- JoNova - https://joannenova.com.au -

The scandal of sea levels — rising trends, acceleration — largely created by adjustments

Headlines across Australia yesterday told us the dire news that a new study finds that “Sea level rising faster in past 20 years than in entire 20th century.   A new paper by Watson et al is driving the headlines, but underneath this Nature paper is a swamp of adjustments, an error larger than the signal, and the result disagrees with many other studies and almost all the raw measurements. Paper after paper kept showing that sea levels rates had slowed (e.g Chen showed deceleration from 2004, Cazenave said in the last decade sea-levels had slowed 30% (but argued post hoc adjustments could solve that). Beenstock used 1000 tide gauges and found no acceleration of sea levels over the last 50 years. A different researcher — Phil Watson, found that Australian sea levels rose faster before World War II then slowed down.)

Firstly,  hundreds of tide gauges show sea level rising at about a third of the rate than satellites do. Worse, the original satellite raw data showed the same slow rise, until it was suddenly adjusted. The real scandal is that the rapidly rising trend was largely created by adjustments in the first place. These latest corrections just adjust down part of the rate which had been created by adjusting up. On top of all that, the long paleo-history of sea levels done by people like Nils-Axel Mörner  show that the current rise is not unusual or unprecedented at all. Could it get more pointless? It can: the acceleration Watson et al found is so small it’s less than the errors. (See the graph below).

The conclusion of the paper is that instead of the sea levels rising at 3.2mm/yr as per the official satellite data, the are rising at 2.3mm/yr + 0.043mm/yr2 of acceleration. Over a century that means the projected sea level rise is revised downwards from 320mm to 251mm. That means sea level rise on current trends has dropped off the bottom end of any UNIPCC projection for sea level rise (AR5 WG1 SPM) for the period 2081-2100, as against 1985-2005. The likely range is between 260 and 820mm under all scenarios. The projection (mid-point 400mm) range is based on succeeding in cutting global emissions to near zero before 2100.

Tide gauges don’t agree with the satellites on sea level. The 68 most stable NOAA tide gauges around the world show about 1mm a year rise. Beenstock use a thousand tide gauges around the world and found the same rise of about 1mm/year. Nils-Axel Mörner has studied arrays of gauges as well but also used the opposite approach and found practically the single most stable beach in Northern Europe. He analyzed long records on all the beaches around it to figure out which way the whole area was tilting — again he found the change of the most stable point is about 1mm/year.

We’re analysing the decimal points of the acceleration of a trend that was largely created by adjustments in the first place. Why bother? The raw satellite data showed almost no rise at all from 1992-2002, and was post hoc adjusted up from less than 1mm to 2.3mm/yr (Aviso, 2003). And the raw low rate was skewed high by the El Nino in 1997. These adjusted figures have been used to generate thousands of headlines about how sea levels are rising faster after 1992. (Anyone going to retract those headlines?) The European satellite data was also adjusted up. Nils-Axel Mörner has described the whole sordid process of sea level adjustments in detail. Knowing this puts the ABC version is a new light. Christopher Watson, lead author, “said the study suggested satellites marginally overestimated the rate of sea level rise in the first six years and that distorted the long-term picture.” He didn’t mention that it was the overestimate of the underestimate and all these numbers were subject to change, post hoc, ad hoc, as the wind blows…

Sea levels are always changing and past changes were often larger.

The rate since 2002 is slowing despite the massive emissions of  CO2: The new adjustments on adjustments bring the 1992 – 2012 rate down (did the ABC tell you that?). This changes the curve, and creates a weak acceleration that was not there before.This also creates new headlines of “acceleration”. At some point in the future, today’s measurements will be adjusted down to create more headlines of “acceleration”. Rinse Repeat Recycle.

If tide gauges were good enough to figure out the rate of acceleration from 1900 – 1992, why are they wrong as soon as the satellites start operating? Does anyone think we should compare highly adjustified satellite data to tide gauges if there are continuous tide gauge records over the same period? Its like a tree-ring spliced to a thermometer: Good PR, bad science.

The acceleration is so small it’s less than the errors. (Be afraid, it’s accelerating at 0.043 +/- 0.058 mm/yr2.) Normal scientists don’t get excited at this. They don’t issue press releases.

The Raw Satellite Data

Before adjustments:

Figure 5. Annual mean sea-level changes observed by TOPEX/POSEIDON in 2000, after technical “corrections” were applied (from Menard, 2000). A slow, long-term rising trend of 1.0 mm/year was identified, but this linear trend may have been largely an artefact of the naturally-occurring El Niño Southern Oscillation event in cycles 175-200.

After adjustments:

Figure 7. Sea-level changes after “calibration” in 2003. The satellite altimetry record from the TOPEX/POSEIDON satellites, followed by the JASON satellites. As presented by Aviso (2003), the record suddenly has a new trend representing an inferred sea-level rise of 2.3 ±0.1 mm/year. This means that the original records presented in Figs. 5-6, which showed little or no sea-level rise, must have been tilted to show a rise of as much as 2.3 mm/year. We must now ask: what is the justification for this tilting of the record?

From the new Watson paper:

Watson et al 2015 | Figure 3 | Adjusted and unadjusted satellite altimeter GMSL time series (each arbitrarily oset and corrected for ocean-basin expansion). Adjusted series use GPS-based VLM estimates (where unavailable for a specific TG, GIACElastic VLM is substituted). GMSL (annual and semi-annual periodic terms removed) is shown as cycle-by-cycle estimates (thin grey line) and after filtering (60-day low-pass Butterworth filter, thick line). Linear and
linear-plus-quadratic fits are shown as continuous and dashed lines, respectively. The inset shows quadratic  components (arbitrarily oset and symmetric about midpoint) highlighting that the adjusted acceleration is invariant to VLM treatment. Equivalent series derived from the CU data set are shown for comparison (thick dashed lines).


Reader Robbo wrote in to say he so was astonished at the ABC story, he read the paper, only to find a very different picture and problems he would fail a first year student for:

Then I carefully read the original paper, and they are completely different from the press release and the ABC version. The paper claims that the rise rate in the last 20 years is actually less than previously thought (that is not mentioned or is at best, carefully massaged by the authors’ press release and ABC piece). But it is true that the title and punchline of the Nature paper is about acceleration: sea level rise is accelerating, they say. And how much is it accelerating? Wait for it: it’s accelerating at 0.043 +/- 0.058 mm/yr2. That’s consistent with zero! I would fail a first year student claiming that 0.043 +/- 0.058 is a Nature-level result.

Finally, how do they get that acceleration result? They fit a second-order polynomial to the data (Fig 3) and take the coefficient of the t^2 term. Again, basic undergrad science, if the linear fit to the data is statistically acceptable, you take the linear fit (the lowest order polynomial that is statistically acceptable). You can always fit the same data with higher and higher order polynomials and get terms in t^2, t^3, t^4,….and of course when you extrapolate those terms to the future your fit goes wild but that is complete rubbish. In their case, all they should have said was that the linear fit is statistically equivalent to the quadratic fit (because a = consistent with 0), therefore we detect no acceleration, end of the paper.

But the ABC distills this uncertainty and the answer to hold back the seas is always the same:

“If we have major mitigation, then we can limit that rise to be somewhere between 30 and 60 centimetres during the 21st century,” [John Church] said.

He said that would require an urgent and significant reduction in greenhouse gas emissions and a big shift away from fossil fuels to renewable energy.

The ABC would never let a mining analyst give his opinion on sea levels, but when a sea level expert tells us to use windmills to change the climate, and transform our nation energy system, that’s all OK. (Sell the ABC.)

h/t  David, Robbo, Geoff, Willie, Tom, Bill, Lance, & John

UPDATE: Ruairi

Small changes in sea-level rise,
Should not come as any surprise,
But a reading adjusted,
Can’t really be trusted,
As it’s not what the reading implies.


9.3 out of 10 based on 107 ratings