Looking For Acceleration In All The Wrong Places

By Willis Eschenbach – Re-Blogged From WUWT

After considering the tide gauge records around Fairbourne in my last post, I wanted to look at a larger picture. Remember that we’ve been repeatedly told that acceleration in sea level rise is not just forecast, it’s actually occurring. I wrote about some of these claims in my post entitled “Accelerating The Acceleration“. Plus we’ve been deluged, if you’ll excuse the word, with endless cartoons and memes and movies and earnest predictions about the Statue of Liberty going underwater, cities being drowned, islands being overtopped by the sea, and the like. And not only that, but we’re assured that we can see and measure the acceleration in both the tide gauge and the satellite sea-level records.

So I went to get the satellite sea-level records from the University of Colorado. But when I plotted them up, I realized that they stopped in 2018. I couldn’t find anything on their website that explained why. Here’s their data.

Figure 1. University of Colorado sea-level record. Note that it is a splice of four satellite datasets that all seem to be in quite good agreement.

I wanted more up-to-date records, so I went to the AVISO site. That’s the French group that is keeping the original satellite records.

I did have to laugh, though, when I looked around the AVISO site and found the following graph:

Figure 2. All nine available satellite sea-level records

YIKES! I truly had no idea that it was all this bad. It seems the good folks in Colorado have simply picked some convenient records from the group above, spliced them together, and called it a valid record fit for all purposes.

I, on the other hand, would say that this is enough data to maybe give us a trend with lots of uncertainty … but teasing acceleration out of that farrago? Don’t make me laugh.

However, I figured I’d look at the AVISO “Reference” dataset. This is the dataset shown in green above. It is basically identical to the Colorado dataset, but it extends to the end of 2019. So I analyzed it.

Now, I’ve recently started to use a sea-level analysis method I developed myself. It’s based on a lovely kind of analysis called “Complete Ensemble Empirical Mode Decomposition” (CEEMD). I described CEEMD in a 2015 post called “Noise Assisted Data Analysis“.

What the CEEMD method does is to identify and remove, one by one, the underlying cycles in the dataset under analysis. And at the end of the CEEMD analysis what’s left is called the “Residual”. It’s what remains when all identifiable cycles have been removed.

Of course, the method can’t identify the cycles that are nearly as long as the dataset itself or longer. So for example, from my last analysis, I looked at 40 to 50 year long datasets. Here’s an example, this one is 44 years long.

Figure 3. A CEEMD analysis of the tidal data from Fishguard, Wales.

As you can see, this has not removed a cycle that’s on the order of 33 years long—too long to resolve in a 44-year dataset.

And this demonstrates a huge problem with trying to determine if the rate of sea level rise is accelerating. It’s well known that the tides have very long-term cycles of fifty years and more. But as I pointed out in my post called “Accelerating The Acceleration“, the people who produced the “US Sea Level Report Card” cut the tidal data short. They removed everything before 1969 … which guarantees that the signal will still contain cycles. And that, in turn, guarantees that any conclusions that they come to will be meaningless.

The other problem is that in the “US Sea Level Report Card”, they don’t even attempt to remove the tidal cycles at all. They foolishly think that you just need to check and see if the raw data is accelerating … but instead, they end up simply measuring some long-term tidal cycle or other.

With that as prologue, I decided to look at the longest sea-level records and see if there is any acceleration. We have a few of these that have 100 to 150+ years of data. This is long enough to remove most of the long-term tidal cycles. As above, I used the CEEMD method to remove the cycles, leaving just the underlying residual. To start with, I looked at the sea-level data for Cuxhaven in Germany. It’s a 176-year dataset.

So just what longer-term sea-level cycles are being removed by the CEEMD method? Here are the empirically-determined groups of cycles that make up the Cuxhaven sea level data:

Figure 4. Periodograms of the groups of cycles removed from the Cuxhaven sea level data by the CEEMD method.

As you might expect, there are a number of short-term cycles between one and five years. There is also energy in cycles that peak at eight, seventeen, and twenty-nine years or so. Note that one of the largest cycles is up near fifty years … highlighting the foolishness of a) not removing the persistent long-period tidal cycles, and b) using short-length datasets to try to determine if there is acceleration.

Finally, note that there is still some energy in cycles longer than fifty years. This is why we need very long datasets in order to determine if there is acceleration.

So what’s left as a residual once we remove all of those cycles from the Cuxhaven data? Here’s the result:

Figure 5. CEEMD analysis of the sea level data from Cuxhaven, Germany. Black/white line is the original Cuxhaven data.

As you can see, there is no sign of acceleration in the Cuxhaven sea level data. Remember that we’ve been warned for the last thirty years that sea level would be accelerating and cities would be drowning … but it appears that the ocean didn’t get the memo.

Let me demonstrate how badly folks are going wrong by using shorter-term data and not removing the underlying tidal cycles from the original data. Here’s the previous graph, plus a Gaussian smooth in blue of the post-1950 original data.

Figure 6. As in Figure 5, but with a 19-year FWHM centered Gaussian smooth of the post-1950 original data.

Now, if all that we had was the 68 years of the post-1950 data, and in addition, we didn’t remove any underlying cycles, we’d look at the blue gaussian smooth and come away firmly convinced that the sea level was running level from 1950 to about 1975, and that it had accelerated since then … none of which is true. That’s just one of the underlying longer-term tidal swings that are removed by the CEEMD method. And unfortunately, scientists around the planet are all too frequently mistaking those tidal swings for an underlying acceleration.

Unwilling to stop there, I looked at a number of the few other long-term sea level datasets we have. As you might expect, most of them are from Europe. Here’s a 170-year dataset from Wismar in Germany.

Figure 7. CEEMD residual analysis. Black/white line is the actual data. 

Again, there’s no sign at all of any acceleration in the Wismar data.

And below, without much in the way of comments, are a number of the other long-term sea-level datasets. In all cases, the black/white line with dots is the original data.

I don’t see the rumored acceleration in those plots. I’d also say that the early data from IJmuiden is very suspect … next, some data from the US.

Note the larger trend in Baltimore, which is known to be the result of land subsidence along most of the US east coast.

And to close out this section, here’s the longest uninterrupted sea-level dataset I know of, that of Stockholm in Sweden, two hundred and seventeen years long …

You can see how the earth in Sweden is still rebounding from being covered with trillions of tons of ice during the most recent glaciation. The land is actually rising faster than the ocean … go figure.

So those are the majority of the long tidal datasets. I gotta say, I am simply not seeing the acceleration claimed by the boffins. I don’t know just how they’ve calculated their results, but the best long-term datasets that we have simply don’t show the acceleration that they claim to find.

In closing, let me circle back to where I started, with the spliced AVISO satellite sea level data. Here’s what the AVISO and the Colorado folks are combining to get their final data:

Figure 8. The four satellite sea-level records chosen by Colorado and Aviso from the nine extant satellite sea-level records.

I gotta say … given that the satellite sea level is supposed to be accurate to tenths of a millimetre per year, why are there such large differences between the different satellite records?

In any case, here is the same data, with a black line showing their final dataset created by combining those four datasets.

Figure 9. The four satellite sea-level records chosen by Colorado and Aviso from the nine extant satellite sea-level records, along with their combined record which is shown in black.

Hmmm … and finally, here is the CEEMD analysis of that combined record.

Figure 10. CEEMD analysis of the AVISO / Colorado satellite dataset. It is composed of four different satellite datasets spliced together. Midpoints of the splices are shown by the vertical red dotted lines.

Now, is there acceleration in that record?

Well … regarding the question of whether there is acceleration shown in that spliced satellite record, I’ll say the three most important words that any scientist can ever say:

We. Don’t. Know.

We don’t know for a few reasons. The first is that it’s a spliced dataset, and the changes in the trend line all occur at and after the splices. Makes a man suspicious, particularly given the differences in the initial individual datasets.

The second is that the record is only 27 years long, so we really don’t have enough data to draw many conclusions. This is particularly true since the variations from a straight line are quite small.

Third, the rise was right along the linear trend line up until 2005. So there was no acceleration before that time. Then the rate of rise started decreasing around 2005 … deceleration rather than acceleration? Why? And then, according to the spliced dataset, it started rising faster around 2011. Again, why? Assuredly those three, first a straight line, then deceleration, then acceleration, are unlikely to be caused by a monotonic rise in CO2. Nor do they conform with any expected pattern of acceleration.

Finally, as with many other tidal records shown above, the satellite seems to be “porpoising” above and below the trend line. There’s no clear acceleration anywhere in the record.


The long-term tide gauge datasets are all in agreement that there is no acceleration, neither in the early nor in the recent parts of the records. Yes, they often porpoise a bit above and a bit below the trend line, but there is no evidence of any CO2-caused recent increase in the rate of sea-level rise.

The satellite dataset, on the other hand, is a splice of a selected four of the nine available satellite sea-level datasets. The changes in trend seem to be associated with the splices. Unfortunately, this spliced record is both too short and too fractured to draw any conclusions about acceleration.


Leave a Reply

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

WordPress.com Logo

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

Google photo

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

Twitter picture

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

Facebook photo

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

Connecting to %s