Dave Burton, you still don’t understand.
You finally commented on this post, showing a graph of San Diego data and saying:
As you can see, there’ve been >112 years of continuous measurements, and still no detectable acceleration.
Not true. I detected acceleration. You don’t believe it. Then you gave us this:
To the eye the trend looks straight as an arrow. As any engineer could tell you, that means there’s been no practically significant acceleration.
“To the eye”? How many times do I have to repeat that “the eye” is a great tool for getting ideas and clues (one of the best there is) but cannot be relied on for significant conclusions? For that, we use math. It’s hard to believe you would say something so foolish, but it’s finally starting to dawn on me: this really is how you think. You think your eye is better evidence than my math. As for “any engineer” concluding there’s “no practically significant acceleration” — you just made that up. It’s bullshit.
Let’s get to the heart of the matter:
Linear regression finds a linear trend of 2.176 ±0.184 mm/yr, and quadratic regression finds an acceleration of 0.00879 ±0.01259 mm/yr², which is neither statistically nor practically significant.
This is the only evidence you have. The rest of your arguments are nonsense or irrelevant or both. I guess this is why you are so convinced there is no acceleration in this series. The only statitistical test you have reported, apparently the only one you are able to apply, is to fit a quadratic. Not all trends with acceleration resemble a parabola, especially in sea level time series. For many a quadratic fit is a weak test. Many.
I’ve already told you that. I’ve said it often. How many times do I have to repeat this? How often must I repeat it before you even acknowledges its existence? I’ve said this often enough that you have no excuse for not being aware of it. Yet here, again, you don’t seem able to do anything else, but talk about it as though it gives the “final answer.”
I didn’t conclude “acceleration” based on fitting a parabola. I didn’t do so based on the lowess smooth (but it’s a great way to show what’s really going on). Conclusions were based on changepoint analysis of a piecewise linear model: two straight lines joined at their endpoints. This is a more realistic model for sea level trends than a parabola. When I applied it to the data from San Diego since 1950, the p-value was 0.017. That’s statistically significant.
You don’t seem to like my starting at 1950, so I ran the same test using the whole time span of data from San Diego. Now the p-value is < 0.004. That's even more statistically significant.
Those are the numbers, Dave. I didn’t make ’em up, I calculated them. They are “statistically significant.” Not just for San Diego, not just for Key West and Boston and St. Petersburg. For LOTS of tide gauge stations. This whole thing started with your statement that many stations show acceleration early but not after the late 1920s. The data contradict you; there are SO MANY that do, your claim isn’t just wrong it’s ludicrous. It’s truly idiotic.
What they don’t show, is response to your “fit a quadratic to the whole thing” test. I’ve told you repeatedly that for finding acceleration in tide gauge data (after the “late 1920s” or not) that’s a weak test. I wasn’t kidding. It really is. The model behind it is that the signal shows constant acceleration, i.e. the acceleration remains exactly the same throughout the entire time span. That particular idea is so clearly contradicted by data — by individual tide gauges, by reconstructions of global sea level, and by satellite data — that to use it as your one and only criterion for detecting acceleration is a dreadful mistake.
In fact given what we do know about sea level — that it has shown all kinds of acceleration over the last century+ including deceleration in the early 1900s and acceleration lately — the “fit a quadratic to the whole thing” test is just about the weakest choice available. If that was your intent, congratulations.
Other methods perform better. I’m not kidding, they really do. The changepoint analysis I used is one of them. For San Diego data, it says “statistically significant.” That’s just a fact, whether you like it or not.
You have not faced this fact. You have ignored this fact. You have been told, multiple times. You can no longer ignore the results I got from changepoint anaysis. Either accept them, or explain why you don’t.
What I find most amusing is your attempt to “blame it on el Niño.”
Do you see it? During El Niño, easterly trade winds diminish and the entire Pacific Ocean sloshes east, raising sea-level at San Diego, and lowering sea-level at Kwajalein. During La Niña and ENSO-neutral conditions, especially of long duration, the opposite happens.
So, let’s compare the measured sea-level plot to what Tamino did, in his effort to find evidence of acceleration. I’ve circled some of the El Niño peaks:
Do you see it? Tamino didn’t measure changes in long-term trend, he just measured ENSO slosh.
That big run-up at the end of his graph corresponds to the transition from the strong La Niña of 1999-2000, to the big 2015-16 El Niño.
No, Dave. “ENSO slosh” creates slosh. It doesn’t create trends. You also “circled some of the El Niño peaks,” but not all of them. Just the ones that you think makes your claim look realistic.
Certainly el Niño affects sea level at San Diego (and Kwajalein). You think it’s responsible for the trend change I found? If that’s true, then if I remove the influence of ENSO it should remove the trend change I found. Let’s find out.
I made my own model of the impact of el Niño. Here’s what it gives:
It’s not a perfect model but it’s pretty good. I removed this ENSO response from the original data, then tested the result for the acceleration I had found originally. It didn’t make the acceleration go away, it made the acceleration clearer and more obvious. Thanks to removing the noise induced by ENSO, the p-value is now below 0.0001. It is definitely, positively, statistically significant. Whether you like it or not.
Since you mentioned Kwajalein, I can estimate how ENSO affects its sea level too:
Then I can do the same, look for acceleration in the ENSO-adjusted data:
Lo and behold, there it is. Statistically significant. Both Kwajalein and San Diego show an increase in the rate of sea level rise, even though el Niño affects them in the opposite direction.
Here’s the rate estimated from the piecewise linear fit (solid blue line with dashed blue uncertainty range) and from a lowess smooth (solid red line with pink shade uncertainty range):
Blaming my result on el Niño doesn’t work. Taking it into account doesn’t make the statistical significance of acceleration go away, doesn’t even reduce it, it makes it a lot stronger.
Your original claim about tide gauge data not showing acceleration since the late 1920s, is the foundation of your denial of the sea level crisis. Your original claim is wrong. The tide gauge data show it. Using only quadratic fits to test for acceleration, ignores it.
Dave Burton: either accept the results of changepoint analysis and stop repeating that bullshit, or explain why you don’t.
This blog is made possible by readers like you; join others by donating at My Wee Dragon.
So how does this compare with the length of the day measurements against angular momentum?
[Response: I don’t know. Ordinarily, this is exactly the kind of question I would welcome. I might look at the data myself, or other readers more knowledgeable than I am might comment and I might learn something.
But this post has a specific purpose. It’s an important one. Comments must remain on topic.]
LB, I know exactly what you are referring to regarding LOD, so send me a message.
I admire your determination to teach DB how to comprehensively evaluate sea level rise. I hope that DB will learn something from your efforts. I think science will have to overcome ideology.
Interesting that Dave Burton seems to believe that in the past 120 years there has been only 4 El Niños and not a single La Niña.
interesting also that he attempts to use “cycles” TWICE to attempt to disavow a deeper understanding of the true nature of things: Once by giving us a Fourier solution which means nothing and once to invoke “sloshes” well at least rising sloshes but not negative sloshes to do the same which also means nothing.
As I said in words, and as tamino expressed ever so much more quantitatively here: Irregular, periodic positive and negative cycles simply don’t have anything at all to do with the trend. All they do is to allow a denier to attempt to confuse the issue.
Regarding irregular sloshes not effecting trends. If you purposefully pick trend periods starting on an ENSO up slosh(eg near 1998) and ending on down slosh (several in next 18 years) then it forces a lower trend in temps.
“If you purposefully pick trend periods starting on an ENSO up slosh(eg near 1998) and ending on down slosh (several in next 18 years) then it forces a lower trend in temps.”
Hence MANY denier “points”…e.g. the”Pause!”
The point stands, however if you pick a long enough period for start point/end point not to matter very much. 30 years generally works well.
Very good, Tamino. Of course, you know it will be to no avail (though there is always a chance). Dave Burton is fixated on something he read about lack of acceleration and is determined to try to justify that belief to himself (and if he convinces others, that would be a bonus for him). It’s a common ploy for deniers to avoid trying to show some real calculation is wrong and instead try to promote their own, usually weak, calculations.
However, maybe this time DB will actually try to seek the truth of the matter?
“Dave Burton is fixated on something he read about lack of acceleration and is determined to try to justify that belief to himself”
If Mr. Burton is an engineer as one comment seems to imply, that would explain a lot. In my experience, the worst offending crank “scientists” tend be engineers rather than scientists.
It even has a name: Google “Salem Effect”. RationalWiki notes that the Salem effect which was originally proposed in the area of evolution denial has now been seen elsewhere like in climate denial.
How do you correct for ENSO?
Mathematically. Put the vector of input data–here the observed SLR values–and the factor to correct for–here the ENSO values at the same moments in time–into a common metric to properly equate the 2. Subtract the latter from the former. The result is a vector of SLR “corrected for ENSO” values.
As a possibly more familiar example, think “seasonally adjusted” economic data. If one is examining say monthly economic data over many years for trends, one wants to subtract out the seasonal cycles in say employment or spending as that variation can cloud seeing the overall trend picture. Note that the longer term trends–if any–are there in the data all along. They are simply harder to see with a bunch of regular, irrelevant to the trend, short term fluctuations going on.
Which of course is why Mr. Burton “accidentally” fails to do so. Far, FAR worse, he actually only considers rather clumsily “correcting” (NONmathematically) for a handful of 4 ENSO highs and ignores the ENSO neutrals and lows all in only a few particular locations in order to convince himself of his particular position. I personally do not know whether he employs such a line of “reasoning” through ignorance or conscious design though I have my suspicions. But regardless, it is simply statistically indefensible.
I can certainly understand your frustration with Burton’s intransigence. However, I believe you are fighting fire with fuel oil. There comes a time in any “discussion” when one (or more) parties cease to listen and seek only to defend. At that point, all further discussion is futile, as knowledge has become secondary to “face.” The next thing that happens is that partisans (yours, his) cease to evaluate arguments and simply come to the defense of their hero. As the “discussion” becomes louder, the defenses burn hotter.
Time to walk away and let the ashes fall where they may.
[Response: Although I wrote this in the style of speaking to Dave Burton, I wasn’t writing it for him. I wrote it for everyone who sees his posts and graphs and claims, and needs to know how imbecilic they are.
A nice side benefit is that, as happened in the previous dust-up with Dave Burton, we might get him to shut up for a while (here, at least). Years ago he visited, posted nonsense, then started posting lots of comments, until I refuted his nonsense so thoroughly that he simply disappeared. He’s back; I guess my recent posts about sea level have inflamed his indignation. There’s more to come.]
Actually, some conflagrations are best fought using fuel oil. Many forest fires, for example.
Burn up a large fire’s fuel source before it can reach an area and the fire dies.
Deniers will similarly die off if and when their inflammable base grows so small they can be ignored as irrelevant. On that score, the tide may be ever so slowly turning (so to speak).
Thanks Tamino, I learn so much at this site, always a pleasure to read. I was just going to say that I don’t think that Tamino is writing this for DB’s sake nor for argument sake, but rather to point out the truth of the matter for the readers sake, but Tamino has already said as much above.
Deniers are a complete irrelevance, i mean they are referred to as deniers for good reason, they deny physics and that’s just ridiculous.
2019, and the arctic is in a mess, never mind DB.
1. 95% of old ice gone
2. ice free by 2021-2031
3. fastest warming place on Earth
4. no evidence sea ice will recover
5. heatwaves 20°C more than usual
6. vast wildfires mean carbon bursts
7. melting is destabilising the planet
8. permafrost collapse is irreversible https://t.co/ZpZpKQ8NM6
Thanks for explaining your purpose in refuting DB’s nonsense. I don’t think he will absorb the math and science and recognize his errors, but creating a googlable link that takes his nonsense down makes sense.
slantcadence, you seem to be treating Burton’s and Tamino’s arguments as being of a similar style. This isn’t the case. Notice that Tamino explains why Burton’s reasoning is flawed but Burton’s tactic is to ignore Tamino’s calculations and explanations, as he puts forward his position. Burton’s approach is what deniers do so you shouldn’t really characterise his position as similar to those explaining the science.
Mike, I was responding, more, to what I perceived as the “tone” of the post – struck me as rather strident (not that stridency is always bad). Both sides preaching to their respective choirs (Tamino with evidence based reasoning, Burton with his tactic of ignorance) accomplishes nothing more than raised volume. Set a backfire – sure…but the burnout leaves smoldering bits that pop back up sometime or somewhere else. Attack, defend, attack, defend, slink off, slink back…it is all so predictable, and so boring.
I appreciate this forum and the information it offers. If Tamino needs to get his blood up, from time to time, well…okay…I am certainly not immune from the practice. But, then, I’m an old guy, and I need to watch my blood pressure…
We’ve been having these discussions for over 20 years now. If the D and F students haven’t got it by this point, they are probably irredeemable. May as well ridicule them and get some fun in the bargain, since they are unwilling to learn.
It would be interesting to have high resolution data for glacial rebound and subsidence.
Maybe Tamino would then extract the stuff out of sea level data just like he did here with the ENSO signals…
Nice to see proper analysis of the Kwajalein data. I guess I could say it’s nice to see proper analysis, period. The eyecrometer thing has limits. I was looking forward to DB’s response but so far it is rather lacking…
Here’s one paper showing land subsidence monitoring along the US east coast:
Subsidence is driven partly by post glacial isostatic rebound but also by groundwater extraction. People are monitoring it, including the US Geological Survey. However, it is politically problematic to say it in public.
Can’t begin to figure out the databases needed but offhand I do wonder if EVERY coastline in the world subsided say 15 cm and all rise was inland inside continents just how much that would affect SLR globally across all the oceans?
I just don’t see coastal subsidence or rise as a key factor re. global SLR. But this is in no way any area of my own personal expertise.
Mitch – and by the way, Tamino
It’s nice to write about the US East coast, but… there are over 1500 PMSL tide gauges worldwide.
Here is an example of what I hope to obtain from Tamino ‘s amazing data processing.
PMSL station 203 FURUOGRUND (1916-2017), located in a postglacial rebound corner (the Bothnian Gulf between Sweden and Finland)
experienced a change in sea level trend from -8 mm/yr for its whole life time up to -5 mm/yr for the period 1993-2013.
The question is here: where did these 3 mm/yr come from?
This you can answer only if you manage to extract the gauge-specific amount of rebound out of their sea level time series.
That’s about right, and one useful technical approach is to use multiple linear regression, so that other time series (volcanic, TSI, etc) can be compensated simultaneously.
“… and one useful technical approach is to use multiple linear regression…”
Yeah. Even a simple layman knows that. But as opposed to professionals, he has no R let alone the Matlab stuff or similar in his toolbox.
I was answering Raymond Horstman above who was asking “How do you correct for ENSO?”, but my response got separated from the thread. Agree that it is a well known technique.
Bindidon, You were asking whether land elevation change was being monitored. The point is that there is monitoring. If there is a geodetic station near PMSL station 203 FURUOGRUND, You would know the change in land elevation and would know the relative sea level change. You would not have to guess
To guess was not at all my intention!
I found what I needed
Moreover, the 2nd SL reconstruction made by Jerevreva & al (2014) integrates Peltier’s GIA:
and here is the data
I’ll compare that with C&W.
[Response: I believe I have identified a serious problem with the “first-difference method” on which Jevrejeva et al. rely. See this.]
Many thanks for your reply, which surprises me a bit.
I read your impressive evaluation of Jevrejeva last year (but not with the necessary attention, having been at that time more fixed on temperature data sets).
Of course I don’t intend to contradict you here! For that I obviously lack both your knowledge and experience.
What nevertheless wonders me is that after having downloaded Jevrejeva’s data from the link below (column 4):
and shifted in my spreadsheets all altimetry and gauge anomalies wrt the mean of 1993-2009, I obtained this for a comparison with Church & White (CSIRO data) for 1880-2009:
and this for a comparison with Church & White and NOAA’s sat altimetry data for 1993-2009:
For 1880-2009, the linear estimates were in mm/year
– C&W: 1.55
– Jev: 1.92
For 1993-2009, they were
– C&W: 3.21
– Jev: 3.12
– NOAA: 2.58 (NASA: 2.64)
In comparison to C&W, Jevrejeva’s PMSL evaluation for 1880-2009 looks a bit like the denialists’ preferred sentence ‘They cooled down the past to make the present warmer’. Maybe they had in mind rather Jevrejeva’s 2006 evaluation? (I didn’t compare it with her 2014 edition.)
Did I make something wrong?
You don’t seem to realize that the Peltier and Jevrajeva products are models of GIA. The models are decent but not necessarily locally accurate. This is why there is major effort to ground truth the actual land surface change in elevation, e.g.
This is not only important to measure sea level change, but also to adjust the GRACE gravity estimates of ice loss.
Oh… I don’t seem to ‘realize’ ! Thanks for this rather redundant comment. You won’t believe it: I really read information, e.g. out of
The files give the present-day rate-of-change of relative sea level and crustal uplift as predicted by a GIA model. The two ascii files refer to the locations of tide gauges in the PSMSL data set. For general interest there are also two netcdf files with values on a one-degree grid.
The ice model used throughout is ICE-5G v1.3. The earth models is VM2 with a 90 km lithosphere.
Furthermore, your ‘hint’ unfortunately is off topic: what matters in my comments above is that I wondered why the ‘skeptic’s (quote needed) appreciated Jevrejeva’s PMSL evaluation, though it gives for the historical period a higher trend than that of Church and White, an evaluation these same people often have aggressively discredited.
Um… I’ve been eyeballing the data. But Promise that I drew no firm conclusions, just found a question. Eyeballing the enso models. Am I imagining it or are the ENSO wobbles getting bigger. Say from 1970-2020 vs the 50yrs before that. and is it large enough to be real? Note Im not sure what definition of bigger i would even use/mean.
The Multivariate ENSO Index (MEI) is one of the likely suspects when folk want a measure of ENSO. A recent update has brought us MEI v2 which only spans the years 1979-to-date. Over this period there is no sign of any long-term trend.
There is also Extended MEI which runs 1871-2005 and this does show a trend in wobble-size (I’m looking at the standard deviation of the index over a 12-month period). There was a falling trend for about 1900-35 and then a rising trend 1935-80.
But do note that the old MEI v1 only ran back to 1950, likely because the underlying data was getting scarse in those early years.
If you want to access MEI’s old revision, use this
The data (until Dec 2018) is here:
last similar post had formatting artifact.
Ta. for the data source. Those will be what I am shortest on knowledge about. I had a quick, spread sheet poke around, using nino3 from somewhere. Using a “simple” subtract the overall mean, abs the anomally, sliding window about 15 years, and “something” came out of the noise. Experienced Eyeballs warned me, mean of a constant noisy linear trend (when analysed like that) would do a dip in the middle(zero crossing) too. And the data is not dead flat. Still pondering what (my eyballs say)”do the wiggles get bigger? means” and how I want to measure it numerically & meanignfully.
Sliding window SD maybe. vary windows sizes and hope intellectual honesty should keep me safe from artifacts. (notice if something real arises)
Not sure Id know what to conclude except that it is curious, or that is just what happens when you turn a tap(driving ocean current) on harder or slower. (it (the dribble stream in the sink) wiggles more or less fast)
perhaps: I might also subtract sliding window av, & subtract the linear fit within that, leave all other ‘frequencies’ intact. SD the data. If some aperiodic signal is increasing and decresing its amplitude other ‘frequencies’ should tell the tale. This is fun. Especially as when I get an answer it won’t mean anything to me. (BTW yeah DFT/FFTs are fun too)