Inside The Acceleration Factory

Guest Post by Willis Eschenbach

Nerem and Fasullo have a new paper called OBSERVATIONS OF THE RATE AND ACCELERATION OF GLOBAL MEAN SEA LEVEL CHANGE, available here. In it, we find the following statement:

Both tide gauge sea level reconstructions and satellite altimetry show that the current rate of global mean sea level change is about 3 mm yr–1, and both show that this rate is accelerating.

So the claim is that tide gauges show acceleration. Let’s start with a look at the Church and White (hereinafter C&W) estimate of sea level from tide gauges around the world, which is the one used in the Nerem and Fasullo paper. The C&W paper is here.

Figure 1. Church and White sea level rise estimate.

Not real scary …

However, there is an oddity. Let’s take a closer look at the C&W sea level estimate shown in Figure 1.

Figure 2. As in Figure 1, but with a different scale.

Now, when I looked at that, the curious part to me was the change in the recent trend. For the last quarter century, we’ve had satellite sea level data, which began in 1993. In the past, the trend of the satellite data (1993 – 2013, 2.8 ± .16 mm/year, or about an eighth of an inch per year) has been almost double the overall trend of the tide gauges (1.6 ± 0.14 mm/year).

But in this most recent C&W estimate, the recent tide gauge trend is much larger. How much larger? Well … a lot. In fact, the recent C&W estimate is greater than the satellite estimate for the overlap period …

Figure 3. As in Figure 2, showing trends for the 21-year periods before and during the satellite era.

Why the increase in trend? Well, since 1993 they’ve mixed satellite data in with the tide gauge data.

To combine the tide gauge and satellite datasets, … Church and White (2011) and Ray and Douglas (2011) use empirical orthogonal functions of the satellite data with principal components derived from the tide gauge records. Church and White analyze changes in sea level over time, enabling them to use many tide gauges, some with short records, without needing to relate the absolute level of different tide gauges. SOURCE

But is this approach justified? I mean, did the tide gauge data itself go up during that time, so that it would be reasonable to use satellite data to refine the results?

Now, that is a tough question to answer, because the tide gauge data is sparse spatially and temporally, and it is also affected by vertical land motion. But you know me … I’m a data guy. So I went and got the full set of 1,512 tide gauge records from the Permanent Service For Mean Sea Level. In passing let me say that I don’t think they could make it harder to collect the data. It is in 1,512 separate files. Not only that, but the so-called catalog looks like this:

Figure 4. PSMSL Catalog. It is great fun to convert this to a simple computer file … but I digress.

To highlight some of the problems with converting tide-gauge data to global sea level data, here are ten typical records in the dataset:

Figure 5. Typical tide gauge records.

I’m sure you can see the difficulties. Some places the land is steadily rising from post-glacial rebound, and it’s rising so fast that the sea levels are actually sinking relative to the land. In other places, the land is sinking due to subsidence and groundwater extraction. Many records are short and have gaps. Generally, it’s a mess.

So … here was my thought about how to get around these issues: You’ll note in Figure 3 above that the increase in trend between the 21 years before the satellite era and the 21-year overlap during the satellite era was 2.1 ± 0.5 mm per year. And while the trends in the tide gauges are all over the place … I can look at the difference in the trends for each individual dataset over the same period. This gets rid of the problem of vertical land movement, which is constant over such a geologically short time period. So here was my procedure.

First, from the 1,512 tide gauge records in the PSMSL dataset, I selected all the records that contained 90% data over the 21 year period before the satellite era and also had 90% data over the succeeding 21 year period during the satellite era. This left me with 258 tide gauge datasets with coverage over the full 42-year period.

Next, I calculated the trend for each of these datasets for the period before and during the satellite era.

Then, for each tide station, I subtracted the pre-satellite trend from the satellite trend. And finally, I got the median and the uncertainty of those 258 trend differences. Figure 6 shows a graphic of those results.

Figure 6. Comparison between the values and the errors of the difference between the 21-year trend before the 1993 start of the satellite record, and the succeeding 21-year trend from 1993 to the end of the Church and White records. The C&W trends are shown in Figure 3 above.

Since the error bars (orange and red) do not overlap, we can say that the C&W estimate does NOT agree with the tide gauge data. And that, of course, means that it has been artificially increased by cross-pollution with satellite data.

Let me close by saying that I think that it is very bad scientific practice to splice together a terrestrial and a satellite record unless they agree well during the period of overlap. In this case, they disagree greatly over the period of record. For the detrended values over the period of overlap (1993-2013), the R^2 value is 0.01 and the P-value is 0.37 … in other words, there is absolutely no significant correlation between the satellite data and the C&W estimate.

And this makes it very likely that Church and White are manufacturing sea level acceleration where none exists … bad scientists, no cookies.

Finally, at the end of my research into this, I find that I’m not the only one to notice the discrepancy …

Figure 7. Different results for the satellite era depending on whether or not the satellite data is illegitimately spliced into the tide gauge records. SOURCE

My best to everyone. Here I’m staying indoors on a rainy Sunday, watching American football and researching the vagaries of sea level …

w.

As Always: I politely request that when you comment, you quote the exact words you are discussing, to avoid misunderstandings.

Data: So that others won’t have the hassles I had extracting and collating the data, I’ve put the full PSMSL dataset as a single comma-separated values (CSV) file here, and the PSMSL catalog here.

Data 2 :Commenter Bear was unable to download the files and suggested I change the end of the link. New links to PSMSL file here, and the PSMSL catalog here.