Christine Shepard has recently published her first peer-reviewed paper as lead author, Assessing future risk: quantifying the effects of sea level rise on storm surge risk for the southern shores of Long Island, New York (Shepard et al. 2012, Nat. Hazards, 60:727–745 DOI 10.1007/s11069-011-0046-8). Such a first is usually (and it should be) a moment of triumph for a young scientist.
But because her research involves the risk of sea level rise, and because that rise is linked to global warming, Shepard’s work has come under fire — not for its results, but simply for using a reasonable result from the research of others. The paper does just what the title says: studies the impact of sea level rise on storm surge risk in Long Island, and the abstract summarizes the result:
We find that even modest and probable sea level rise (.5 m by 2080) vastly increases the numbers of people (47% increase) and property loss (73% increase) impacted by storm surge.
The whole thing is pretty non-controversial. Sea level rise of half a meter or more is supported by quite a bit of scientific literature (Rahmstorf 2007, Vermeer and Rahmstorf 2009, Grinsted et al. 2010, Nicholls and Cazenave 2010), in fact there’s reason to believe it may be a lot greater. But Shepard et al. isn’t about how much sea level will rise. It’s about what will happen to Long Island if it does.
If you disagree with the sea level forecast, take it up with Rahmstorf and Vermeer and Grinsted and Nicholls and Cazenave.
Unfortunately, the editor of the journal Natural Hazards, Tad Murty, has described global warming as “the biggest scientific hoax being perpetrated on humanity.” He’s one of the authors of “Independent Summary for Policymakers IPCC Fourth Assessment Report” from ultra-conservative “think tank” the Fraser Institute. The report is so outlandish, it claims such things as “Arctic sea ice showed an abrupt loss in thickness prior to the 1990s, and the loss stopped shortly thereafter.” Remind you of anything?.
That may be why Shepard et al. generated so many comments in the journal — three (so far) — and it may be why the comments were published in spite of clear indications of faulty science. Did I say “faulty”? I meant “the absence of.”
There’s a common theme to all three comments: that sea level hasn’t accelerated over the last century or so, so the models used for forecasting are wrong, and that the only acceptable projection to use is the linear trend from the available past data. Does that remind you of the idiotic legislation proposed in North Carolina? The bill that was pushed by “NC-20,” an economic development group? It should. The first comment came from David Burton, who just happens to be on the board of directors of NC-20. The other comments come from A. Boretti and Albert Parker, Australians who specialize in … internal combustion engines.
The argument has four severe problems.
First, sea level rise has accelerated over the last century or so. Although that’s the overall pattern, in detail it has shown a complex pattern of accelerations and decelerations, something which Burton, Boretti, and Parker seem to be blissfully unaware of.
Second, there is good reason to expect dramatic acceleration throughout the 21st century, based on something we like to call “physics.”
Third, when it comes to time series, anyone who knows the difference between a sphincter and a hole in the ground also knows that for complex physical processes, making a long-range forecast by extrapolating a simple trend far into the future is a fool’s errand.
Fourth, perhaps the most severe and most glaring problem with the arguments of Burton and Boretti and Parker and their ilk is that it’s predicated on a false premise: that sea level rise must have accelerated in the last century or so or all the models used for forecasting are wrong. It’s also predicated on the childish notion that the over last century or so, sea level must have accelerated consistently rather than show a complex pattern of changes which the best models (based on something we like to call “physics”) are in good agreement with. Their argument shows the lack of comprehension — nay, the outright rejection — of the complexity and subtlety of the very phenomenon of sea level change.
There’s not a word about changes of land ice in Greenland and Antarctica, or of Alpine glaciers, or the thermal expansion of seawater and the impact of much greater temperature increase throughout the 21st century. There’s no consideration of the fluctuations induced by constant back-and-forth movement of water between land and sea (witness the tremendous transfer of water to the continents in recent mega-flooding events, a process confirmed by the results of the GRACE satellite mission). There’s no understanding of, or apparently even knowledge of, the variety of time scales on which sea level can change, the ubiquitous decadal fluctuations which invalidate a simplistic view of how sea level is changing. Instead, we are confronted with the simpleton’s attitude — that if sea level hasn’t continuously risen at an ever-increasing rate, then all the models are wrong and the impact of global warming is not a problem. It’s just as naive as the idea that we could project the future by extrapolating a simple trend line on time scales of a century or more. Expecting any shred of nuance, or heaven forbid, complexity is asking too much of them. Seriously, you could get more sophistication from a bright middle-school student’s science report.
Furthermore they fail to mention, and apparently to comprehend, that in the last century-and-a-half or so sea level has both accelerated and decelerated. This is no way contradicts what we expect from global warming; the idea that “it must be ever accelerating” which so permeates the comments betrays complete ignorance, even disdain, for the complexity of reality. Tectonic shifts, the natural hydrological cycle, significant human impact on the hydrological cycle, and yes even global warming, can create a much richer and more varied response than they seem able to wrap their minds around. But Burton, Boretti, and Parker have a serious case of “Bastardi’s disease” — if reality is even the least bit complicated, if it can’t be summed up in a 10-second (or less) sound bite, if it involves any amount of “take some time and think about that before moving on” effort, it seems too much for them.
David Burton’s comment on the Shepard et al. paper begins thus:
At its western tip, Long Island meets Manhattan Island, at a place called The Battery. There is an excellent GLOSS-LTT tide gauge there which has been measuring sea levels since 1856.
Due to local land subsidence, sea level is rising faster at The Battery than at 85% of the other GLOSS-LTT tide gauges in the world, but the rate of rise has been nearly constant for over a century, at 2.77 +/- 0.09 mm/year (95% confidence interval).
About 32 Km away, at Kings Point, on the north shore of Long Island, there’s another good GLOSS-LTT tide gauge, which has been measuring sea level since 1931. Sea level there has been rising at only 2.35 +/- 0.24 mm/year.
Despite over 2/3 century of major anthropogenic CO2 emissions, these tide gauges have measured no statistically significant increase in the rate of sea level rise.
It’s the opening salvo of the “sea level hasn’t risen at an ever-increasing rate so all the models are wrong and the impact of global warming is not a problem” meme. And it isn’t even right.
As at most locations, sea level data from New York (the Battery) shows an annual cycle superimposed on any trend present. The first thing to do is remove the annual cycle to compute anomaly rather than raw data. Then we’re better able to estimate the linear trend, and to look for possible changes in the trend. The graph, which is from NOAA, already has the seasonal cycle removed.
If we fit a quadratic trend to the anomaly data, or equivalently remove the linear trend from the anomaly data (to clarify the changes visually) then fit a quadratic trend, we find overall acceleration.
Is that acceleration statistically significant? Why, yes it is, at greater than 99% confidence. Yes, I accounted for autocorrelation in the analysis. Burton’s claim that there has been no acceleration of sea level rise in this tide gauge record is false.
We can even compute the linear trend for different time spans. Here it is, together with its 2-sigma (about 95% confidence) uncertainty range, for time spans ending with the final data values but starting with each year from 1856 through 1990:
Note that the recent trend is faster than the “linear trend of the entire time span.” A lot faster. And the increase, once again, is statistically significant. It puts the lie to Burton’s claim that “the rate of rise has been nearly constant for over a century.” No. It hasn’t.
Why didn’t Burton mention any of these facts? Did he botch the analysis of data from the Battery? I suspect that he simply didn’t do any. In fact I’m rather confident in that conclusion, because the linear trend rate he quotes in the text is the one straight off the NOAA graph, 2.77 mm/yr. But if you look at the text below the graph you’ll note that the given trend is based only on the data through 2006, but the available data go up to December 2011. The linear trend rate including the most recent data is 2.83.
All Burton did was show a graph — which he didn’t even produce himself — and make a false claim based on no analysis. This isn’t so much faulty science as it is the absence of science.
Characterizing changes in sea level rise only by looking at a quadratic fit amounts to assuming constant acceleration, so it’s an extremly crude way to study the issue. The best you can hope for is a rough indication of the long-term pattern, and even that is all too easily obscured by the high noise level in the data. If we compute a lowess smooth of the detrended anomalies, we can see that the pattern of changes has actually been quite complex:
Plotting the smooth alone shows just how extreme are the changes:
It also shows just how rapid has been the most recent sea level rise in New York, and again puts the lie to Burton’s claim that “the rate of rise has been nearly constant for over a century.” No. It hasn’t.
Rather than attempt to take anything like a nuanced view of sea level changes, Burton chooses to wallow in the simpleton’s view by referring to this:
The same thing is true at most other tide gauges around the world. In fact, the best and most comprehensive analyses of sea level measured by tide gauges around the world show slight decelerations in the rate of sea level rise over the last 80 years. [Houston, J.R. and Dean, R.G., 2011. “Sea-Level Acceleration Based on U.S. Tide Gauges and Extensions of Previous Global-Gauge Analyses.” Journal of Coastal Research, 27(3), 409–417, DOI:10.2112/JCOASTRES-D-10-00157.1]
Houston & Dean? Yes, that paper. In my opinion it’s one of the worst travesties to appear in the scientific literature, and as long as Burton maintains that it’s among “the best and most comprehensive analyses,” his credibility goes down the toilet along with the Houston & Dean paper.
Burton does discusses the global data sets based on combining tide gauge data from Church & White (the most up-to-date is described here). His exposition is a how-to guide for cherry-pickin’. IF you use a selected version of the Church & White data (not the most recent version), and IF you carefully cherry-pick your start year, then you can claim the absence of acceleration.
If you want to know how cherry it is to use a start year selected for its result, and how it affects the statistics, then read this. Carefully.
Here’s the latest global sea level dataset based on tide gauge data from Church & White (2011):
When looking for acceleration, greater visual clarity results from subtracting a linear fit from the data, then plotting the residuals — often called “de-trended” data but more precisely “linearly de-trended.” Here they are:
Yes, it has accelerated. And decelerated. Several times. The overall pattern is one of acceleration — and the most recent decades show the fastest rise of all
Let’s again compute the linear trend rate for time spans starting with all possible years and ending with the present day. I’ll insist on at least 10 years of data, so here are the rates for all start years from 1880 through 2000 (error bars are 2-sigma and compensated for autocorrelation):
Note that the recent trend is much faster than the “linear trend of the entire time span.” About twice as fast. The “faster than” is statistically significant. In fact the most recent central estimate of the trend in the Church and White data is about equal to the recent trend in satellite data — around 3 mm/yr, compared to the overall trend of 1.5 mm/yr. Imagine that.
By the way, those who are interested in acceleration of sea level rise on longer (millenial) time scales, as well as some of the further complexities regarding future sea level rise, may enjoy this presentation by Jerry Mitrovica.
Clearly, Burton’s claim that sea level rise has not shown acceleration is just plain wrong. Yet he so easily convinced himself of it, and in my opinion there’s a simple explanation — he doesn’t want it to be true. His interest is in economic development, and any inconvenient scientific truth which threatens to increase the cost of that development isn’t allowed to be true. Evidence or not.
We’re interested in actual science, so let’s look closely at the smoothed version of the linearly detrended Church & White data:
The overall (century-long) pattern is acceleration, plain and simple. But on decadal and even multi-decadal time scales, the data show both acceleration and deceleration, repeated ups and downs that violate the simpleton’s model. This is a clue that sea level rise is more complex than simple linear trend or constant acceleration. It’s that pesky “nuance” and “complexity” thing again. But if sea level doesn’t follow a strictly, perpetually accelerating path, Burton and others will suggest that the scientists who study the impact of man-made global warming have grossly misjudged the whole phenomenon. Such a suggestion is appallingly naive — it’s the simpleton’s viewpoint of Burton and others which is grossly misguided.
The scientific thing to do is to try to understand these changes using — oh, I don’t know, how about … physics. Let’s compare the simpleton’s viewpoint to that of Vermeer & Rahmstorf (2009). Burton takes a pot-shot at such efforts when he says:
Yet Shepard, et al, postulate 20 inches of sea level rise for a shorter 69 year period (and call it “modest and probable”), based on models which indicate that sea level rise should be accelerating in response to higher atmospheric CO2 levels. But the data indicate that those models are wrong, sea level rise is not accelerating.
I suspect this is his attempt to smear the results by association, since fake skeptics have made it a point to smear the results of computer models (i.e., computer simulations). Not only are they wrong about computer models, the “model” of Vermeer & Rahmstorf isn’t even a computer model. It’s a simple (but not simplistic) physical model based on the radical notion that sea level rise is influenced by temperature.
Vermeer & Rahmstorf fit a physical model of sea level rise to temperature change, finding that
When applied to observed data of sea level and temperature for 1880–2000, and taking into account known anthropogenic hydrologic contributions to sea level, the correlation is > 0.99, explaining 98% of the variance. For future global temperature scenarios of the Intergovernmental Panel on Climate Change’s Fourth Assessment Report, the relationship projects a sea-level rise ranging from 75 to 190 cm for the period 1990–2100.
Three things should be noted. First, the correlation of their model output with observations, in excess of 0.99, puts the lie to Burton’s claim — the foundation of his (and others) insistence on ignoring everything but the existing linear trend — that “the data indicate that those models are wrong.”
Second, the Vermeer & Rahmstorf physical model shows both acceleration and deceleration of the rate of sea level rise. Just as the observed data show.
The thin red line shows annual average observed sea level data, to which Vermeer & Rahmstorf have done what actual scientists who want to understand reality do — they’ve corrected for known sources of variation, tectonic and reservoir effects. The thicker red line shows the smoothed version of same. The dark blue line shows the result of their model, which reveals that both acceleration and deceleration are features of the 20th century in this physical model. Also shown is the estimate using only the first half of the data (green) or the second half of the data (light blue). Note that the model result (dark blue) almost obscures the observed sea level (red) because of the close match.
It should be taken as a lesson in understanding, that even a simple physical model can reproduce both the acceleration and deceleration observed in the sea level data. It’s really only the first steps toward nuance and complexity — their model doesn’t include a lot of things, such as the observable impact of el Nino on sea level fluctuations — but even the most rudimentary of physical models not only reproduces obvious features that linear or constant-acceleration models can’t, it also fits observations better.
Third, projecting their model into the future they estimate sea level rise 1990-2100 of between 0.75 and 1.9 meters. That’s quite a bit larger than the actually rather conservative value of 0.5 meters used by Shepard et al.
As for the third severe problem with the comments, it really is true that when it comes to time series, making a long-range forecast of a complex physical system by extrapolating a simple trend far into the future is a fool’s errand. Let me give you an example.
For several years now I’ve forecast the annual minimum in Arctic sea ice by computing the existing trend and extrapolating that trend into the future. One year into the future. But I’ve said repeatedly that such extrapolation can only be considered valid in the short term. Suppose we had projected Arctic sea ice decline by extrapolating an existing trend into the more distant future? Let’s take the annual minima of Arctic sea ice extent from 1979 through 1999, and use it to forecast 2000 through 2012. I’ll use the annual minimum of daily extent data from NSIDC. Here’s the data used to compute the trend, for which I’ll choose a quadratic regression to allow for acceleration:
Now let’s compute the extrapolation:
Notice that for the immediate future, up to about 2006, it’s not too bad. But for even a modestly greater time the projection falls flat on its face. The actual decline in Arctic sea ice was way faster than the existing trend suggested. If we look just at the errors from this simplistic model we see this:
By the time we get to this year — a mere 12 years into the future — the projection is off by 2 million square kilometers.
Yes, projections of future sea level rise really should be based on physical models, not on extrapolation of existing trends whether they include acceleration or not. Yet Burton, in his comment, closes by suggesting this:
However, all is not lost for this report. Long Island will probably experience 20 inches of sea level rise by about 2200. So if they just change “2080” to “2200,” the report will be useful for projecting the impact of sea level rise on Long Island residents 190 years from now.
Not only does Burton utterly fail to realize the folly of projecting sea level rise for the 21st century based on extrapolating a simple trend, not only has he suggested that laws should be enacted preventing the state of North Carolina from officially doing otherwise, he actually suggests that this is a valid way to plan for the risk of sea level rise nearly two centuries into the future.
That’s worse than just a fool’s errand. It’s reckless.
As far as the Shepard et al. paper is concerned, the salient point of all this brouhaha is that If you disagree with the sea level forecasts, take it up with Rahmstorf and Vermeer and Grinsted and Nicholls and Cazenave. In my opinion, it is downright unethical and somewhat cowardly to use the simple application by Shepard et al. of very plausible results from others, as an excuse to attack those very plausible results from others. Why then did this happen?
I can only offer my opinion, which is this:
Burton and Boretti and Parker, and the journal’s editor Tad Murty, can’t reply directly to Vermeer and Rahmstorf and Grinsted and Nicholls and Cazenave in the peer-reviewed literature because they don’t have a leg to stand on. Based on the comment by Burton, and the similarly foolish ones by Boretti and by Parker, any such attempt would be laughed right out of the peer-review process. Instead, they used the paper by Shepard et al. as a convenient excuse to put their foolishness in the peer-reviewed literature in response to a paper which doesn’t even deal with the issue of forecasting sea level rise.
Just my opinion.