Sea Level Acceleration since the 1960s

Dangendorf et al. have made a new estimate of global sea level since the year 1900, based on data from tide gauges around the world. I’ve compared it (which I’ll call Dang) to the most trusted data set (in my opinion) from Church & White (which I’ll call CW), to the dataset I have criticized from Jevrejeva et al. (which I’ll call Jev), and to my own reconstruction (which I’ll call me) based on my own method of correcting for VLM (Vertical Land Movement). The first thing to note is that my own data doesn’t include proper area weighting, and can only be considered seriously flawed. But it is my own, so we’ll see how the new kid on the block compares to it, as well as to well-known data sets. Here’s the new data from Dang:

The method they use is a hybrid of two methods, one designed to get a better estimate of the trend, the other for better estimates of year-to-year fluctuations. I’m not sure yet how I feel about their methods, but I will say that when they compare their results to satellite data (since 1993 when we have both) the match is impressive. But this post isn’t about thier methods, it’s about their results.

The new data show a very steady trend, but with the same accelerations and decelerations as other data sets, albeit with different magnitudes and timings. Here are yearly averages for each of the four data sets based on tide gauges (aligned to have the same average value after 1993):

Although my own reconstruction has a larger noise level than the others, especially before 1950, its overall trend seems to match the new data from Dang, better than CW or Jev do.

I used a lowess smooth to estimate how the rate of sea level rise has changed over time, for each data set. Here’s what I got:

The most obvious “outlier” is the data from Jev. They show a high rate of sea level rise in the early-to-mid 20th century which is not shown by the others.

All reconstructions show a rate that is rising consistently since around 1990; a rise in the rate is acceleration. The biggest difference is that while the CW and Jev data show acceleration consistently since around 1990, the data from Dang and from me show it starting before that, in the 1960s.

In fact after 1960, the new data set very closely follows a constant acceleration curve, also known as a “parabola” or “quadratic.” For those who insist on such things, yes, it’s statistically significant. Way.

We’ve known for some time now that sea level rise is accelerating upward at ever-faster rates. This has been cofirmed since 1993 using satellite data, and appears also in the data from CW and from Jev. We now have evidence that sea level has been consistently accelerating since the 1960s.

The next step for me, is to look at individual tide gauge records with a focus on the period since 1960. It will be quite difficult to confirm acceleration even if present, for two reasons. First, the noise level is very high in tide gauge records. Second, restricting analysis to the time since 1960 leaves precious little data to work with (given the low signal-to-noise ratio), making identifying acceleration even more difficult.

I suspect that acceleration in tide gauge records is strong enough that some of them will show the signs clearly, despite the obstacles. The real challenge will be differentiating between “sudden acceleration in 1993” and “consisten acceleration since the 1960s.” It’s the kind of challenge which might be fun.

This blog is made possible by readers like you; join others by donating at My Wee Dragon.


10 responses to “Sea Level Acceleration since the 1960s

  1. The full text of Dangendorf et al (2019) is not immediately available on-line (it is ‘on request’ via Researchgate) but there is coverage of the paper by CarbonBrief which includes a trio of minute-long YouTubes with Dangendorf.

    [Response: Thanks for the link. I got the paper on researchgate, but didn’t know about those videos.

    Maybe there’s *some* “sudden acceleration” around 1993 after all, because that’s when ocean heat content accelerated suddenly. But the Dangendorf data (and mine) don’t really show that, so now I have to wonder.]

  2. I will refer him to this blog, he is in our civil engineering department. Maybe he likes to comment

    [Response: Thanks.]

  3. As I think I’ve commented before, I’ve liked the Hays approach (which Dangendorf is based on) for a while. A few advantages:

    1) I like the physical pattern matching
    2) The approach can use more tide gauges (particularly at high latitudes): though Hays et al. also showed that their method is robust when using station sets consistent with those used by the other authors
    3) The approach can predict tide changes at arbitrary locations (which means that it can be tested by removing a station from the calibration, and then seeing how well it is predicted)
    4) The GPR approach (though not the Kalman) can also allow for non-stationary vertical land movement (e.g., due to tectonic events).

    Given that your approach seems to be more similar to the CW approach, I’m surprised that the results more closely resemble Dangendorf… given that you don’t include area weighting, I’m not sure I’d consider this extra weight for Dangendorf yet.

    I’d be interested in seeing your quadratic fit to Dangendorf extrapolated to 2100 (not that it is a good way of estimating future SLR, but given 50 years of fairly constant acceleration, it is an interesting question).

    Thanks, as always, for interesting posts,


    [Response: I too was surprised by the close match of my own estimate to Dangendorf. And, I too think the lack of any area weighting in my reconstruction means it can’t really be thought of as “endorsement” of the new estimate. I’ve been playing with new alignment methods more than I’ve been seriously estimating global sea level change.

    The unique aspect of my method is to estimate vertical land movement based on the tide gauge data itself. Perhaps it offers a genuine advantage. Still, my instinct is that their hybrid method is a better approach; I haven’t yet studied the details sufficiently to form a more thorough opinion.

    Extrapolating the parabola leads to 475mm sea level rise from the year 2019 to 2100.]

  4. The quadratic fit gives an additional 5m sea level rise in 2010-2100.

    [Response: No, I think it’s 0.5m.]

    • nzcpe

      “The quadratic fit gives an additional 5m sea level rise in 2010-2100.”

      Could you please cite your source(s)?

      On the page

      we see:

      Global mean sea level rise during the 21st century will very likely occur at a higher rate than during the period 1971–2010. Process-based models considered in the IPCC AR5 project a rise in sea level over the 21st century (2100 vs. 1986–2005 baseline) that is likely (i.e. 66 % probability) in the range of
      – 0.28–0.61 m for a low emissions scenario (RCP2.6) and
      – 0.52–0.98 m for a high emissions scenario (RCP8.5).

      The page refers to even higher levels (up to 1.5 – 2.5 m), but this info is not directly related to any scenario.

  5. Everett F Sargent

    A public domain copy of this paper is available here …

    Click to access s41558-019-0531-8.pdf

    I mentioned this paper back in the “Sea Level Rise” thread on 2019-10-08 …

    The SOM/SI is free per the direct link to the paper in the post above.

  6. David B. Benson

    Moving a lighthouse to save it, for a while, from the North Sea:

  7. Tamino

    Thanks for continuing this sea level article sequence.

    I’m happy that you decided to make an own evaluation of that stuff. I would greatly enjoy you providing us with a link to a monthly time series of your evaluation, so people interested in doing a similar job would have the chance to compare their results with yours in a spreadsheet rather than by simple eye-balling.

  8. Sönke Dangendorf

    Holger has forwarded me this link. Very interesting article, Grant. My personal believe is that indeed methodological differences between the different reconstruction approaches are comparably small and that the majority of the differences is explained by the different handling of vertical land motion. The latter is also the result of our 2017 paper (, in which we use (more or less similar to Jev) a virtual station approach, but accounting for GPS-based vertical land motion at individual locations. The results are, on a first view, very consistent with your average presented here. An indeed big advantage of the Kalman Smoother, as it has been used in Hay et al. (2015) and our newest hybrid estimate (Dang), is that it contains a residual term with a certain spatial length scale, which accounts for local processes such as vertical land motion unrelated to GIA. Therefore, although the information is not coming directly from GPS observations, it takes potential vertical land motion into account.

    We are currently working with a large number of experts on a comparison of the different published techniques in ocean reanalyses and climate models as well as with coherent observational datasets ( and we’ll going to submit the results very soon, so stay tuned.:-)