Swanson and Tsonis 2009 on Regime-shifts

RealClimate was so concerned with our paper Structural break models of climatic regime-shifts that they felt compelled to post a paper by Swanson and Tsonis in response ;-), see Warming, interrupted: Much ado about natural variability.

There are a number of similarities and differences. Firstly, ST09 uses a very different method, fitting periodic fluctuations in ocean temperature to spatially distributed data. Regime-shifts are recognized with a measure of coherence between areas, and their theory is that shifts are more likely when synchronized. With this approach, they arrive at major shifts in 1976 as we do, and in 2002, after the big 1998 El Nino.

Using a simple Chow test for structural breaks, we get a break in 1978 and also in 1997, just before the big El Nino. These dates are within bounds of error, and differences in the method, so one might as well say they are consistent.

Tsonis and Swanson previously published on the topic, and its hard to see how this paper contributed any additional work. The only change I can see is stated below:

In contrast to the definition of coupling used by Tsonis et al. [2007], a clear statistical definition of ‘strong’ and ‘weak’ coupling is possible simply by calculating the coupling using surrogate data generated from an AR-1 process with the same autocorrelation as the observed mode time series.

But the paper does not provide any data, any means of verifying their claims, anything really other than a statement of results. Very dissapointing that this should qualify as a new paper, as nothing seems to have changed, except for the order of the authors.

Another difference is that Swanson and Tsonis do not attempt to forecast temperatues (except for a figure in the RC post), only asserting that their results indicate this current regime will persist for a few decades, with probably flat temperatures. I agree, that an underlying warming could still be occurring, and would be expected to resume when this current regime ends.

However, when an estimate is made of the underlying rate of warming, assuming the temperature increase from 1976 to 1998 was based on the sum of the underlying rate, enhanced by the regime shift, as shown in our paper the underlying warming does not intersect with the current flat temperature regime until around 2050, and only increases by 0.2C by 2100.

Another difference is that while our model is created with empirical fits to data, it is backed up by reference to physical studies of changes in ocean currents, and physically based determinations of climate sensitivity to CO2 by Spencer.

Advertisements

0 thoughts on “Swanson and Tsonis 2009 on Regime-shifts

  1. It’s kinda funny, though…the entire basis of the AR4’s attribution claim was, given the assumption that natural variability in models is realistic, the inability of such models to reproduce variability since 1950 without anthropogenic forcing “proves” that the warming must be due to man. Except that not only the recent shifts but pretty much all model results show that the models do NOT simulate natural variability well.

    RC does seem to see how this undermines the attribution argument at all, they just want to “spin” the issue away from reduced alarm.

  2. It's kinda funny, though…the entire basis of the AR4's attribution claim was, given the assumption that natural variability in models is realistic, the inability of such models to reproduce variability since 1950 without anthropogenic forcing “proves” that the warming must be due to man. Except that not only the recent shifts but pretty much all model results show that the models do NOT simulate natural variability well.RC does seem to see how this undermines the attribution argument at all, they just want to “spin” the issue away from reduced alarm.

  3. Here is my take on David’s ‘step’ paper; in Australia there is no discernible AGW effect on temperature trend during the 20thC; the 1978 step up generates the temperature difference; prior to the break-point the temperature is flat because there was a -ve and +ve PDO which in effect nullified each PDO’s effect on temperature; the 1978 step-up is the entirety of the new +ve PDO; this is not a stored or accumulated heat build up but, imo, a product of the oceanographic effects which occurred in 1976; on this basis the contrary oceanographic effect in 1998 should produce a gradual decline in trend from the plateaued trend from 1998.

    The global temperature history is more complex; there is a gradual trend from 1910 to 1976 of about 0.3C; the trend after the step-up in 1976 is about 0.2C until the 1997 break; after 1997 the trend declines. For what its worth my take on that is the pre-1976 trend is solar; the trend between 1976 and 1997 cannot be natural variation because the PDO effect is in the break so the problem for CO2, which is monotonically increasing, is, if the CO2 post 1976 caused the trend to 1997, after 1997 with the PDO effect also at 1997 there cannot be a Keenlyside type masking of AGW to explain the flat to slightly declining post 1997 trend because the masking effect stopped at 1997; so there is no amplification of AGW between 1976 and 1997 and no depression of AGW post 1997. My best bet for the 1976 to 1997 trend is UHI with some measurement hiccups thrown in and for post 1997 a declining sun.

    The Tsonis paper seems to be channeling the Keenlyside ‘effect’ but the Keenlyside effect doesn’t exist in a step scenario because there is neither masking or amplification; what you see is what you get unlike the “heat is in the ocean” mantra of the government, an expression which is right up there with “the cheque is in the mail”.

    • “We hypothesize that the established pre-1998 trend is the true forced warming signal,” I think this is faulty reasoning from Swanson and not supported by anything, just an assumption. As you say, if post 1998 increase is depressed by -ve PDO, the pre 1998 is enhanced by +ve PDO, the underlying rate of warming must be in between. Cant suck and blow simultaneously.

    • Cohenite – Temperature between 1978 and 1997 is characterized by two things underneath ENSO noise: a step change relating to the 1986/7 El Nino and the effects of Mt. Pinatubo. The step-change drives the trend. I’ve posted links to show this in a comment below.

      And about UHIE, let’s avoid this question all-together and go straight to SST – after all, land and lower tropospheric temps are just a product of SST. Much less problematic variables, especially in modern times.

      • I’m not buying it. The big drop it claims to explain occured in air temperature too. It marked the end to the long-term response to the four-year-long El Nino in the early 40s.

      • I’m not referring to “Thompson, D., et al., 2008. A large discontinuity in the mid-twentieth century in observed global-mean surface temperature. Nature, 453, 646-650, doi:10.1038/nature06982” but to the issue more generally.

        http://www.climateaudit.org/?cat=67

        The issue has the potential to be serious, Thompson et al. aside.

      • Andrew: Sorry for jumping in so late but… How would buckets and inlets issues impact Carl’s comment which covered a period from 1978 to 1997? Much of the SST variability over that period is verified with satellite data that’s been available since 1981.

      • Thanks. I’m often unclear on the methodology for putting SST data together. Can you point me to an exposition of what’s what?

  4. Here is my take on David's 'step' paper; in Australia there is no discernible AGW effect on temperature trend during the 20thC; the 1978 step up generates the temperature difference; prior to the break-point the temperature is flat because there was a -ve and +ve PDO which in effect nullified each PDO’s effect on temperature; the 1978 step-up is the entirety of the new +ve PDO; this is not a stored or accumulated heat build up but, imo, a product of the oceanographic effects which occurred in 1976; on this basis the contrary oceanographic effect in 1998 should produce a gradual decline in trend from the plateaued trend from 1998.The global temperature history is more complex; there is a gradual trend from 1910 to 1976 of about 0.3C; the trend after the step-up in 1976 is about 0.2C until the 1997 break; after 1997 the trend declines. For what its worth my take on that is the pre-1976 trend is solar; the trend between 1976 and 1997 cannot be natural variation because the PDO effect is in the break so the problem for CO2, which is monotonically increasing, is, if the CO2 post 1976 caused the trend to 1997, after 1997 with the PDO effect also at 1997 there cannot be a Keenlyside type masking of AGW to explain the flat to slightly declining post 1997 trend because the masking effect stopped at 1997; so there is no amplification of AGW between 1976 and 1997 and no depression of AGW post 1997. My best bet for the 1976 to 1997 trend is UHI with some measurement hiccups thrown in and for post 1997 a declining sun.The Tsonis paper seems to be channeling the Keenlyside 'effect' but the Keenlyside effect doesn't exist in a step scenario because there is neither masking or amplification; what you see is what you get unlike the “heat is in the ocean” mantra of the government, an expression which is right up there with “the cheque is in the mail”.

  5. Can only caution that the temperature data behind your analysis could be showing instrumental as well as climatic effects. e.g. the mid 1990s saw a swing from mercury thermometry to resistance devices. Apart from the calibration, drift and longevity differences, there were changes to the number of reading made per day and almost certainly different treatments of “spikes” which can be seen on continuous data but not on max/min thermometers. I’d really be surprised if there were no effects from the changeovers. There are indirect suggestions that tapered splicing was used and this might well interact with detection of break points and steps.

    An attribution of effects to PDOs or El Ninos etc runs the risk that CO2 has faced. You have to be sure you can show the quantitative cause/effect link before you rely on it for explanation.

  6. Can only caution that the temperature data behind your analysis could be showing instrumental as well as climatic effects. e.g. the mid 1990s saw a swing from mercury thermometry to resistance devices. Apart from the calibration, drift and longevity differences, there were changes to the number of reading made per day and almost certainly different treatments of “spikes” which can be seen on continuous data but not on max/min thermometers. I'd really be surprised if there were no effects from the changeovers. There are indirect suggestions that tapered splicing was used and this might well interact with detection of break points and steps.An attribution of effects to PDOs or El Ninos etc runs the risk that CO2 has faced. You have to be sure you can show the quantitative cause/effect link before you rely on it for explanation.

  7. I wouldn’t even suggest it of there was not evidence in the ocean that 1997-8 was a shift, as mentioned in our paper.

  8. “We hypothesize that the established pre-1998 trend is the true forced warming signal,” I think this is faulty reasoning from Swanson and not supported by anything, just an assumption. As you say, if post 1998 increase is depressed by -ve PDO, the pre 1998 is enhanced by +ve PDO, the underlying rate of warming must be in between. Cant suck and blow simultaneously.

  9. David –
    Your paper is great!
    This is exactly what I have been looking for.
    I have made many posts on the subject of ENSO-induced step-changes (most notably 1976-8, 1986/7, and 1997/8), but have found no way to statistically present them. Perhaps the reason the Chow test missed the 1986/7 step change is because of the effect of Mt. Pinatubo from 1991-1995. Looking at the SST data, it’s clearly there, perhaps in even greater magnitude than the 1997/8 step change.

    I have made a post in response to Swanson, here:
    http://climatechange1.wordpress.com/2009/07/13/swansons-not-so-novel-post-at-realclimate/

    For an introduction to my thoughts on the topic:
    http://climatechange1.wordpress.com/2009/05/22/ten-questions-for-alarmists-about-the-el-ninosouthern-oscillation/

    • One more thing. Another reason the Chow test may have missed the 1986/7 step change is that ENSO happened to “hide” it more than it hid the 1997/8 step change. ENSO variation just happened to make the step-change look more like a gradual rise. Take out ENSO, and both step-changes are even more clear. A graph showing this is in the first link of my previous post.

    • Carl, It is an area of growing interest. In the paper, I worked on the major patterns, and did not worry too much about multiple breaks that could have picked it up. One reason is that I don’t really trust automated algorithms at that level. Even though it found that one break was optimal in all cases, I would want to do a lot of sensitivity testing if I was going to use it for more detail.

      I just focused on the claims, and trying to address them with a rigorous procedure. No doubt you could use strucchange at the level you require, as it does multiple breaks.

  10. David – Your paper is great! This is exactly what I have been looking for.I have made many posts on the subject of ENSO-induced step-changes (most notably 1976-8, 1986/7, and 1997/8), but have found no way to statistically present them. Perhaps the reason the Chow test missed the 1986/7 step change is because of the effect of Mt. Pinatubo from 1991-1995. Looking at the SST data, it's clearly there, perhaps in even greater magnitude than the 1997/8 step change.I have made a post in response to Swanson, here:http://climatechange1.wordpress.com/2009/07/13/…For an introduction to my thoughts on the topic:http://climatechange1.wordpress.com/2009/05/22/

  11. One more thing. Another reason the Chow test may have missed the 1986/7 step change is that ENSO happened to “hide” it more than it hid the 1997/8 step change. ENSO variation just happened to make the step-change look more like a gradual rise. Take out ENSO, and both step-changes are even more clear. A graph showing this is in the first link of my previous post.

  12. Carl, I is an area of growing interest. In the paper, I worked on the major patterns, and did not worry too much about multiple breaks that could have picked it up. One reason is that I don't really trust automated algorithms at that level. Even though it found that one break was optimal in all cases, I would want to do a lot of sensitivity testing if I was going to use it for more detail. I just focused on the claims, and trying to address them with a rigorous procedure. No doubt you could use strucchange at the level you require, as it does multiple breaks.

  13. “Another difference is that while our model is created with empirical fits to data, it is backed up by reference to physical studies of changes in ocean currents”

    David – as I challenged Coho – still not enough IMVeryHO. All still just stats stuff. You need some accounting of various effects. Maybe even some “modelling” – choke choke – to see how all this adds up and has a mechanistic as well as statistical basis.

    So 2 challenges:

    Get Chris Folland (Prof IPO) to give you some critique.

    Get in published in GRL.

    No excuses.

  14. “Another difference is that while our model is created with empirical fits to data, it is backed up by reference to physical studies of changes in ocean currents”David – as I challenged Coho – still not enough IMVeryHO. All still just stats stuff. You need some accounting of various effects. Maybe even some “modelling” – choke choke – to see how all this adds up and has a mechanistic as well as statistical basis.So 2 challenges:Get Chris Folland (Prof IPO) to give you some critique.Get in published in GRL.No excuses.

  15. Cohenite – Temperature between 1978 and 1997 is characterized by two things underneath ENSO noise: a step change relating to the 1986/7 El Nino and the effects of Mt. Pinatubo. The step-change drives the trend. I've posted links to show this in a comment below.And about UHIE, let's avoid this question all-together and go straight to SST – after all, land and lower tropospheric temps are just a product of SST. Much less problematic variables, especially in modern times.

  16. David,

    I read the earlier Tsonis et al 2007 as it was cited in the Lindzen 2007 paper I was interested in a while back. Lindzen interpreted it thus, “A very recent paper (Tsonis et al., 2007) suggests, in fact, that the surface temperature record can be accounted for by essentially superpositions of known oceanic fluctuations such as the Pacific Decadal Oscillations and the Atlantic Multidecadal Oscillations.”

    So it came as a huge surprise to me to finally see RealClimate dare to go anywhere near something as heretical as this Swanson & Tsonis paper. But after reading Swanson’s post I see that the whole thing looks like an elaborate exercise in apologetics, or perhaps a forced recantation.

    It seems to be saying, “Look you can read our paper — and believe the alarmists — if you want to! Here’s how…”

    I made some comments at RealClimate about that trend from 1976-1998 that they said “wasn’t cherry-picking”. How can an assumption of a single linear trend beginning 1850 and fitting through 1976-1998 be anything other than cherry-picking, or wishful thinking, rather?

    • Yeah, as you can see from Roger’s post (see my comment above) RC is “spinning” (Roger’s words, not mine, but I wouldn’t disagree a bit) these results.

      An alternative spin:
      http://www.worldclimatereport.com/index.php/2009/04/08/has-the-climate-recently-shifted/

      Bottom line: it is pretty easy in this business to torture results until they confess to one thing or another. But personally I think WCR’s spin is more plausible than RC’s spin. What’s shocking is that Swanson actually got spin into the paper itself!

      I mean, in a rational world, what respectable scientific journal would publish a paper with a statement like:

      “there is no comfort to be gained by having a climate with a significant degree of internal variability, even if it results in a near-term cessation of global warming.”

      Apparently the otherwise highly respectable Geophysical Research Letters

      Weep for science friends. Weep.

  17. David, I read the earlier Tsonis et al 2007 as it was cited in the Lindzen 2007 paper I was interested in a while back. Lindzen interpreted it thus, “A very recent paper (Tsonis et al., 2007) suggests, in fact, that the surface temperature record can be accounted for by essentially superpositions of known oceanic fluctuations such as the Pacific Decadal Oscillations and the Atlantic Multidecadal Oscillations.” So it came as a huge surprise to me to finally see RealClimate dare to go anywhere near something as heretical as this Swanson & Tsonis paper. But after reading Swanson's post I see that the whole thing looks like an elaborate exercise in apologetics, or perhaps a forced recantation.It seems to be saying, “Look you can read our paper — and believe the alarmists — if you want to! Here's how…”I made some comments at RealClimate about that trend from 1976-1998 that they said “wasn't cherry-picking”. How can an assumption of a single linear trend beginning 1850 and fitting through 1976-1998 be anything other than cherry-picking, or wishful thinking, rather?

      • My apologies. I have just posted it and it is still awaiting moderation (hopefully they will let it through. It is a bit of a lottery with RC). I have included the text below:

        “149.
        Richard Steckis says:
        Your comment is awaiting moderation.
        15 July 2009 at 12:33 AM

        You identified a breakpoint in the dataset at 1997. I analysed the HadCRUt3 simplified dataset (monthly data http://hadobs.metoffice.com/crutem3/diagnostics/global/simple_average/monthly) and identified three breakpoints in the dataset one at 1977, one at 1986 and the breakpoint at 1997. This was using the strucchange package in R. The 1977 breakpoint can be explained by a paper by McGuirk (Planetary-Scale Forcing of the January 1977 Weather, Science 199:4326 293-295 1978). The 1986 and 1997 breakpoints both coincide with El Nino events.

        My hypothesis is that each of these are cumulative adding one on the other. Much of the observed warming from 1977 through to now may be a result of this cumulative effect where the next event caused more heat to be added to the system before the heat from the previous event radiatively dissipated.
        150.
        Richard Steckis says:
        Your comment is awaiting moderation.
        15 July 2009 at 12:52 AM

        With regard to my analyses described above. I used the HadCRUt3 data from 1950 to present. This is the same period used the first graph.”

        Regards

  18. Yeah, as you can see from Roger's post (see my comment above) RC is “spinning” (Roger's words, not mine, but I wouldn't disagree a bit) these results.An alternative spin:http://www.worldclimatereport.com/index.php/200…Bottom line: it is pretty easy in this business to torture results until they confess to one thing or another. But personally I think WCR's spin is more plausible than RC's spin. What's shocking is that Swanson actually got spin into the paper itself!I mean, in a rational world, what respectable scientific journal would publish a paper with a statement like:”there is no comfort to be gained by having a climate with a significant degree of internal variability, even if it results in a near-term cessation of global warming.”Apparently the otherwise highly respectable Geophysical Research LettersWeep for science friends. Weep.

  19. My apologies. I have just posted it and it is still awaiting moderation (hopefully they will let it through. It is a bit of a lottery with RC). I have included the text below: “149. Richard Steckis says: Your comment is awaiting moderation. 15 July 2009 at 12:33 AM You identified a breakpoint in the dataset at 1997. I analysed the HadCRUt3 simplified dataset (monthly data http://hadobs.metoffice.com/crutem3/diagnostics…) and identified three breakpoints in the dataset one at 1977, one at 1986 and the breakpoint at 1997. This was using the strucchange package in R. The 1977 breakpoint can be explained by a paper by McGuirk (Planetary-Scale Forcing of the January 1977 Weather, Science 199:4326 293-295 1978). The 1986 and 1997 breakpoints both coincide with El Nino events. My hypothesis is that each of these are cumulative adding one on the other. Much of the observed warming from 1977 through to now may be a result of this cumulative effect where the next event caused more heat to be added to the system before the heat from the previous event radiatively dissipated. 150. Richard Steckis says: Your comment is awaiting moderation. 15 July 2009 at 12:52 AM With regard to my analyses described above. I used the HadCRUt3 data from 1950 to present. This is the same period used the first graph.”Regards

  20. Bob Tisdale has looked at this interesting issue of whether ENSO produces an accumulation of heat;

    http://bobtisdale.blogspot.com/2008/04/is-there-cumulative-enso-climate.html

    An accumulation of heat effect does not preclude a step but tends to mitigate the step’s primacy or even exclusivity as a determinant of temperature level, as opposed to trend; structural breaks are not trends but a hiatus in one temperature level which is replaced by another; the steps achieve this transition fairly rapidly so this also vitiates the applicability of a trend to the temperature. Given this 2 conclusions seem to follow; firstly the step does not accumulate or even cause the increase in temperature but rather is an indice of a redistribution of the existing heat in the system; secondly, if there is a trend in the temperature caused by an accumulative factor such as CO2 levels then the periods between the steps or structural breaks should be examined for trends and correlations with the nominated cause; David’s paper shows no such trends in Australia, while globally the reverse side to the Keenlyside principle of AGW masking, that is amplification, does not seem to be present thus removing AGW as a potential cause for the trends between the steps in the global temperature data.

    • cohenite,

      I do not think these structural breaks (as determined by the Chow test) represent step functions, so much as changes in power and slope of a trend in a time series.

      The accumulation of heat may have nothing to do with CO2 and probably doesn’t. It is possible that rapid succession events may such as the ones I described may power warming episodes such as those of the Roman Climate Optimum and the Medieval Warm Period.

      • Richard, the breaks were step functions in the case of Australia, because the slopes tested not significant outside the break. I’d like to know how common steps are in regions, rather than trends. Its possible that global trends are composed of a lot of regional steps, at different times, as you suggest.

        You have to be careful not to read too much into this, as the test is also sensitive to choice of start and end point, scale. For example, breaks will look more significant near the centre, than at the end. Have to really check out each pattern.

      • Thanks. I will look into this a bit further. I am somewhat new to the Chow test and so will take you advice on board.

  21. Bob Tisdale has looked at this interesting issue of whether ENSO produces an accumulation of heat;http://bobtisdale.blogspot.com/2008/04/is-there…An accumulation of heat effect does not preclude a step but tends to mitigate the step's primacy or even exclusivity as a determinant of temperature level, as opposed to trend; structural breaks are not trends but a hiatus in one temperature level which is replaced by another; the steps achieve this transition fairly rapidly so this also vitiates the applicability of a trend to the temperature. Given this 2 conclusions seem to follow; firstly the step does not accumulate or even cause the increase in temperature but rather is an indice of a redistribution of the existing heat in the system; secondly, if there is a trend in the temperature caused by an accumulative factor such as CO2 levels then the periods between the steps or structural breaks should be examined for trends and correlations with the nominated cause; David's paper shows no such trends in Australia, while globally the reverse side to the Keenlyside principle of AGW masking, that is amplification, does not seem to be present thus removing AGW as a potential cause for the trends between the steps in the global temperature data.

  22. cohenite,I do not think these structural breaks (as determined by the Chow test) represent step functions, so much as changes in power and slope of a trend in a time series. The accumulation of heat may have nothing to do with CO2 and probably doesn't. It is possible that rapid succession events may such as the ones I described may power warming episodes such as those of the Roman Climate Optimum and the Medieval Warm Period.

  23. Hi Richard; I’m no statistician but I understand the Chow test determines whether there is a trend in a series of data or whether that nominal data series has distinct parts so that there is not a trend over the whole series. The Australian results show that there is no trend before or after the 1978 break or step; whatever “power and slope” there is in the data occurs in 1978 and in fact seperates the series into 2 halves. The global data is more complex and I have given my take on that above.

    • Thanks for your input. I too am not a statistician (I am a fishery biologist) so am still finding my way with this stuff.

  24. Hi Richard; I'm no statistician but I understand the Chow test determines whether there is a trend in a series of data or whether that nominal data series has distinct parts so that there is not a trend over the whole series. The Australian results show that there is no trend before or after the 1978 break or step; whatever “power and slope” there is in the data occurs in 1978 and in fact seperates the series into 2 halves. The global data is more complex and I have given my take on that above.

  25. Richard, the breaks were step functions in the case of Australia, because the slopes tested not significant outside the break. I'd like to know how common steps are in regions, rather than trends. Its possible that global trends are composed of a lot of regional steps, at different times, as you suggest. You have to be careful not to read too much into this, as the test is also sensitive to choice of start and end point, scale. For example, breaks will look more significant near the centre, than at the end. Have to really check out each pattern.

  26. Thanks. I will look into this a bit further. I am somewhat new to the Chow test and so will take you advice on board.

  27. Thanks for your input. I too am not a statistician (I am a fishery biologist) so am still finding my way with this stuff.

  28. I'm not buying it. The big drop it claims to explain occured in air temperature too. It marked the end to the long-term response to the four-year-long El Nino in the early 40s.

  29. I'm not referring to “Thompson, D., et al., 2008. A large discontinuity in the mid-twentieth century in observed global-mean surface temperature. Nature, 453, 646-650, doi:10.1038/nature06982” but to the issue more generally.http://www.climateaudit.org/?cat=67The issue has the potential to be serious, Thompson et al. aside.

  30. It’s bad form to refer to one’s work, but from a limited number of Australian station plots of Tmax and Tmin (reported elsewhere on Niche Modeling a few months ago), there is a feature which I’m seeing more and more as I do more plots in detail, I’m taking daily data and getting the feel for it, at about 40 stations for up to 130 years a time. The strange feature is that many stations show a “necking” of values about 1994 +/- 2 years. The max and mean move closer together and to the linear 40-year average, for about 5 years in a row. Then they blow out again to the earlier spread and pattern.I’m wondering if this textural effect should be examined to see if it qualifies as a different type of break point. I have no explanation for it unless it’s a consequence of changing from thermometry to thermocouples and more frequent readings each day.

    Davids: Changes in variance are big in finance (aka volatility). I am interested in the 1930-80 period that seems to be flat but higher volatility. Seems like measurement factors could be important though.

  31. It's bad form to refer to one's work, but from a limited number of Australian station plots of Tmax and Tmin (reported elsewhere on Niche Modeling a few months ago), there is a feature which I'm seeing more and more as I do more plots in detail, I'm taking daily data and getting the feel for it, at about 40 stations for up to 130 years a time. The strange feature is that many stations show a “necking” of values about 1994 +/- 2 years. The max and mean move closer together and to the linear 40-year average, for about 5 years in a row. Then they blow out again to the earlier spread and pattern.I'm wondering if this textural effect should be examined to see if it qualifies as a different type of break point. I have no explanation for it unless it's a consequence of changing from thermometry to thermocouples and more frequent readings each day.Davids: Changes in variance are big in finance (aka volatility). I am interested in the 1930-80 period that seems to be flat but higher volatility. Seems like measurement factors could be important though.

  32. Andrew: Sorry for jumping in so late but… How would buckets and inlets issues impact Carl's comment which covered a period from 1978 to 1997? Much of the SST variability over that period is verified with satellite data that's been available since 1981.

  33. Bob, thanks for drawing attention to that post. Your last figure -riveting.<img src=”http://i42.tinypic.com/iom6ab.jpg”>

  34. Thanks Bob, I had looked at your reemergence thread before; as usual, very interesting! You say that a predominance of El Nino, or at least a series of El Nino which are not counterbalnced by La Ninas of equal or greater power will gradually increase temperature over the duration of the period of El Nino dominance; wouldn’t it also be fair to say that the effect of such a period of dominance on temperatures could be an initial break and then the maintainenece of the new plateau created by the break up in temperature at the beginning when the La Nina dominated period ends and the El Nino dominated period begins as happened in 1976-78?

    It seems to me that depending on the nominated period if there is a balance between the disparate phases then temperature trend will be flat or subject to other exogenous factors; any period where there is a dominance of say El Nino will have a break and plateaus either side; any AGW effect has to be in the slopes either side of the break[s]; this appears to be what David’s paper suggests, imo.

    • cohenite: You asked, “…wouldn’t it also be fair to say that the effect of such a period of dominance on temperatures could be an initial break and then the maintainenece of the new plateau created by the break up in temperature at the beginning when the La Nina dominated period ends and the El Nino dominated period begins as happened in 1976-78?”

      Could you rephrase the question?

  35. Thanks Bob, I had looked at your reemergence thread before; as usual, very interesting! You say that a predominance of El Nino, or at least a series of El Nino which are not counterbalnced by La Ninas of equal or greater power will gradually increase temperature over the duration of the period of El Nino dominance; wouldn't it also be fair to say that the effect of such a period of dominance on temperatures could be an initial break and then the maintainenece of the new plateau created by the break up in temperature at the beginning when the La Nina dominated period ends and the El Nino dominated period begins as happened in 1976-78?It seems to me that depending on the nominated period if there is a balance between the disparate phases then temperature trend will be flat or subject to other exogenous factors; any period where there is a dominance of say El Nino will have a break and plateaus either side; any AGW effect has to be in the slopes either side of the break[s]; this appears to be what David's paper suggests, imo.

  36. cohenite: You asked, “…wouldn't it also be fair to say that the effect of such a period of dominance on temperatures could be an initial break and then the maintainenece of the new plateau created by the break up in temperature at the beginning when the La Nina dominated period ends and the El Nino dominated period begins as happened in 1976-78?”Could you rephrase the question?

  37. Bob, I guess Figure 1(a) from David’s paper sums up what I’m trying to say; the flat period after the break indicates no build or upward trend after the break; at least in Australia; the global situation in Figure2 is slightly different.

  38. Bob, I guess Figure 1(a) from David's paper sums up what I'm trying to say; the flat period after the break indicates no build or upward trend after the break; at least in Australia; the global situation in Figure2 is slightly different.

  39. Bob, I guess Figure 1(a) from David's paper sums up what I'm trying to say; the flat period after the break indicates no build or upward trend after the break; at least in Australia; the global situation in Figure2 is slightly different.

  40. Pingback: www.videogaming24.co.uk

  41. Pingback: bateria do laptopa dell

  42. Pingback: witryna

  43. Pingback: witryna firmowa

  44. Pingback: montaz sufitów armstrong

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s