware fo that, however, when I did pairwise comparrisons on Tmin, I saw exactly what I thought I would see. UHI. Not only UHI in the trends, but UHI in different parts of the records depending on time. And these were the homoginized numbers coming out of USHCN v2. The Earth is 2/3 ocean only tells you that in the final scheme of things, UHI will have less of an impact globally. So what is the effect of , say .45 UHI bias on the land record? 25% of the global signal (or less depending on what data set you are looking at)
“My own claim is you can account for about 0.03°C/century of the trend from urban heating for a period where 1.6°C/century was observed… this is land + ocean… land only it’s closer to 0.1°C/century for the contribution.”
What drives me crazy on the UHI effect is that some observers do not appreciate how the UHI effect can affect temperature trends. Most observers agree that UHI is real and relatively large on an absolute basis, but the problem then becomes what effect will it have on a temperature trend during a period of time x. If the local climate change occurred many years ago vis a vis affecting the temperature of the local and then more or less stabilized, we would see little of no trend over that time period.
Many climate scientist will also, correctly in my view, point to the fact that the exact location of where the temperature is measured in an urban setting is important – for example an airport outside the populated area. This observation in turn points to what I view as a much larger influence on temperature measurement than UHI and that is the micro climate at the measuring site – be that site urban or rural.
What also bugs me is this blind attachment of population growth with UHI effects – like adding a few thousand people to an already heavily populated area should have any effect on temperature. The ivory tower exercises like satellite observations of night lights and connecting that to a division of rural and urban for the purposes of calculating UHI I find a particularly lazy endeavor. The correct approach in my view is what the Watts team initiated when they did there on site observations of USHCN stations and gave them CRN ratings. Obviously in order to relate these ratings to potential effects on longer term temperature trends requires knowing when the observed changes occurred. For example, if the micro climate changed 30 years ago and stabilized the trend over those 30 years would not be affected by the change.
Carrick, I have no reason to doubt your claims in the quote above, but from the warming rate you show, I would guess that you are looking at the past 30 years and the UHI effect you claim over that time period would not be surprising.
I think he has been “uncivil†and is ruining the spirit of the balanced debate. He is an idiot and he trys to cheat along with Michael Mann.
.
🙂
MikeC:
There are two UHI effects. One of the potential systematic effect from oversampling urban areas compared to rural. The second is the “real physical impact”.
I guess I would have to challenge the suggestion that the second is purely a “global” effect. There should be a latitudinal effect, so that latitudes that have large (relative) urban samplings should show comparatively larger effects than those latitudes that don’t.
The problem I see with invoking too much of an effect (either systematic or physical) from UHI is that the pattern of warming is strongly biased towards high latitudes, with the latitude of maximum contribution to warming trend being around 62*N.
As to not having the hardware… I guess I’ll have to show a bit of unimpressed there. The resources needed for studying this are not great, with the biggest cost being ones own time.
This is a little off topic but does anyone know which item is greener, a paper towel or a hand dryer, and how much estimated warming do we get from each? For example, is there a record showing the impact of hand dryers on gat from coal usage? Is there a record showing temperature increases from the use of paper towels due to deforestation?
Kenneth:
It was based on data over a 60-year period, starting in 1950. I’ve looked at the 1980-2010 period too, that doesn’t significantly impact the conclusions.
Carrick, The geographical concerns you expressed are addressed in the gridding, arent they? And as for how I spend my retirement… well… mostly working (although I’ve taken the last month off and have been enjoying retirement)… but I usually focus on ENSO since that’s where the money is… I let those with the puter power and know how do the fun part with the temp data.
MikeC, my concerns relate to a change in the mean geographic latitude of the measurements over time, which is a bit more of a complicated issue than “just” the effect of gridding.
“It was based on data over a 60-year period, starting in 1950. I’ve looked at the 1980-2010 period too, that doesn’t significantly impact the conclusions.”
I would agree that the overall warming trend for the globe was what you say over the past 60 years simply by looking a Zeke’s graphs above. How did you estimate the UHI effect during that time period?
It’s true! It’s true! The crown has made it clear.
The climate must be perfect all the year.
A law was made a distant moon ago here:
July and August cannot be too hot.
And there’s a legal limit to the snow here
In Camelot.
The winter is forbidden till December
And exits March the second on the dot.
By order, summer lingers through September
In Camelot.
Camelot! Camelot!
I know it sounds a bit bizarre,
But in Camelot, Camelot
That’s how conditions are.
The rain may never fall till after sundown.
By eight, the morning fog must disappear.
In short, there’s simply not
A more congenial spot
For happily-ever-aftering than here
In Camelot.
Camelot! Camelot!
I know it gives a person pause,
But in Camelot, Camelot
Those are the legal laws.
The snow may never slush upon the hillside.
By nine p.m. the moonlight must appear.
In short, there’s simply not
A more congenial spot
For happily-ever-aftering than here
In Camelot!
Hello denizens of Camelot!
MikeC : ” PDO is ekman transport along the west coast of N. America causing cool onshore flows on a large scale… is what Liza has been talking about it being so cold in S. California at the moment”
It is not cold here at the moment during the day, sun is finally shining. I would say we are having a normal sunny time finally for So.Cal. (what do you say about C02 when the temp is normal? 😉 (still cold at night and in the morning (40°F or lower) when the sun is gone ) And not just “at the moment” its been cooler for years now; including in the summer; and this summer especially. And we had winter water temps this summer as well for the beach. We can gauge some of this by our backyard pool use over the last ten years too. Less and less use because heating it is a waste when the days are not warm enough for it and the nights are not warm enough too. It has changed and it’s not a change to “warmer” temps. 😉