Climate Science Glossary

Term Lookup

Enter a term in the search box to find its definition.

Settings

Use the controls in the far right panel to increase or decrease the number of terms automatically displayed (or to completely turn that feature off).

Term Lookup

Settings


All IPCC definitions taken from Climate Change 2007: The Physical Science Basis. Working Group I Contribution to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Annex I, Glossary, pp. 941-954. Cambridge University Press.

Home Arguments Software Resources Comments The Consensus Project Translations About Donate

Twitter Facebook YouTube Pinterest

RSS Posts RSS Comments Email Subscribe


Climate's changed before
It's the sun
It's not bad
There is no consensus
It's cooling
Models are unreliable
Temp record is unreliable
Animals and plants can adapt
It hasn't warmed since 1998
Antarctica is gaining ice
View All Arguments...



Username
Password
Keep me logged in
New? Register here
Forgot your password?

Latest Posts

Archives

Climate Hustle

How much is sea level rising?

What the science says...

Select a level... Basic Intermediate

A variety of different measurements find steadily rising sea levels over the past century.

Climate Myth...

Sea level rise is exaggerated
 

"We are told sea level is rising and will soon swamp all of our cities. Everybody knows that the Pacific island of Tuvalu is sinking. ...

 

Around 1990 it became obvious the local tide-gauge did not agree - there was no evidence of 'sinking.' So scientists at Flinders University, Adelaide, set up new, modern, tide-gauges in 12 Pacific islands.

 

Recently, the whole project was abandoned as there was no sign of a change in sea level at any of the 12 islands for the past 16 years." (Vincent Gray).

 

Gavin Schmidt investigated the claim that tide gauges on islands in the Pacific Ocean show no sea level rise and found that the data show a rising sea level trend at every single station.  But what about global sea level rise?

Sea level rises as ice on land melts and as warming ocean waters expand. As well as being a threat to coastal habitation and environments, sea level rise corroborates other evidence of global warming 

The blue line in the graph below clearly shows sea level as rising, while the upward curve suggests sea level is rising faster as time goes on. The upward curve agrees with global temperature trends and with the accelerating melting of ice in Greenland and other places.

Because sea level behavior is such an important signal for tracking climate change, skeptics seize on the sea level record in an effort to cast doubt on this evidence. Sea level bounces up and down slightly from year to year so it's possible to cherry-pick data falsely suggesting the overall trend is flat, falling or linear. You can try this yourself. Starting with two closely spaced data points on the graph below, lay a straight-edge between them and notice how for a short period of time you cancreate almost any slope you prefer, simply by being selective about what data points you use. Now choose data points farther apart. Notice that as your selected data points cover more time, the more your mini-graph reflects the big picture. The lesson? Always look at all the data, don't be fooled by selective presentations.

graph from Church 2008

Other skeptic arguments about sea level concern the validity of observations, obtained via tide gauges and more recently satellite altimeter observations.

Tide gauges must take into account changes in the height of land itself caused by local geologic processes, a favorite distraction for skeptics to highlight. Not surprisingly, scientists measuring sea level with tide gauges are aware of and compensate for these factors. Confounding influences are accounted for in measurements and while they leave some noise in the record they cannot account for the observed upward trend.

Various technical criticisms are mounted against satellite altimeter measurements by skeptics. Indeed, deriving millimeter-level accuracy from orbit is a stunning technical feat so it's not hard to understand why some people find such an accomplishment unbelievable. In reality, researchers demonstrate this height measurement technique's accuracy to be within 1mm/year. Most importantly there is no form of residual error that could falsely produce the upward trend in observations. 

As can be seen in an inset of the graph above, tide gauge and satellite altimeter measurements track each other with remarkable similarity. These two independent systems mutually support the observed trend in sea level. If an argument depends on skipping certain observations or emphasizes uncertainty while ignoring an obvious trend, that's a clue you're being steered as opposed to informed. Don't be mislead by only a carefully-selected portion of the available evidence being disclosed.

Current sea level rise is after all not exaggerated, in fact the opposite case is more plausible. Observational data and changing conditions in such places as Greenland suggest if there's a real problem here it's underestimation of future sea level rise. IPCC synthesis reports offer conservative projections of sea level increase based on assumptions about future behavior of ice sheets and glaciers, leading to estimates of sea level roughly following a linear upward trend mimicking that of recent decades. In point of fact, observed sea level rise is already above IPCC projections and strongly hints at acceleration while at the same time it appears the mass balance of continental ice envisioned by the IPCC is overly optimistic (Rahmstorf 2010 ).

Basic rebuttal written by doug_bostrom


Update July 2015:

Here is a related lecture-video from Denial101x - Making Sense of Climate Science Denial

 

Last updated on 5 July 2015 by pattimer. View Archives

Printable Version  |  Offline PDF Version  |  Link to this page

Further viewing

Comments

Prev  1  2  3  4  5  6  7  Next

Comments 51 to 100 out of 304:

  1. Daniel - you're apparently missing the point from both Peter and myself.

    The data points from that paper (core samples) fit a linear trend. There is no evidence from those samples of any higher frequency changes (short term), since if there were short term changes those data points would be extremely unlikely to all fall on a linear fit line.

    Peter said that quite clearly, so did I. The evidence in that paper supports a linear fit - not a long term linear fit with lots of short term excursions.

    I think Peter said it better: "If there were short term variations of the magnitude which you suggest between the sparse points then the probability of all of these randomly sampled points fitting any smooth long term curve is small."

    If a rise such as seen in the last 200 years occurred some time in the remaining 600 years covered by the Donnally paper, those carbon dated sediments would not fall on or near the linear fit line for 1400-1800, especially given an average sampling of 60 years.

    There is no evidence in the data that supports your assertion of short term variations, and hence none are postulated by Donnelly. If you think that there are, you need to find some evidence for them. There's certainly no such evidence in the Donnally data points.

    If you don't get that, there's really nothing I can say other than to suggest you find out more about data fitting and evidentiary rules in science.
  2. KR I think you should reconstruct the data from table 1 and include sample 1 like I asked Peter to and you will find that Donnely's data supports my conclusion. The linear fit is quite possibly an undersampling of a higher amplitude trend.
  3. daniel at 08:21 AM on 6 July, 2010

    If "wordy" bothers you I can draw some pictures?

    As a professional scientist I have concerns about your recent comments and general analytical approach to this data set. As we have sparse data and Donnelly gives error estimates, might I suggest again a statistical approach?

    It also worries me that you have not followed through on physics or measurement based evidence which suggest your “short term large variation” hypothesis is highly improbable. High resolution temperature reconstructions are available and do not support your proposal - Grinsted covers this well. Please read it.

    Back to Donnelly, we have more plenty of points here near point 1. The uncertainty in point 1 is shown. It is consistent with the other data which has higher certainty and fits the given trend, statistically speaking. If your argument had any chance of surviving critical scrutiny you would expect deviation of the error envelope from the “curve” for at least some of the other points, which we do not see. The total error envelope would have smooth upper and lower bounds, and I agree we could see variation within these bounds, but this is relatively small and does not compare with the overall rise or the recent rise in rate. This is the point of Donnellys paper. Coupled with evidence of lack of variability in the drivers for your hypothesised variations, this greatly diminishes the probability of your hypothesis being correct. Add to this data from other sources which also give points with high uncertainty but which also fit on the shallow curve (given the error envelope) increase the probability of the low amplitude variation fit being statistically robust, and diminish further the probability of your hypothesis being correct. The recent acceleration in sea level is well documented as are the physics based drivers for this. The overall picture is consistent. For you to hide behind "ad hominem" when it is suggested that Donnelly has access to a great deal more data than he is presenting in one paper on one specific site, reflects poorly on your argument, and I expect better from anyone who claims a science background. Here are some further references (in no particular order) on proxy records, physical basis of sea level rise, recent acceleration, and extended tide gauge data, I hope you will read them, and follow through on a few of their references, and come back with a bit more knowledge and a bit less uninformed opinion: Woodworth 2008, Engelhart 2009, Woodworth 2009, Yasuda 2008, Romundset 2009, Engel 2009, Gonzalez 2009, Leorri 2008, Miller 2008, Goodwin 2008, Wopplemann 2008, Merrifield 2009
  4. Yikes pete, sure I'll read them. But did you reconstruct the data from table 1? We can see that using high uncertainty low res data we still have roughly the same linear trend into the modern age. This is evidence of undersampling of a higher amplitude trend. A high amplitude trend directly measured in recent times using a tide gauge. You seem to think that the data points lie perfectly on a linear trend when even the centres of the boxes don't do that. Coupled with the error margins there is indeed enough slack im the data for higher amplitude trends. Therefore the link to Anthropogenic CO2 as a driver is undermined.
  5. daniel,
    you keep asking people to plot the data from table 1. Could you please show your graph with the linear fit to the data and the line with a slope of 2.8 mm/yr? This would be more convincing.
  6. Ok but it will take me sometime as, like the rest of you, I am a busy man. I would also like to point out to Peter that this comment:

    "Add to this data from other sources which also give points with high uncertainty but which also fit on the shallow curve (given the error envelope) increase the probability of the low amplitude variation fit being statistically robust, and diminish further the probability of your hypothesis being correct."

    has already been discussed in part by both of us. He asked me to examine Donnely 2006 and made the claim that the papers support each other but Donnelly 2006 is an even sparser data set than 2004 with all of the 2004 data fitting within the two latest samples of 2006. The 2006 data has just as much or even more uncertainty than the 2004 data. So how is it that either refine the other? I guess the answer will be read the other papers and they will do better? I will, but I doubt they will do the job.

    Also I have been told that high res studies are in agreement with the overall conclusion of Donnelly's paper. But I have read the Gehrels 2006 high res study and the uncertainties still undermine the validity of a recent unnatural uptrend (not to mention a highly suspicious uptrend when the method of age determination changes). Will the other papers cited like Grinsted do any better? We'll see but from what I've seen and keep seeing when asked to read these studies is that they will probably all have much the same faults because the nature of the measurement doesn't have the certainty required to detect a moderm uptrend.
  7. Riccardo at 00:05 AM on 7 July, 2010

    Ok Riccardo, I have provided a graph here using data obtained from Donnelly 2004 cited by the original article. By averaging the limits of the C14 error bars you can obtain the centres of the date ranges. The mean heights are plotted in table 1 so you can then find the centres of the uncertainty boxes. The height error limits are described in the text and also in table 1.

    In this graph I have tried to reconstruct the Donnelly linear fit by taking the mean height and oldest error limits for the dates of samples 4 and 11. I had to use these visual markers to reconstruct Donnelly's linear trend because I can't find any indication of what the true linear parameters are from the paper. You can see that it seems to be correct visually and rounds adequately to the 2 sig. figs. quoted by Donnelly.



    You can see that I've labelled all the paleo-samples with their respective numbers and have put a linear regression line through all samples (including sample 1) spanning the entire 700 years. I have included the recent linear uptrends mentioned in the text as you asked. They both initiate from the 1mm trend as the text suggests they do. Some of the trends are extrapolated with dashed lines up to 2000AD.

    I think it is fairly obvious from this graph that there is no statistically significant uptrend in the last 150 years detected by this paper.

    I dare any of you to argue with me any further on that point.
  8. johnd - Thank you for the chart.

    A question of note, however: You have the 11 paleo proxies with a slope of 1.023mm/year, the recent tide data with a slope of 2.4 or 2.8mm/year. And you then fit all the data with a slope of 1.2mm/year?

    Modern levels of SLR are KNOWN to be ~2.4 mm/year.

    Donnelly's fit of ~1 +/- 0.2 mm/year average over the 1300/1900 period still holds. Perhaps, just perhaps, there were major swings in SLR between the Donnelly sample points (although as Peter Hogarth points out, lots of other data indicate that this is not the case, filling in the spaces between these linear fit samples). There are certainly no physical phenomena that we know of that could cause reversible SLR changes on that order. The data provided in this paper still demonstrates an average (read that word again) average SLR of 1+/-0.2 mm/year for 1300-1900.

    I think you are really missing the point. The current SLR is known to be ~2.4 mm/year. Donnelly's paper establishes that over the period of 1300-1900 it averaged ~1+/-0.2 mm/year. Therein lies the conclusions of interest, that SLR rates are changing.

    Are you arguing that the current SLR is NOT 2.4mm/year??? Then you need to disprove all of the satellite and tide data. Are you arguing that it did not average ~1+/-0.2 mm/year for the 600-700 years prior to the 1900's? Then you are disagreeing with yourself - your reconstruction and graph support Donnelly, well within his error bars.

    And - if we had a 150 year change in SLR of this magnitude in the previous 1000 years, the paleo data points wouldn't all be on the fit line!

    I suspect Peter will have something to say about this as well...
  9. daniel,
    I did not follow your previous discussion so i'm basing my comments on the last couple of your comments. In particular, the claim "I think it is fairly obvious from this graph that there is no statistically significant uptrend in the last 150 years detected by this paper." (emph. mine).

    The behaviour of the data should be clear and we should also come to the same conclusions reported in Donnelly paper, which is:

    "A linear rate of rise of 1.0 ± 0.2 mm/year intersects all the 2s uncertainty boxes of the record from the 14th to the mid-19th century (Figure 2). Linear regression of the NYC tide-gauge data reveals an average rate of SLR of 2.8 mm/year from 1856 – 2001 A.D. Coupling the Barn Island record and regional tide-gauge data indicates that the rate of SLR increased to modern levels in the 19th century (Figure 2).
    [...]
    The NYC tide-gauge data further support the late 19th century timing of the SLR increase. Linear regression of segments of the NYC tide-gauge data indicate an increase in the rate of SLR from about 1.0 mm/year between 1856 and 1878 to 2.4 mm/year between 1893 and 1921 A.D. [Donnelly and Bertness, 2001]."


    Indeed, you (and Donnelly) get a statistically significant trend of 1.0 mm/yr before about 1850. The last data point lies above this line, although barely statistically significant; including it rises the rate at 1.2 mm/yr but both R and chi2 decrease. Statistics indicates that there has been a change in slope but a weak conclusion, i'd say.
    Adding the NYC tide gauge data between 1893 and 1921, the 2.4 mm/yr line nicely match sample #1. Then, sure i'd not say that there has been an acceleration after the 19th century from sedimentary data alone, afterall there's just one data point supporting this conclusion. Note that not Donnelly nor John in this post claimed otherwise. But overall, i.e. including all the data presented in the paper, the conclusion of an increase in the sea level rise rate from the late 19th century is solid.
    Back to you claim quoted above, i think that the mistake is in the last few words "by this paper", you should have referred only to sedimentary data.
  10. KR at 02:02 AM on 10 July, 2010
    johnd - Thank you for the chart. - ?????????????????????????
  11. johnd - sorry about that; I really need to get off this cold medicine!

    That should be a reference to daniel.
  12. daniel at 19:25 PM on 9 July, 2010

    Daniel, thanks for the chart. It would take me a little longer to enter the data and do one with error envelopes and the tide gauge data, but I think we now get a better explanation of why this misunderstanding has rolled on...
  13. KR at 02:02 AM on 10 July, 2010

    "You have the 11 paleo proxies with a slope of 1.023mm/year, the recent tide data with a slope of 2.4 or 2.8mm/year. And you then fit all the data with a slope of 1.2mm/year?"

    Please read the graph and post carefully. Samples 4-11 (that's 8 count em... 8, have you read the paper KR?) have the Donnelly linear fit (psst... it's not a least squares fit) of 1.023mm/yr. I used visual indicators/markers from the Donnelly graph (fig 2.) to construct it, it rounds to 1.0mm/yr. The dashed portion is extrapolated for comparisons to the other fitted lines etc. The least squares I have fitted to all paleo data produces a 1.2mm/yr average long term trend over the entire 700 years (thats a slope just inside Donnelly's error bars.... pennies dropping yet?)

    "Modern levels of SLR are KNOWN to be ~2.4 mm/year."

    That's nice.....

    "Donnelly's fit of ~1 +/- 0.2 mm/year average over the 1300/1900 period still holds."

    It holds to 2000AD....... look at the graph KR

    "Perhaps, just perhaps, there were major swings in SLR between the Donnelly sample points"

    I am suggesting short term swings that lie within the error bars. They are easily there, as I keep asking you.... read paper.... look at graph. The fact that a shallow linear trend extends up until 2000AD with tide gauge data that deviates from it but remains within the large error estimates of the most recent paleo sample is more than enough evidence to support my critique of this paper. Such deviations could have easily existed

    "(although as Peter Hogarth points out, lots of other data indicate that this is not the case, filling in the spaces between these linear fit samples)"

    He tried using Donnelly 2006 and failed miserably, sure there are other papers and I need to find time to read them but my first impression was not a good one.

    "There are certainly no physical phenomena that we know of that could cause reversible SLR changes on that order."

    I don't know what you mean by "reversible" (probably some exaggerated claim about the short term trends I'm suggesting involving unicorn plasmas). There seems to be alot that the climate science community doesn't fully understand about the hugely complex system known as planet earth. I don't really care if you have or haven't found drivers for ancient SLR swings.
    You can't claim they didn't exist from an amateur non least squares line fit!

    I'm not saying that recent SLR can't be 2.4mm/yr or that the long term trend isnt ~1mm/yr +/- 0.2mm/yr. Actually KR.... if you read carefully.... I'm saying its 1.2mm/yr..... :0 ..... wha!!!???

    "And - if we had a 150 year change in SLR of this magnitude in the previous 1000 years, the paleo data points wouldn't all be on the fit line!"

    I am moved to laugh... You mean like the centres of sample boxes 8, 11 and 10?

    "I suspect Peter will have something to say about this as well..."

    Yes that's right KR, if it wasn't for him we'd barely have a discussion. Why don't you let him do the talkin while you do some readin, not skimming.
  14. Riccardo at 02:19 AM on 10 July, 2010

    "The behaviour of the data should be clear and we should also come to the same conclusions reported in Donnelly paper,"

    Yes it's clear, a simple linear regression of the box centres shows that the ~1mm/yr trend extends up to 2000AD. Donnelly implies that the tide gauge data is unusually high compared to 1300-1850 AD. But it clearly isn't since we use the same methods to obtain a "modern" paleo sample and find no significant uptrend in the data. Donnelly's rate error limits are 0.8-1.2mm/yr over 1300-1850. The simple linear fit of all the paleo data up to 2000AD has a rate within these limits (no detail as to how the limits are acieved in the first place). Are you going to go on again and say that a significant uptrend has been detected by the tide gauge when there is no high certainty paleo data to compare it to? Please save your breathe (fingers).

    "Indeed, you (and Donnelly) get a statistically significant trend of 1.0 mm/yr before about 1850."

    Well actually I don't know what statistical analysis Donnelly has performed on his trendline since it's not mentioned. I have simply tried to reconstruct it using visual markers. Linear regression of the centres of sample boxes 4-11 gives a rate of ~1.1mm/yr. Donnelly was trying to marry up the earliest tide gauge trends with a proposed linear trend through 1300-1850.

    ".... including it rises the rate at 1.2 mm/yr but both R and chi2 decrease."

    Can you do some calculations to show this please and by how much they decrease? I won't have time over the next couple of days.

    You then go on to say that you agree the paleo data doesn't support a recent acceleration and that nobody was claiming otherwise or at least not the paleo data alone. But they were claiming samples 4-11 did and I am showing that inclusion of sample 1 undermines that conclusion. The short term tide gauge data compared to the much less certain, long term paleo data is invalid and I believe I have shown by inclusion of sample 1 in a simple linear regression that short term variance is easily achievable amongst samples 4-11.

    After agreeing with me on the insufficiency of the paleodata you then say that the conclusion is solid. (Throws hands up in air as a sign of frustration).

    I didn't make an error by claiming "by this paper" Donnelly only provides sedimentary data. Do you think he collected the tide gauge data? Do you still think the conclusion drawn from the comparison between the two data sets is valid?
  15. Peter Hogarth at 06:10 AM on 10 July, 2010

    Pete I can't quite tell if this comment was supposed to be taken as a backdown on Donnelly 2004 or if you intend to argue further with a graph of your own.

    If a backdown I acknowledge that it would be of this paper and this paper alone. I cannot then use this to say that all of Donnelly's work is invalid.

    But I do have some personal doubts and I feel that this discussion should prompt those reading on to look again with a more critical eye as to what is published in both in climate literature and other disciplines.
  16. daniel,
    the paper shows two data sets and draw conclusions explicitly based on both. It really does not matter who actually collected the data.

    The large error in the sedimentary data does not imply an equivalent large error in the trend. Indeed, the latter is 0.2 mm/yr and assuming a comparable error for the tide gauges data the difference in the trends is still significant. Or do you think it's not possible to compare two different data sets?
  17. Riccardo,
    you have been absent from the discussion and it may pay for you to back track a little. I do not dispute any of the trends discussed or proposed by anyone I only dispute the argument made against me that significant deviations from the long term trend are impossible, highly unlikely or have been shown to not exist at Barn Island. Such short term deviant trends in the paleo data set would undermine any conclusiom of an unusual modern uptrend. I also have gone so far as to say that this is a good example of poor science swallowed by people who should know better but are blinded by fear of impending doom.
  18. Sorry Riccardo,
    to answer your question more directly, I don't think this particular comparison is valid at all. You also say that the tide gauge data has more deviation than 0.2mm/yr. Could you explain what you mean by that? Donnelly believes his long term trend may deviate overall by 0.2mm/yr over a ~550 year period. You can't compare that to the deviations in the tide gauge data and say that the tide gauge data is more noisy. There are no error limits placed on the short term linear trends discussed by Donnelly for the tide gauge data. You cannot compare visual non-quantified scatter to statistically determined error limits of a linear trend. Even if there were error limits on the proposed short term trends you should be able to see that I have shown that the tide gauge data can lie on a shallow long term trend of paleo data (samples 1-11) with a rate (1.2mm/yr) within the error limits proposed by Donnelly (1.0 +/- 0.2mm/yr)..
  19. daniel,
    i see your point, but you're comparing the values of sea level with sea level rise, i.e. the trend. It doesn't matter if the recent "high" resolution sea level data fall within the uncertainty of the sedimentary trend, what matters for Donnelly conclusions is that the two trends are significantly different.
  20. daniel at 14:18 PM on 10 July, 2010

    To summarise: The original paper provides evidence used to suggest a relatively recent acceleration in sea level rise. The recent trend is taken from tide gauge data. This is accurate and errors small. We can take this as the "real" local relative sea level rise since 1856. The question is, what was the local sea level doing before 1856? Several data points derived from peat sediments are given where the dating of the sediment has uncertainties. Two more points are derived from heavy metal polution and pollen. A linear trend is fitted to these points which is less than half of the tide gauge trend, hence, acceleration. You argue that the longer term trend fitted to the sparser points may be hiding short term variations and other possible periods of acceleration. I argue that any variation is constrained by the error envelope of the points and the physical processes which might cause fast variations. I have provided a considerable body of evidence which supports this, but you have still focused on this data from this paper and provided a simple chart (but without any statistical analysis). I'll see about a chart with the actual tide data etc.
  21. Daniel - Something you haven't really addressed (at all, and I don't count your ad hominem statements) is that the paleo data indicates sea level, not sea level rise rate. The rise rate is extrapolated by looking at multiple measurements of sea level.

    I've said it before, but it didn't seem to register - if a major upswing like the currently observed 2.4 mm/yr SLR occurred in the 1300-1850 period, the paleo data points of sea level couldn't fall anywhere close to the linear fit line for the 1.0+/-0.2 mm/year SLR trend.

    If there was a brief 100-150 year rise at 2.4 (as currently observed) there would be a step change in the observed paleo sea levels. That is not shown in the data. The only way a rise at those rates could occur and still fit the data (including the error bars) would be if the SLR went negative (or close to it) long enough for the long term sea levels to still average 1.0+/-0.2 mm/yr. If you have a physical process for something like that, I would love to see it.

    If this was measuring rate of rise, there are a lot more degrees of freedom. But these measurements are of sea level, and the long term historic rate is clearly about half the current rate.

    That's why I said "reversible", and why I don't feel your hypothesis of high variability and equally high rise rates in the past can hold, unless you also postulate extremely low SLR levels.

    The current rate of 2.4mm/yr is well established by tide gauge data, Donnelly submits evidence for ~1.0mm/yr for the 1300-1850 era - and your graph fit falls within his error bars for that rate. Sparse data or not, if a high SLR rate occurred in that period, it would have to be matched by a low SLR rate for the long term trend in sea level to still be ~1mm/yr.

    If you don't get that, well, end of discussion for me. I'm not going to waste my time yelling at the deaf.
  22. Peter Hogarth at 07:48 AM on 11 July, 2010

    "Several data points derived from peat sediments are given where the dating of the sediment has uncertainties."

    A candidate for understatement of the year. Sample 7 has a height uncertainty of +/- 10.4cm and a date range of 172 years, Sample 8 does not do much better. Sample 9 generally dates older than sample 10 despite attempts using the principle of superposition to eliminate such assignment errors and sample 11 cannot be definitively defined between two date ranges which span 123 years in total.

    "I argue that any variation is constrained by the error envelope of the points..."

    And here we have overstatement of the year. Gentlemen I will provide a graph with linear fits between small groupings of the data points if you lik and you will see that on the short term deviations fall easily within the error envelopes. But then you will claim that such 2-4 point linear trends are statistically insignificant. But that's the point! Donnelly's study can't make short term comparisons so any comparison to the recent data is invalid. Donnelly doesn't give us enough data to produce statistically significant trends on the short term.

    " ...and the physical processes which might cause fast variations."

    and

    "I have provided a considerable body of evidence which supports this, but you have still focused on this data from this paper and provided a simple chart (but without any statistical analysis)."


    Well we could discuss how well these physical processes are known and you could cite specifically which papers deal with causation but I would at least like you to acknowledge that this paper, using the data it provides only, has no case to make. Donnelly has gone to all this effort to determine a linear trend in the hope it would alone provide the evidence. He only cites other climate studies not SLR studies to support his conclusions and he admits they only roughly do so. I fear that this is the kind of study that is used to determine causation of SLR (a rough correlation to paleoclimate data) where other, as yet unknown, ocean heating or seismic factors may also be playing a role. It is often claimed that climate science is in it's infancy so you should not be surprised when others claim SLR drivers may not be properly understood. If this statistically insignificant kind of study is used to determine causation then I am very worried indeed about what we do and don't know about SLR drivers and what the known drivers were actually doing in the recent and distant past.

    You do realise that conceding this point and accepting the poor quality of this paper does not mean you lose the war. I would love to discuss other papers but if you can't bring yourselves to accurately critique this paper what point is there in moving on to others? You can't say that all the other papers validate this study without discussing them in detail (Donnelly doesn't). You can't use dodgy materials to build a strong house. There needs to be other studies that are of higher quality than this one in order to support this one. Citations please (I haven't read your review articles Pete, I've been busy defending an obvious and very pertinent point... amongst living my life).

    As to the comments about my lack of statistical analysis can you please state exactly what you want? I can do a chi squared test if you like? Can you see the R squared value in my least squares fit? Do you think these kinds of analyses are going to support your cause? They will only support mine further. If I do a chi squared test on whether the 1mm/yr trend is significant on short term collections of the paleo data do you think we will get a statistically significant trend? That pendulum swings both ways and mostly in my favour (if not completely)
  23. Riccardo at 04:34 AM on 11 July, 2010

    " ...but you're comparing the values of sea level with sea level rise, i.e. the trend."

    I'm only really comparing the trends.

    "It doesn't matter if the recent "high" resolution sea level data fall within the uncertainty of the sedimentary trend,"

    It clearly shows that modern swings in SLR fit inside the error envelope of a shallow, linear long term trend in paleo data that is within the rate error limits of Donnelly's 1.0mm/yr +/-0.2mm/yr. Therefore there is no evidence to suggest tjhy could not have occurred in the past or that recent swings are unusual at Barn Island.

    " what matters for Donnelly conclusions is that the two trends are significantly different."

    They have not shown to be different. Directly measured, high certainty, high res, short term data cannot be compared to it's exact opposite ie. indirectly measured, low certainty, low res, long term trends.
  24. "I've said it before, but it didn't seem to register - if a major upswing like the currently observed 2.4 mm/yr SLR occurred in the 1300-1850 period, the paleo data points of sea level couldn't fall anywhere close to the linear fit line for the 1.0+/-0.2 mm/year SLR trend. "

    ...and I've said it before KR, that's total bunkem. Provide a graph to show your claim is true. You will find yourself dissapointed by the rashness of your claims.

    "If there was a brief 100-150 year rise at 2.4 (as currently observed) there would be a step change in the observed paleo sea levels. That is not shown in the data. "

    I can see two possible points where that rate or a rate closer to 2.4mm/yr than 1.0 mm/yr. could occur. Best candidate is samples 9-11 followed by sample ranges including 7-8.

    " If you have a physical process for something like that, I would love to see it. "

    Please discuss the evidence for a lack of active drivers in detail citing the relevant papers.
  25. So, daniel - you look at the larger gaps in the paleo data from this (and only this) paper, and assert that large SLR rates could occur in the spaces between the data points? I point out between, because the slope between 9/11 and 7/8 that you use as an example shows an average slope of ~0.9-1.0mm/year, as far as I can determine with a quick examination (certainly not >2mm/year!!).

    You do realize that these data points for sea level show the integral of the SLR over time, and that for such a large value of SLR (derivative) to occur in that period, there would have to be a corresponding low/negative SLR in that same period in order for the integral over that period to still yield ~1mm/yr? If you recognize that, great, we're half-way there, and perhaps past arguing and back into discussion over what could cause such high variation around the 1.0mm/year mean for 1300-1850.
  26. daniel at 14:18 PM on 10 July, 2010

    A quick chart with all of the data used by Donnelly 2004, except the tide gauge data (NY) uses monthly mean values and extends to today.



    The tide gauge data consists of very frequent samples (modern readings are every six minutes or so) averaged over each month. As such errors are extremely low. The seasonal MSL variations are retained in this chart and represent measured mean sea level. The overall long term linear trend is 2.8mm/year, however in this tide gauge data the trend has gradually accelerated over the measurement period.

    The sediment, pollution and pollen derived data is shown with approximate two sigma and one sigma error envelopes. However it should be noted that the variance from a straight line fit for all points is very low, and in addition a smooth curve fit is further suggested by the relatively low level of decadal and inter-annual trend variation in the 150 year tide gauge record. The accelerating trend in sea level rise is evident in the overall data set. The error bars on the sediment data overlap the actual tide measurements even when an annual MSL average or overall fitted linear trend for tide gauge data is used.
  27. KR at 14:34 PM on 11 July, 2010

    "you look at the larger gaps in the paleo data from this (and only this) paper, and assert that large SLR rates could occur in the spaces between the data points?"

    But also the large error limits of the points which indicate a high uncertainty as to where the true sea level was.

    "I point out between, because the slope between 9/11 and 7/8 that you use as an example shows an average slope of ~0.9-1.0mm/year"

    Well you clearly misunderstood me. I was saying within the small group 9-11 and also within a small group of 3-4 data points which include 7-8 either 7-10 or 5-8.

    "for such a large value of SLR (derivative) to occur in that period, there would have to be a corresponding low/negative SLR in that same period in order for the integral over that period to still yield ~1mm/yr?"

    Yeah, so, why is that so unlikely? I don't think we have to evoke negative trends to support my analysis.
  28. daniel,
    could you please explain this sentence:
    "It clearly shows that modern swings in SLR fit inside the error envelope of a shallow, linear long term trend in paleo data".

    As I understand it, you're saying that the tide gauge seas level data (not trend) are within the trend error of paleo data.
    Comparing trends is to compare 1.0 mm/yr (+/- 0.2 mm/yr) with 2.4 mm/yr. Comparing sea level, instead, is to see if tide gauge data are within the limit of variability of sedimentary data. You're doing the latter, so you're not comparing trends.

    Is it possible, at least in principle, that there has been similar up swings at 2.4 mm/yr in the past? Yes, but putting physics aside. If you take the lower error bar of a point and the higher of another you might get such trends. But it's unlikely, there's nothing that make us think it's actually happened and it's hard to immagine a mechanism producing such up swings. Are you aware of a sudden increase in temperature or large land ice melt between 1300 and 1800 A.D.? Are you aware of other, higher resolution, datasets showing this swings?
  29. Riccardo at 18:04 PM on 11 July, 2010

    No you see the error envelope on Peter's graph. They are continuous lines that extend along either side of the paleo trend. What I am saying is that the tide gauge data points fit inside of that if it were extended over sample 1. The MSL does anyway. You can see this more clearly in Donnely which uses averaged tide gauge data which lie over the sample 1 uncertainty box. Donnelly and Peter seem to think that this overlap supports their case but actually it supports mine. You can see that paleo samples do not adequately show us where the true sea level is because they have such high uncertainties. A big deal is made about the centres of the boxes but clearly the tide gauge data shows the centres aren't that important because the centre of sample box 1 does not lie directly on the averaged data point trend of the tide gauge data. The mean height of sample 1 fits the tide gauge data at 1950 rather than the center of the assigned date range which is 1975. That's directly measured data lying on the very extreme of the 95% confidence interval for the mean height estimate of sample 1 and then I'm told tbat short term deviations in the older data are unlikely to be far from the centres of the boxes or the linear model. For at least three samples 8, 10 & 11 the linear model is far from the box centres. You guys can't have it both ways.
  30. daniel at 13:42 PM on 11 July, 2010

    "Understatement of the year" "overstatement of the year" etc. Histrionics does not make up for lack of knowledge, nor does it engender respect. To put these sea level numbers into context, the sea level varies by around 1m every day close to this site, and through year seasonal and tidal variations add to this. I believe I have shown from the data that your argument (that large intersample variations containing trends similar to those of the past 150 years are "likely") is weak.
  31. read above pete
  32. daniel at 23:20 PM on 11 July, 2010

    No, you have not said this until now. What you have argued is that extending the approx 1mm/year long term linear trend from pre-tide gauge data beyond 1856 by including point 1 gives a 1.2mm trend and error bars which can fully accomodate the total tide gauge data record anywhere inside this trends error envelope, therefore you argue this paper should not be used as evidence that the recent 2.8mm/year trend is unusual.

    Your case for including point 1 in an overall 1.2mm trend is invalid. Clearly, the density of tide gauge data makes any influence of point 1 marginal on the measured 2.8mm/yr trend since 1856, clearly, the recent measured trend is diverging away from the long term paleo trend. That point 1 error bars overlap the tide data possibly validates point 1, but the fact that the tide gauge data overlaps point 1 does not validate extending the pre-1856 trend. The post 1856 trend is already determined as measured fact by the tide gauge data. Point 1, statistically speaking, is on this trend.

    The question you are asking really, is: Can the 150 year tide gauge data trend be accomodated inside the pre 1856 error envelope, and show that the longer term trend could most likely have missed similar episodes of acceleration in the past? This is the question I address in my chart.

    The chart I have given shows the error envelopes based on the statistics given for each point, up to the start of the tide gauge data, which has much lower standard deviation and represents the measured data. It is not the same as a linear trend. It is likely that the envelope will follow or at least accomodate whatever the true trend was. We can only ascribe a probability to this.

    It is barely possible to fit the tide gauge data into the 2 sigma envelope in one pre 1800 place, but it is not a "good" statistical fit anywhere over any approx 150 year period except after 1800 (by comparing with at least three paleo points, though this is not a strong test). It is not possible to fit inside the 1 sigma envelope anywhere, and the data series deviates away in opposite directions above and below the envelope, indicating significant systematic trend error.

    I repeat that your suggestion is possible, but not "likely" from this data. Of course we need a reality check, and the temperature record and other work I referenced strengthens the case presented in the paper, and further weakens your hypothesis.
  33. daniel at 23:20 PM on 11 July, 2010

    One thing slightly bugged me about my chart, the slight offset between the tide gauge data and the "centre" of the error envelopes, where we would expect the real trend to probably be. I checked the tide stations closer to the site. One (New London, CT, with a shorter record from 1938) is very local, but the other (from 1856, NY) has the longest record. I charted the NY record for this reason. I've now noticed there is a small vertical offset (not unusual between stations) between the two that (subjectively at least) neatly resolves the visible offset in the chart.
  34. In order to also start some fresh productive discussion, hot off the press is the following from Llovel 2010



    This is a new measured result from GRACE which goes at least some way to confirm explanations of seasonal variations in mean sea level, and partially answers previous points on this post and also answers some questions on the visualizing sea level post.

    The further relevance is that land storage contribution to mean sea level trend and budget is slightly negative, giving -0.22 +/-0.05 mm/year, confirming modeling results for this with direct measurements for the first time.
  35. Peter Hogarth at 16:21 PM, Peter, the comment you make about modern tide gauges taking measurements every 6 minutes or so interests me.
    Given that at some locations the time it takes for the tide to rise can be considerably longer than it takes for it to fall, just a quick look at one site showed over 15 hours rising, 9 hours falling, would taking such regular measurements introduce a bias into the average when comparing against those records where the average is that of the high and low tide measurements?
  36. daniel,
    "What I am saying is that the tide gauge data points fit inside of that if it were extended over sample 1."
    They are inside the sedimentary data uncertainty and we expect them to stay there for a while given the difference in the slopes. Ok, but they do have a different slope. Does this tell you something?
  37. Riccardo at 06:08 AM on 12 July, 2010

    Yes, it tells me that short term variations are quite possible within the paleo data set. It is verified by the highly deviant tide gauge results still lying within the uncertainties of the modern sedimentary data.
  38. Peter Hogarth at 02:15 AM on 12 July, 2010

    "No, you have not said this until now."

    Pretty much what I've been saying the whole time Pete. Sure it's a different approach but what I'm saying should have been obvious from the very beginning.

    "What you have argued is that extending the approx 1mm/year long term linear trend from pre-tide gauge data beyond 1856 by including point 1 gives a 1.2mm trend and error bars which can fully accomodate the total tide gauge data record anywhere inside this trends error envelope,"

    No not anywhere. For instance it is unlikely to have occurred between 1700-1850 but there is no evidence to suggest it couldn't have occurred between 1300-1700.

    "Your case for including point 1 in an overall 1.2mm trend is invalid. Clearly, the density of tide gauge data makes any influence of point 1 marginal on the measured 2.8mm/yr trend since 1856, clearly, the recent measured trend is diverging away from the long term paleo trend."

    Uh huh.... but the divergence is covered by the uncertainty of modern sedimentary data. Sedimentary data that lies on a 1.2mm/yr 700 year trend. Hmmmmm.....invalid? Maybe, I dunno, highly relevant?

    "That point 1 error bars overlap the tide data possibly validates point 1, but the fact that the tide gauge data overlaps point 1 does not validate extending the pre-1856 trend."

    Why? Too inconvenient a fact for you?

    "Point 1, statistically speaking, is on this trend."

    It is also.... statistically speaking...... on the 1.2mm/yr trend. Therefore so is the tide gauge data. Statistically speaking of course......

    "It is not the same as a linear trend. It is likely that the envelope will follow or at least accomodate whatever the true trend was. We can only ascribe a probability to this."

    Amen brother, finally we agree on something.

    "It is barely possible to fit the tide gauge data into the 2 sigma envelopein one pre 1800 place,...."

    Mere opinion, show it with an overlay.

    "...but it is not a "good" statistical fit anywhere over any approx 150 year period except after 1800 (by comparing with at least three paleo points, though this is not a strong test)."

    Which is why Donnelly's paleo data is insufficient for a valid comparison to tide tide gauge. He doesn't have the resolution let alone the certainty.

    "It is not possible to fit inside the 1 sigma envelope anywhere,..."

    Mean height of paleo sample 1 has tide gauge lying over 95% (2 sigma) confidence interval extreme. Care to retract?

    "...and the data series deviates away in opposite directions above and below the envelope, indicating significant systematic trend error."

    Perhaps only if compared to an irrelevant long term trend on the short term scale. An overlay to help make your point?

    "I repeat that your suggestion is possible, but not "likely" from this data."

    Simply not shown at all. On the short term, with the data we have, it's just as likely.

    "Of course we need a reality check, and the temperature record and other work I referenced strengthens the case presented in the paper, and further weakens your hypothesis"

    I think what you'll find pete is that if you apply my kind of reality check to all that supportive evidence you will find the AGW hypothesis severely weakened. You can't support poor data and conclusions with more poor data and conclusions. If you can't see the error here you can't thoroughly critique the mass body of evidence you claim is out there.

    Peter Hogarth at 03:52 AM on 12 July, 2010

    Please provide the graph with the relevant data then and provide me with your source so I can also provide graphs of a similar nature.
  39. daniel,
    here the discussion becames circular and I can only repeate what i wrote in #78.
    Do we have any evidence of such "short" (one century and a half) term variations in the past or are they just hypothetical? What should have produced these accelarations? What do other datasets say?
  40. Riccardo at 16:26 PM on 12 July, 2010

    You tell me. Donnelly can't.
  41. To Riccardo #89

    Ok Riccardo again I'l answer a bit more directly.

    Do we have any real evidence that the 1mm/yr trend existed without short term variations from the paleo data? How can we know if drivers existed or not when the links of SLR to temperature fluctuations are so poorly achieved using paleo data?

    What do other data sets say Riccardo? Post some links we can discuss it.

    Here's a graph of Donnelly's data with sample 11 reassigned to the younger date range which was never adequately constrained by Donnelly in the first place. (But he was happy to fit the older date range to suit his trend).



    Are you sure it's not possible for the recent sea level trends, or very similar slightly less severe uptrends, to have existed in the past given the uncertainty in the paleo data? Does this make the recent uptrend look less alarming? Is all of pro AGW science done this way? You wanna discuss drivers? Post some links.
  42. daniel,
    step by step things are getting clearer in my opinion and your last graph is very helpful. Now, in light of this new graph, could you tell what should drive these sudden accelerations in the past? Known science can tell just for the last one, the one in the last century, it's mainly ocean thermal expansion and land ice melting. It's up to who's trying to cast doubts on the accepted explanation to provide evidence and reconcile it with other data that support his hypothesis.
  43. Riccardo at 00:00 AM on 13 July, 2010

    Well actually I would question whether your claim that SLR is so strongly linked to the percieved warming of the 20th century or if the temperature record is so reliable into the past, both over this period and beyond.

    But since this is a pro AGW website it is up to people like myself to disprove the unproven. Are there any papers you rely on Riccardo to undoubtedly satisfy yourself of SLR driven strongly by climatic factors and also the reliability of the reconstructed temperature record? Post a link and we'll discuss it. I've only just started to find some time to read some of Peter's links throughout the above discussion.
    Response: Might I suggest starting with Vermeer 2009 that examines the link between sea level rise and temperature, finding the two are closely linked:


    Observed rate of sea-level rise (red) compared with reconstructed sea level calculated from global temperature (dark blue with light blue uncertainty range). Grey line is reconstructed sea level from an earlier, simpler relationship between sea level and temperature (Vermeer 2009).

    Of course, where the paper gets really interesting is when it uses these results to predict future sea level rise which we examine here...
  44. daniel,
    I'd not look for cutting edge research paper, it's just simple physics. Global sea level may rise due to thermal expantion of the ocean or to the melting of land ice; locally it may rise for a change in the ocean circulation pattern. It has nothing to do with AGW and the like.
    Add to this that even simple statistics would tell you that going from the lower to the higest end of the uncertainty range is statistically very unlikely.

    Beyond this, I might we willing to say that global sea level rose because of a rise of the ocean floor with respect to the continental crust or because of a giant movent of masses that locally produced an increased gravitational pull. In principle they are both possible, in practice they're both untenable. And this is also the line we should never cross if we want to talk about the science of any phenomenon instead of producing pseudo-scientific hypothesis.
  45. daniel at 12:31 PM on 12 July, 2010

    I suggested the tidal data barely fits into the 2 sigma envelope and does not fit anywhere except where it is using the 1 sigma curves. I have extended these up to point 1 for sake of argument only. You state “Mere opinion, show it with an overlay”. My statement stands.

  46. daniel at 12:31 PM on 12 July, 2010

    On data point 1, you state “It is also.... statistically speaking...... on the 1.2mm/yr trend” if we are talking just the non-tide gauge data, I agree, but we are not. “Therefore so is the tide gauge data”. This is incorrect. I have included error bars on the tide gauge data. Do the error bars on the paleo data overlap the tide gauge trend? Yes. Do they overlap the paleo trend? Yes. Could this point be on both trends, Yes. Do the error bars on the tide gauge data overlap the extended paleo trend? No. Could they be on this paleo trend? No. Clearly the trends diverge (and continue to do so). We must accept that the tide gauge record from close to the site in question shows a 150 year relative sea level trend which is in excess of 2 mm/yr. It is important to see that this trend is smooth at decadal timescales over the entire 150 year period. There is no inconvenience to me here, perhaps a basic misunderstanding on your part?



    The light grey line is the paleo trend. Red is NY tide gauge MWL. Blue is New London tide gauge MWL. New London is relatively local to site. Next I will look at the paleo trends in more detail.
  47. daniel at 12:31 PM on 12 July, 2010

    I have also looked at what we can say about short term trends from the paleo data. If your hypothesis was “likely” we should see random or high variations in trend away from the overall trend when looking at any approximate 150 year period in the overall time period. These variations are not random and not large.



    These are trends from every four consecutive paleo points. I have also taken gradients from three consecutive points, as a three or four point record from this data encompasses a period most similar to the tide gauge data (roughly 150 years). I have also added the shorter New London tidal series (running annual average) for comparison. We should be cautious with this crude approach, but it’s not a “trick”, we treat all paleo data consistently. We allocate a “centre” date for these mini-series, then finally add the measured trends from the two tide gauge series with a centre date for these (solid points below). I am aware I have argued against this breaking up of time series elsewhere, but on this occasion we are actually looking for any evidence of short term changes in gradient - whilst still honouring all of the points in the overall analysis by generating a series of overlapping trends. If deviation of the points from mean trend was high, we would expect inconsistent or noisy results. If large variations had occurred, we would expect to see some significant changes in gradient, especially on such short series. If there was no acceleration in trend, we would see a flat line, or an insignificant trend.





    The results are surprisingly consistent, all linear fit trends are significantly positive. The derived rate of increase in trend (just from the paleo data) is clear and consistently positive. The addition of the tidal data confirms and increases this trend, showing late period acceleration and giving strong evidence that the most likely earlier and later trends are close to those suggested by Donnelly, whilst also suggesting short and medium term variability similar to the tide gauge data, (which is intuitive). The second order fits are not meant to represent possible variation, but simply to indicate the significant positive second order component, ie acceleration. This analysis supports the conclusions of the paper, and does not support your suggestion.
  48. Surprisingly consistent, indeed. Good job Peter.
  49. Peter Hogarth at 02:45 AM on 13 July, 2010

    The 1 sigma bounds are irrelevant Peter. I've already commented on this. Do you see the overlay of the tide gauge data on sample 1? That's right, it crosses the uncertainty box at the mean height and date range extreme. That is the most probable height of sample 1 has a date at the 2 sigma extreme in the date range, the least probable according to you. On this one Donnelly sees it my way since his linear fit does this also at samples 11, 10, 5 and 4 not to mention the extra paleo data used to constrain the C14 ranges.
  50. daniel - or more, properly for those reading daniel's comments, since he's apparently missed the point:

    The data referred to comes from two different data sources measuring the same quantity. The first is recent tide gauge data (last 150 years), indicating a sea level rise (SLR) of 2.4 mm/yr. The second data set comes from a fairly sparsely but well distributed sampled set of paleo data, based on core sample species abundance, indicating sea levels at various points over the last 1000 years, with a statistically tight best fit of ~1 +/- 0.2 mm/year. The Donnelly paper (single paper) daniel has been discussing therefore concludes that SLR increased from ~1 to 2.4 mm/year starting ~150 years ago, a conclusion completely in agreement with all other data sets (which daniel does not discuss).

    Daniels argument appears to be that natural variation could account for the last 150 years of 2.4 mm/year SLR. In other words, the ocean isn't warming (contradicts the data), icecaps aren't melting (contradicts the data), it's variability, not global warming.

    Any similar 2.4 mm/year rise extending over 100+ years in the past would have greatly offset later paleo data points. It's barely statistically possible (but certainly not the statistically justified hypothesis!) that such a rise could have occurred, if a corresponding <1.0 or even negative SLR trend balanced it out, returning later paleo samples to the ~1 mm/year trend. If there was a situation of high variability, that would have to cycle +/- around the 1.0 mm/year level to still show that on the long term averages.

    This is exceedingly physically unlikely, however, as sea level is driven by steric expansion and icecap/glacial melt, rather low variability processes. In addition, no records, evidence, or data supports sufficient temperature drops or increased icecap/glaciation in the past 1000 years. Additional data sets for paleo sea level also fall on the ~1.0 mm/year slope, with a total sample spacing much closer than the 11 samples in the Donnelly paper, thus leaving no room for such high variability periods.

    Claiming extreme variability that is somehow missed in the paleo data for even this single paper under discussion is not justified by the evidence. And opening your view to additional data clearly eliminates such variability in the time period under discussion. Asserting that it somehow snuck in between the data points is simply wishful thinking, not science.

Prev  1  2  3  4  5  6  7  Next

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.

Link to this page



The Consensus Project Website

THE ESCALATOR

(free to republish)


© Copyright 2019 John Cook
Home | Links | Translations | About Us | Privacy | Contact Us