Prev  1  2  3  4  5  

Comments 101 to 123 out of 123:

101. KR at 09:02 AM on 15 May, 2012
     SRJ - Got it, my mistake on what you were discussing in terms of autocorrelation. While I still feel that varying trend lengths are of some use, if properly considering uncertainties, I would agree that looking at equal spans provides rather more information.
102. skywatcher at 09:38 AM on 15 May, 2012
     SRJ - I quite agree on the general point that for a more detailed analysis, using trends of comparable length is preferable, but this is only relevant for detailed analysis, which the IPCC FAQ figure was not. Your numbers in #100 neatly illustrate the point with consistent trend lengths. The reason I produced the figures in #76 was that, as KR says in #99, the IPCC figure was merely illustrating a point. Not proving some part of atmospheric physics, or anything like that. Helena repeatedly implied that there was a kind of cherry-picking going on with the choice of the four trends by the IPCC - my figure neatly demonstrated that this was not the case, given the more-or-less progressively increasing trend rates to the present. Most other choices of four trends spaced across the profile would have given the same result, even if a few choices would have produced slightly different patterns (relevant to Dikrans points about forcing and trends). The patterns produced in both graphs in #76 are indicative of a temperature profile that is broadly concave, rather than convex or flat. I think that general point is mathematically defensible, as with Tom Curtis' sketches. Mathematically, whether you start in 1850 or in 2005, both will show increasing gradients for a concave profile, but one will have rapid increase near the end (2005 fixed point), and one will have slower increase at the end (1850 fixed). But both will increase. There are, statistically, much better ways to support the overall acceleration in temperature (or just look at any of the 'hockey stick' graphs!), but these methods would not be appropriate for an illustrative FAQ answer. Helena #92, a "decreasing, decelerating" profile would have negative gradients. So your point was? There are four possible curve solutions - draw the four possible curves, like the two in Tom Curtis' #77, one has increasing positive gradients, one has increasing (steepening) negative gradients, a third has decreasing positive gradients, and the last has decreasing (flattening) negative gradients. All would show up differently in my figures in #76.
103. Tom Curtis at 09:48 AM on 15 May, 2012
     SRJ @86, the question under discussion here is not, "What is the best method to analyze temperature trends over the last 150 years?" The IPCC's analysis did not consist of just the illustration under discussion, which as Dikran Marsupial has pointed out, only appears in summaries of the data, not in direct analyses of the data. In FAQ 3.1 where a similar figure to that above appears, the IPCC says:

     "Expressed as a global average, surface temperatures have increased by about 0.74°C over the past hundred years (between 1906 and 2005; see Figure 1). However, the warming has been neither steady nor the same in different seasons or in different locations. There was not much overall change from 1850 to about 1915, aside from ups and downs associated with natural variability but which may have also partly arisen from poor sampling. An increase (0.35°C) occurred in the global average temperature from the 1910s to the 1940s, followed by a slight cooling (0.1°C), and then a rapid warming (0.55°C) up to the end of 2006 (Figure 1). The warmest years of the series are 1998 and 2005 (which are statistically indistinguishable), and 11 of the 12 warmest years have occurred in the last 12 years (1995 to 2006). Warming, particularly since the 1970s, has generally been greater over land than over the oceans. Seasonally, warming has been slightly greater in the winter hemisphere. Additional warming occurs in cities and urban areas (often referred to as the urban heat island effect), but is confined in spatial extent, and its effects are allowed for both by excluding as many of the affected sites as possible from the global temperature data and by increasing the error range (the light grey band in the figure)."

     Clearly, therefore, the IPCC makes people aware of the high trends in Global Mean Surface Temperature (GMST) in the middle of the 20th century. These high trends correlate with high trends in total Top of Atmosphere (TOA) forcing, although the forcings do not completely explain it. Because of that correlation, it would be an embarrassment to the IPCC if there where not high trends in GMST in the middle of the 20th century. As the IPCC makes people aware of the mid 20th century trends, their purpose in the diagram we are discussing is not to conceal them. Rather, it is to illustrate the general tendency to higher trends as the century progresses. If that illustration is not to be misleading, then an inductive inference from the pattern shown to the acceleration of the warming over the period in question must be justified. That is what has been questioned by Monckton and by Helena. So, while I agree that the additive model is better than the IPCC's chart, that is not relevant to this discussion. Further, while I agree that for most purposes comparing trends of different length is unwise, that is also not relevant to this discussion. What is relevant is the question as to whether or not, if you do compare trends as the IPCC has done, you can validly make an inductive inference that the warming has accelerated. Please note that simply comparing the peak 25 year trend in different periods does not answer that question. Consequently, for this purpose your first chart is worse than the IPCC's. That is because a high trend ending in a particular year may be followed by a low trend shortly thereafter. Consquently you can have more total warming with a lower peak trend which is reached year after year than with a higher peak trend that is reached and simply falls away. Another way of putting that is that that you get greater warming with more robust trends, and hence it is the robust trends which better indicate increased warming. Finally, the attack on the chart above by Monckton is not based on statistical qualms. Rather it is based on a desire to attack the IPCC to undermine action on global warming. You can be sure that had the IPCC used the additive model shown by you, it would have come under equally virulent, and unwarranted assault. If you do not believe this, just consider the continuing unjustified assaults against the validity of the temperature record itself.
104. Tom Curtis at 10:15 AM on 15 May, 2012
     Two points for Helena: 1) I believe I have finally figured out what you mean be a decreasing decelerating curve. It is a curve which below the x-axis, ie, negative, but in which it approaches the x-axis over time such that curve satisfies the condition that slope(t) < slope(t+a) where a is greater than 0. That, however, is mathematically just an accelerating curve. Over time, the slope of the curve gets larger. The situation is described as - it is cold, but it is getting warmer, and over time it is getting warmer faster. So, yes, the IPCC technique will detect acceleration in this case - and so it should, for the situation is one of accelerated warming. If I have misunderstood your phrase (which is as clear as mud), specify what you mean mathematically, or plot the curve on a graph and show it to us. 2 Helena @93 said:

     "In the same way, when you start in the present (his graph1 and the IPCC), you're not being "fair" to the past as its trend gets mechanically diluted. You "force-flaten" long trends. That's why you can't compare trends that have different T. "

     This theme that increasing the length of the trend will always flatten it is a constant theme of Helena's, and (so far as I can tell) the core of her argument. Indeed, SFAICT, apart from a few failed counterexamples and constantly reiterating that she disagrees, she has no other argument. And it is simply false. Consider the function such that If x < or = 0, f(x)= 0; else f(x)=x^2. We can consider two series of progressively longer trends. In the first series the trends are all centered on 1. In this case, each successive trend will have a higher slope than the previous trend. That directly contradicts Helena's contention. In the second series, the trends all terminate at 25. In this case, each successively larger trend will have a lower trend than the preceding one. This is the case that parallels the IPCC diagram. Helena's central assumption is in fact false of any data with an accelerating slope (and in some other cases as well). By sticking rigourously to her "intuitive statistics", Helena simply starts with an initial premise that no data can show acceleration.
105. Tom Curtis at 10:25 AM on 15 May, 2012
     In my concluding paragraph of 104, I should have said her assumption is false of any data with an accelerating slope and a signal to noise ratio sufficient to allow determination of the underlying signal. If Helena has wanted to argue that the technique illustrated by the IPCC can work in some cases, but that in the particular case used by the IPCC, the noise is to great for it to be valid, she may have had a point. There are relatively simple mathematical methods for determining that, and I believe they would show that she was incorrect. But you do actually have to look at the data to know if she would have been incorrect or not, if that had been her contention. Instead, Helena has argued that the technique cannot work under any circumstances, even as a statistical inference. She thereby demonstrates that she does not even understand the basic maths of the situation.
106. Tom Curtis at 10:33 AM on 15 May, 2012
     Skywatcher @102 writes:

     "The patterns produced in both graphs in #76 are indicative of a temperature profile that is broadly concave, rather than convex or flat. I think that general point is mathematically defensible, as with Tom Curtis' sketches. Mathematically, whether you start in 1850 or in 2005, both will show increasing gradients for a concave profile, but one will have rapid increase near the end (2005 fixed point), and one will have slower increase at the end (1850 fixed). But both will increase."

     I have probably misunderstood what you said, however, with an accelerating (concave upward) curve a sequence of progressively longer trends with a common start point would each have a progressively greater trend, ie, the longer trends would be larger. In contrast, sequence of progressively longer trends with a common end point would have progressively lower trends, ie, the longer trends would be lower.
107. skywatcher at 10:51 AM on 15 May, 2012
     Tom #106, I see your point - the clarifying statement is this - whether you fix the start or end point, always read the trend changes forwards in time. So fixing the 1850 point begins with short trends and ends with longer ones, the later trends being larger than the earlier (2nd fig in #76). Fixing the 2005 point (as done by IPCC) begins with longer trends, that get progressively shorter through time (1st fig in #76), and in both cases the positive trend rises over time. That feature is unique to the rising, accelerating curve, and not present in a falling or decelerating curve. The rate of trend rise is lower if you fix the 1850 point than if you fix the 2005 point, due to the increasing amount of data involved in the longer trend (so the acceleration is less pronounced for the 1850 fix), but the trend still rises.
108. Tom Curtis at 11:08 AM on 15 May, 2012
     skywatcher @107, thanks for the clarification. We should note that it is still preferable to used the end point method rather than the start point method in this case. That is because the HadCRUT3v temperature index typically covered around 20% of the globe in the 1850's, compared to around 80% in the 2000's. The result is that the temperature record has a much lower signal to noise ratio early in the record, particularly prior to 1880 which would defeat the technique. (It may show the correct pattern, I have not checked. But it would be illegitimate to draw any inference from it.)
109. skywatcher at 13:30 PM on 15 May, 2012
     Tom #108 - agreed , it's prefereable to use end point method due to signal/noise issues early in the series. This is seen in my second graph in #78, the early noise is what causes the trend to vary substantially for trends between 1875 and 1920. After that, you have the progressive rise, which tells you nothing more clever than that the overall profile is slightly convex, but it is weak at inferring anything much else. Helena, just saying 'no' will never make you right. Tom's statement is quite correct, as demonstrated multiple times upthread. You have failed to provide any examples, real or conceptual, that demonstrate otherwise.
110. Nick Stokes at 16:59 PM on 15 May, 2012
     The Moyhu trend viewer helps here, though it only goes back to 1901. You can click buttons for the dataset and period you want - say Hadcrut from 1901 to present. In the triangle plot, you can see the trends to present along the right hand axis. They do show reddening colours going up untli about 20 years duration; shorter term trends to present are quite noisy and can be negative. If you want to see trends starting 1901, look along the x-axis, Again after some initial noise, there's quite a high trend reflecting the 1900-1940 period, then a dip, and then a rise to the century duration. You can click anywhere to see numerical values, with significance and a graph, and you can click the red square buttons to show significance. And if you want to cherry pick, it's helpful there too.
111. les at 20:54 PM on 15 May, 2012
     95 - SJR "I think that Skywatchers contribution suffers from the same problem as the IPCC graph, namely comparing trends over periods of different length." Really your post 100 and KR's post 99 answers this. There is not problem - despite the psudo-science from Helena 92 - to have identical periods so long as the uncertainty is treated properly. as 99 shows, the shorter period has higher uncertainness - however the effect is still substantially greater so there's no problem... skywatcher 76's graphs could have +/- bars or lines which would widen as they reached the early/late fixed point... if one doesn't explicitly calculate the error-bars, you are more or less forced to 'fix' the errors by having constant periods as per 100.
112. muoncounter at 23:14 PM on 15 May, 2012
     It is interesting that the word 'accelerated' is not present in the WG1 discussion of this graph. They merely state the obvious: The rate of warming averaged over the last 50 years (0.13°C ± 0.03°C per decade) is nearly twice that for the last 100 years. If the underlying temperature-time function is linear with random noise, trends calculated from any starting point and over any time period long enough to deal with the noise are all very close. This is distinctly not the appearance of the IPCC graph, clearly indicating that there is some curvature to the function. In this case, the curvature is up.
113. Helena at 05:39 AM on 16 May, 2012
     It seems that first and second order draft of the IPCC report didn't show the published graph, but this one instead : http://pds.lib.harvard.edu/pds/view/7768990?n=383&printThumbnails=no I cannot find any critics to it, but it takes a while to go through all the comments. I sure wouldn't criticize it. Does this mean that the published graph did not go through the full review process ? Did it go through any ?
     Response: [RH] Fixed typo.
114. Tom Curtis at 08:54 AM on 16 May, 2012
     Helena @113, while interesting that the second draft of the IPCC AR4 included a simpler graph than that included in the final product, your concluding question is simply fatuous. Any revision to the first or later drafts of IPCC reports, by there nature, have not gone full the "full review process". The only way to ensure that everything in the IPCC reports has gone through the full process is to avoid any revisions, so that the final report is identical to the first draft. The more interesting point is, was the graph a product of the full review process, and quite plainly it was. This attempt by you to beat up a fake controversy over the graph shows to my mind that your interest here is only in generating controversy, with the purpose of obscuring the obvious. Your comments are the result of deliberate trolling rather than genuine inquiries or differences of opinion. I suggest readers bear that fact in mind when reading any of your future posts.
115. skywatcher at 09:02 AM on 16 May, 2012
     Agreed Tom, Helena's post smacks of rather desperate trolling, as clearly she has no supportable quantitative substantive issue with the published graph, that was in an FAQ and is correspondingly uncomplicated. Recent warming is faster than earlier warming.
116. Tom Curtis at 10:41 AM on 16 May, 2012
     skywatcher @115, while I agree that the graph was intended to summarize information in a simpler form, it was also in the Technical Summary of the IPCC report. Indeed, the version above is the version from the technical summary, which differs from that in the FAQ in the relative placement of the maps and line graph. Further, the FAQ is still intended to be scientifically accurate, with FAQ's being referenced by other parts of the report. Far more importantly, the IPCC explicitly discusses the rapid warming in the early to mid 20th century in the report. There is no question of their attempting to hide relevant information here. So the question of the appropriateness of the graph realy comes down to two issues - is the "inference" from the graph of accelerated warming justified; and is the graph simple enough to be understood by its intended audience of politicians and policy makers. Some of the utterances of politicians and policy makers give me reason to doubt that any graph could be simple enough for them to understand, but there is no doubt that the "inference", ie, the point it was designed to illustrate, is justified. It is interesting to note that Monckton has given us an example of what he thinks the graph should have looked like: This graph he calls "the unvarnished truth". Careful examination shows otherwise. To start, the three clearly marked lines, which are given as being "equal slopes" in the legend, are not equal slopes, with the 1910-1940 line being slightly steeper than the 1975-1998 line. In contrast, the actual trend for 1975-1998 is greater than that for 1910-1940 (see table below). What is worse, the duration (length on the x-axis) of the lines do not correspond to their stated duration, with the 21 year 1860-1880 line having the same width as the 31 year 1910-1940 and the 1975-1998 interval. Further, although the 1975-1998 trend is only just greater than the 1910-1940 trend, the 1975-2005 trend, ie, the trend corresponding in duration to the actual "trend line" shown by Monckton is 0.188 C per decade, a difference from the 1910-1940 trend if Monckton had drawn in the actual trend lines. Being fair to Monckton, he does not call those bold pink lines "trend lines". But if they are not trend lines, then they can have no legitimate purpose on the graph. Their purpose is solely to deceive the eye, which is notoriously bad at estimating trends. By giving the eye a bold target, Monckton seeks to exaggerate the similarity between the three warming periods, both as to duration and slope. No doubt, as a fair minded person, Helena has noticed these "errors" by Monckton, and can point us to her public criticism of them? Or is she so busy trying to reheat false "skeptic" talking points that she has no time to criticize the genuine errors of fake "skeptics". HadCRUT3v +/- HadCRUT4v +/- 1860-1880 0.105 0.159 0.109 0.16 1910-1940 0.153 0.056 0.135 0.056 1975-1998 0.156 0.08 0.071 0.077 1975-2006 0.189 0.052 0.195 0.05 1850-2006 0.042 0.007 0.042 0.008
117. skywatcher at 11:37 AM on 16 May, 2012
     Tom, points taken re FAQ, my bad! Not that I have ever suggested that the graph is anything but scientifically accurate of course. Monckton's chart, complete with its egregious line painting (the "1860-1880" line is particularly misleading as he's drawn a line from ~1852 to ~1885) ... is it the "anti-Escalator"?? According to Monckton's chart, global warming is almost always happening very rapidly (more rapidly than the long-term mean), and periods of apparent hiatus are mere illusions. We should be even more worried! Monckton is an alarmist!! Does Helena think Monckton's graph is better than the one in the IPCC report?
118. Bob Lacatena at 22:59 PM on 16 May, 2012
     This alone should be clear evidence of Monckton's ability to accurately present data. Compare his hand-drawn graph in Tom's comment to these... using the exact same time frames, with both the HADCRUT and BEST data. It should be noted that the GISTEMP data cannot be used, because they don't go back prior to 1880, deeming the data before 1880 to be to unreliable to use. Here is the same graph with one more trend added, for 1979 to the present... [Click on each to go to the original woodfortrees.org plot] Honestly, how can anyone trust his interpretation of the data when it is as sloppy as this?
119. Helena at 03:49 AM on 17 May, 2012
     Tom : "The more interesting point is, was the graph a product of the full review process, and quite plainly it was." I cannot find any review comment asking for any kind of change to the original graph. I obviously cannot prove that i cannot find any, as it is impossible to prove a negative. Can you therefore provide me with any review comment(s) that would suggest that the final graph is "the final product of the full review process" ? Thanks.
120. skywatcher at 08:18 AM on 17 May, 2012
     Helena is that the best you can do? Nothing quantitative, no analysis, and a request that others should go away and do work for you? I think, from this, we can safely say that Helena has nothing to add to this discussion.
121. Tom Curtis at 08:41 AM on 17 May, 2012
     Helena @119, as the graph was not present in the first and second drafts, it was obviously added in response to comments on the second draft (see section 4.2.5 of the IPCC procedures). As such revisions are explicitly described in the procedures, such revisions are part of the the full IPCC review process. Following that revision, the final draft is voted on and adopted by the panel (section 4.4). For the graph to not be a product of the full review process, it would need to have been added after adoption by the panel, a claim you have no evidence for. As it stands, a brief skim of early pages of the second round of review shows at least two reviewer comments that may have suggested to the lead authors the need for the revision. In particular, the reviewer for the US Gov suggested "more discussion of better characterized shorter period trends" (3.34), and the reviewer for the Australian government suggests that using simple linear trends obscures important details. (3-11) While neither reviewer requests the specific modification made, the addition of the shorter trend lines to the graph would certainly help satisfy each reviewers stated concern. Regardless, the issue here is whether or not an unjustified inference is made from the chart, and as clearly and repeatedly demonstrated above, it has not been. In contrast, Monckton not only makes unjustified inferences from his alternative chart, but grossly distorts the data in the chart to begin with.
122. Helena at 13:11 PM on 17 May, 2012
     (-Snip-)
     Response:

     [DB] Moderation complaints snipped. Please construct comments to adhere to this site's Comments Policy.

     Please note that simply disagreeing without then providing a foundation based in the science/literature is construed as sloganeering. As you have consistently been doing this it also constitutes excessive repetition.

     If you have something of substance to support your position then your are welcome to provide it. However, as it stands your position has been completely refuted on this thread.

123. SRJ at 05:51 AM on 9 June, 2012
     # 86 SRJ and the discussion between me and Dikran in # 90, #95 and # 96 I should have mentioned that the graph of the GAM model shows 99% significance, marked with blue. Using the usual 95%, the GAM is significant until 2006.

Prev  1  2  3  4  5  

Post a Comment

Political, off-topic or ad hominem comments will be deleted. Comments Policy...

You need to be logged in to post a comment. Login via the left margin or if you're new, register here.