Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyses

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

Good call with the reference to Tamino. Very nice examination of the results. I was far too charitable in what I said above, though I did catch a couple of things that Tamino caught. One, that they constantly said they'd found deceleration, when in fact the results were not statistically significant -- what they found was nothing. Two, that taking a simple average of individual tide gauges is just a very dumb way to go about it (by which I mean, an incredibly inefficient estimator). Also kudos to Tamino for using what is, in effect, a non-parametric approach -- no restriction to some particular functional form for the curve fit. All you need to do is look at Tamino's graph and read the accompanying statistical analysis, and it's clear that the Houston and Dean paper is baloney. Well-chosen baloney, but fundamentally wrong.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

Well ya'know, not everyone will view Tamino as centrist.

Um, but is there a centrist on this broad topic area?

Anyway, Seems to me that Church/White have squeezed everything possible out of the combined tide gauge and sat data so far. Including the impoundments, ground water withdrawals, and continental precipitation changes. I thought it was quite thorough. Variability precludes a strong pronouncement on SL trends so far.

Up to the thoughtful readers here to conclude whether that means ' continue busness and usual and wait for undeniable changes'. Or hop on to the precautionary principle. Meanwhile govts (perhaps excluding some in Europe) are sticking with the former choice.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

Thank you for the updated information. The trend from this additional data from Church does not look like a deceleration but a slow acceleration, with greater variability. The greater variability will allow people with different view points to promote their own theories, but the evidence is against extremist views of rapid acceleration of the melting of the ice caps. I think honest scientist is more important than centrist. Does the evidence support the theory? The ice caps are melting, sea levels are rising, but a tsunami is a more likely natural disaster than the earth being swallowed up in a rapidy rising ocean.

But this is really the question, most of us readers don't have the power to instruct the biggest trend setting governments. Would money be better spent attempting to get china and India not to build massive coal power plants, or would it be more prudent to spend it planning for a 0.5 meter rise in sea levels over the next century. I would say the first is not going to happen, the second is much more doable.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

I would agree with you IF a .5 meter sea level rise was the only consequence.
Since this is only one of many consequences, doing both may be our best bet.
Don't make it more difficult though, change our own habits rather than trying to change China's.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

I don't want to leave you with the impression that we should do nothing to reduce co2 emissions, only AFAIK the solutions currently proposed will not avert those consequences. The rain patterns are changing, but the fossil record shows they changed before and small reductions in co2 will not change that.

Say the US produces X amount of CO2 and reduces it to 0.8X. If china and india are increasing by 0.3X in the same time period what will our reduction do to the total global temperature. There is an elephant in that EU global warming plan. US cutting by 40% would cut global CO2 by a little less than 8%. Europe reduced part of their production by exporting (shifting co2) manufacturing. AFAIK an 8% reduction in new CO2 production does not greatly change the climate models.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

I posted this originally to Tamino's blog, but it didn't show up for some reason. So I am reposting it here. I have several concerns about his critique of Houston and Dean, and I am interested in any responses to or critiques of my concerns. Thanks.

The reason Houston and Dean are using a quadratic fit is that they are trying to ascertain the long term average rate of acceleration. They note that Douglas (1992) has determined that it is necessary to analyze upwards of 50-60 years of continuous data in order to determine a significant trend, because decadal scale oscillations in the acceleration are common in the observed record. By significant I do not mean “statistically†significant: rather, I mean significant with respect to the determination of the type of long term trend that the climate models forecast in response to the increase in atmospheric greenhouse gases.

If we had thousands of years of data, this could be done by first decomposing the data with a Fourier series analysis, and then analyzing the low frequency signals for statistics on acceleration. But we don’t have enough data to do that, so the next best thing is to fit the data to a quadratic curve, which by definition yields the “average†acceleration rate of the entire sample space.

Houston and Dean are explicit with respect to their reasons for starting the analysis at 1930. Quoting from the paper:

“A review paper on sea-level acceleration by Woodworth et al.
(2009) notes that the analysis by Church and White (2006)
shows a positive acceleration, or ‘‘inflexion’’ point, around
1920–30. They do not use the mathematical definition of an
inflexion point as the point where the curvature (second
derivative) changes, but instead define it as a change in sea level
trend. They say that the inflexion point around 1920–30 is
the main contributor to acceleration from 1870 to 2004.
Woodworth et al. (2009) concluded there was consensus among
the authors that acceleration occurred from around 1870 to the
end of the 20th century; however, with the major acceleration
occurring prior to 1930, the sea-level rise (Figure 1) appears
approximately linear from 1930 to 2004. Church and White
(2006) did not separately analyze this specific period.â€

So, Houston and Dean analyze from 1930 forward to determine if the acceleration is a persistent (long term) trend or not. If the acceleration is a response to atmospheric greenhouse gas emissions, the trend should in theory be persistent and observable in the record since 1930.

Tamino's method of analysis consists of averaging over shorter and shorter time scales as one moves along the time axis. This effectively weights higher frequency signals more and more as one approaches the present time. This method ensures that such a plot will look more “active†on the right hand side of the plot.(i.e. the left hand side of such a plot will always have less high frequency events than the right hand side) Since the most recent data indicate a short term increase in acceleration , this ensures that Tamino's plot will appear to show this acceleration as if it were a novel event, which is misleading.

Houston and Dean acknowledge that there is a recent trend showing increased acceleration, especially evident in the satellite altimetry data. They discuss this observation at length. They conclude that this type of higher frequency event is not unique in the observational record, and that although it may represent a long-term change in the rate of acceleration, it is not possible to determine this with the limited amount of data available (again, 50-60 years of data is needed) They speculate that it is more likely just another decadal oscillation ,for several reasons they give in the paper, but they do not form any definite conclusions on this point.

The Houston and Dean analysis is appropriate for use in determining the long term trend of acceleration in the 20th century (at least since 1930) . Whether this is important information or not depends on the results of the hindcasting verification runs of the climate models . If the climate models hindcast a significant positive 20th century long-term acceleration, and if, as Houston and Dean demonstrate, such an acceleration is not present in the observational record, it gives good reason to be concerned that the process descriptions that govern the models are in error.

I don’t know enough about the climate models to know what their hindcasts indicate, so I can’t asses the relative significance of these results with respect to climate modeling.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

Hey aardvark, thanks for joining PC! Our discussions here are quite aside from the main goals of this site, but a few of us at least enjoy tossing the ideas around.

We can certainly wonder/disagree on whether H&D or Tamino has a lock on this. I rather look to Church and White, mostly because they use the entire data set, not a subset.

But when you get down to nubs, the tidal-gauge record has enough noise such that it is a difficult place to find second-order trends. The satellites would seem to offer more, but probably a decade before they can show/not show trends to everyones' satisfaction.

We can only hope that waiting for this information to arrive will not seriously hamper control/adaptation/mitigation efforts. Few to none governments now are jumping at the chance to reduce emissions of IR absorbing gases. Energy = money, and fools alone would set that fact aside.

As you are approaching this topic with fresh interest, may I suggest that you track down

Journal of Climate November 2009, Vol. 22, No. 21 : pp. 5772-5781
An Anomalous Recent Acceleration of Global Sea Level Rise
M. A. Merrifield, S. T. Merrifield and G. T. Mitchum
(doi: 10.1175/2009JCLI2985.1)

because I am simply snowed under here and haven't the chance. If it merits our attention please let us know.

Meanwhile, tell us about your Prius

(kidding there - there is no Prius ownership requirement to chat this group. Tell about your car, or bicycle, or shoes...)

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

China and India are highly motivated to halt global warming because they can't afford to let it wreck their domestic agriculture. (For that matter, can the US?) Also, China doesn't want Shanghai to drown, and India doesn't want 50 million Bangladeshi refugees swarming over their border driven by sea level rise.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

Just state your hypothesis clearly. As far as I can tell, you seem to be saying this:

"Because a quadratic fit to US tidal gauges over the period 1930 to present shows no statistically significant second-order term, the general circulation models that show global warming must be incorrect."

Start with the state of knowledge at the time the paper was written, as embodied in the IPCC 4th TAR.

5.5 Changes in Sea Level - AR4 WGI Chapter 5: Observations: Oceanic Climate Change and Sea Level

Mainly, they don't know a whole lot about this, there's a large degree of uncertainty in the measurements, and many factors affect sea level. But, with respect to acceleration of sea level rise, in a nutshell:

1) Sometime between the 19th and 20th centuries, the rate of sea level rise increased.
2) The rate of increase post 1990 is larger than the rate of increase in the 1960-1990 period.
3) Model projections agree reasonably well with the observed sea level rise in the 1960-1990 period and the 1990-present period.

If you put all this together, you will see that your hypothesis is a straw man. Nobody says that the GCMs predict a smooth acceleration of sea level over the past 70 years.

In particular, starting at 1930 means this method cannot identify point 1) above. Fitting a quadratic over the entire period means it won't identify point 2) above. (Tamino demonstrated that directly in his analysis.) The entire approach ignores the significant analysis that goes into point 3) above.

The better question to ask is this: Given that the IPCC summary was common knowledge at the time the paper was written, why would anyone proceed with that analysis?

With regard to a simple quadratic fit over this time period:

The notion that a quadratic fit "by definition" measures acceleration is wrong, as Tamino pointed out in detail. It's one way to do it. It's certainly easy. But if sea level isn't quadratic in time, then that's just one lousy parametrization of the underlying process.

In particular, if the long-term driver is surface temperature, and you know that there was a period of little temperature increase post WWII, then even under a naive model (temperature = acceleration) you would not fit a smooth quadratic curve to the data. (And, interestingly, Tamino's cubic fit shows just that -- a mid-century "flat spot" in the acceleration).

So the notion that a quadratic fit is somehow "by definition" the way to measure acceleration is just not right. It is one of many approaches that could plausibly show higher growth later in the time period. Just fitting a piecewise linear model would do that. But in particular, if you naively predict that acceleration is proportional to temperature, it's clearly a wrong choice of parameterization. We know that surface temperatures rose only slightly from WWII to about 1970. Fitting a quadratic over the entire time period runs contrary to the observed temperature data.

If you want to see how well the model predictions of sea level rise agree with observed sea level rise, well, then go look at that. It's not as if it hasn't already been tabulated.

9.5.2 Sea Level - AR4 WGI Chapter 9: Understanding and Attributing Climate Change

That's a fairly nuanced discussion. They show more rapid increase post-1990. They note that this could be natural variation. But they also note that the (IPCC ensemble of GCM) models predict that increase, and those models (largely) do not mimic the decadal oscillations observed in nature. Hence, their weak assertion that this is likely an anthropogenic acceleration in sea-level rise.

Once you've read the IPCC findings, you can understand Tamino's point in his analysis of shorter time periods ending with the present. He clearly knows the IPCC findings. He was using the tidal gauge data to show the IPCC findings, and to show that the approach of focusing on the entire period from 1930 makes you blind to that finding. In other words, he directly showed that the author's method simply can't see the finding that the IPCC has already discussed in detail. But they could have found the IPCC result if they had bothered to look. Hence his statement that boils down to the authors having been willfully ignorant.

So, I'll restate it. The IPCC summary is the state of knowledge as of 2007. Given that, what does a naive quadratic fit to individual US tidal gauge data from 1930 forward add to our understanding of this? My answer is, not much. Even if you were going to fit some parametric curve for the entire period, based on what we know about 20th century temperatures, it wouldn't be a quadratic curve. And, in particular, if you want to know whether this finding is a rejection of the prediction of the (IPCC ensemble of) GCMs, the answer is no. The IPCC already showed the agreement of observations and model projections for sea level, cited above. Further, it's the combination of the observations and the models that allowed the IPCC to state, tentatively, that the observed increase in sea level rise since 1990 is probably anthropogenic and not part of some natural decadal oscillation.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

not surprising.. warming up produces more vapor, which produces more snow fall at polar caps.

But wait when it gets warm enough to cause massive melt off at Greenland and Antarctica

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

Chogan2, thanks for the thoughtful reply. It will take me some time to investigate what you've said and respond, because it's a lot of information and I have to do my day job too. But I just wanted to let you know I appreciate the response and I'll get back as soon as I can.

 

Re: Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyse

Your summary of my hypothesis is a bit exaggerated. I would restate it this way:
"Because a quadratic fit to US tidal gauges over the period 1930 to present shows no statistically significant second-order term, the general circulation models that show global warming may require some revision. If these models do show a definite positive long-term acceleration over this same time frame, it may indicate a weakness in the process descriptions that inform the models."
I read the IPCC documents you cited, and they provide a lot of interesting information and demonstrate a lot of good and diligent work. So that’s encouraging. However, they also tell me that, indeed, the sea level rise “budget†comprising all the terms that are known to contribute the rate of sea level rise, is not closed. So this indicates that the process descriptions are known to be incomplete.
Now, of course, this doesn’t mean the models are hopelessly wrong, or that they should be ignored, or anything crazy like that. It means simply that they are a work in progress, which is utterly unsurprising for such a complicated system. But this also means that papers like Houston and Dean’s paper are useful for informing the process.
You noted 3 main summary points from the first IPCC link: I quote them here:
1)Sometime between the 19th and 20th centuries, the rate of sea level rise increased.
2) The rate of increase post 1990 is larger than the rate of increase in the 1960-1990 period.
3) Model projections agree reasonably well with the observed sea level rise in the 1960-1990 period and the 1990-present period.
I would dispute your point 3: I think it is better to say that they may agree. That is, the recent observed acceleration may be a long term trend associated with AGW (as the models predict), or it may be another decadal scale oscillation. Both the IPCC report and Houston and Dean say the same thing about this recent acceleration, although they differ in their speculations about whether or not it will persist.
All of the analysis that is involved in generating point 2 is designed to check the validity of the models and by inference to detect acceleration terms that can be associated with AGW (since the primary cause of the acceleration in the models is AGW). Again, the results are in decent agreement, but the uncertainties and speculative explanations of anomalies are so significant that no definite determination can be made as of yet. A reading of the IPCC analysis of the results says the same thing: correlation does not necessarily equal causation .
But, in any case, it is still useful to determine if a long term signal is present in the tidal record, because I l believe the models predict that such a signal should have been present. In the 19th and 20th centuries. I am inferring this from the discussion on acceleration at the end of the second IPCC link you posted. I’d love to see an actual plot of the acceleration terms that the models hindcast for the 19th and 20th centuries, and perhaps that is readily available and I should know where it is, but I don’t. So if you can point me to it, I’d be very grateful.
The only way to interrogate the existing record to see if such a trend exists is to look at least 50-60 years of data. So this is what Houston and Dean have done.
A quadratic fit of the time series yields by definition a measure of the AVERAGE acceleration of the time series, just as the linear fit yields by definition a measure of the average velocity, and a constant fit yields a measure of the average position. These guys evidently did the same thing for their analysis: (I can't post the link, but just google "Recent global sea level acceleration started over 200 years ago")

I am afraid I am beginning to ramble. The general point I am trying to make is that the paper is designed to look at the available data and determine if a long-term acceleration can be detected. They note their reasons for starting at 1930, which I quoted in my earlier post. They do not try to avoid the more recent acceleration, indeed they address it at length. But their conclusions are ultimately the same as the IPCC: we don’t’ yet know if it is a short term anomaly or the beginning of a long term trend. And, importantly, even if it is, if the models hindcast a long term acceleration and the observations don’t indicate that it exists, it would be negligent not to consider this discrepancy when evaluating the fidelity of the model process descriptions.