Posts Tagged ‘Martin Vermeer’

h1

Rahmstorf (2011): Robust or Just Busted (Part 5): Why a paper about “robustness”

September 29, 2012

This is part 5 of a multi-part series about “Testing the robustness of semi-empirical sea level projections,” Rahmstorf, et. al., Climate Dynamics, 2011. You can see an index of all parts here. I frequently refer to this paper as R2011.

I will refer to Stefan Rahmstorf’s ”Testing the robustness of semi-empirical sea level projections” as R2011 [1].

What does R2011 mean by “robust?”

What does Rahmstorf mean when he says his model linking sea level to temperature is “robust?”  Simply this: when the inputs that he deems acceptable are inserted into his model, he gets the results he likes.

How does he decide which inputs are acceptable?  Easy – if they yield the results he likes, then they are acceptable.  It is a very simple and efficient system of logic!

Why a paper about “robustness?”

Rahmstorf and his associates have a pressing need to defend their sea level rise projections.  I have presented a host of reasons why his model is bogus.  One of the most embarrassing is that one of his fit parameters, that he expected to be positive, is in fact negative for every combination of input tried.  This leads to all kinds of bizarre results (see here, here and here , for example).  The other is that his sea level projections dropped dramatically when his preferred source of 20th century historical input data updated their data set.

This “robustness” paper (R2011) is a stumbling attempt to dismiss the revised sea level data from the source that he had previously enthusiastically used.

A quick recap

Rahmstorf’s model, which I will refer to as the VR2009[2] model, attempts to relate global sea level rise to global temperature through the following formula…

where H is sea level and T is temperature.  Insert historical data for H and T,  and solve to a, b, and To.  Then insert projected temperatures for the 21st century and calculate projected sea level rises for the 21st century.  The VR2009 model and approach have an amazing number of problems and the list just keeps getting longer.  There is a whole family of realistic temperature scenarios for the 21st century that cause this model to yield ridiculous results (see here).  The root of most of these problems comes from the fact that every set of historical sea level inputs and temperatures that Rahmstorf and associates have tried result in a negative b.  That includes every set of input data considered in R2011 (see figure 1, below).

Model inputs and projections in R2011

(click to enlarge) …

FIGURE 1. R2011’s projections of 21st century sea level rise and baseline temperatures under the RCP45 emissions senario (Moss, 2010)[3] for various temperature and sea level input data sets.

I have circled the results R2011 likes.  As you can see, nothing involving the Church’s and White’s 2011 sea level data (CW11)[4] meets R2011’s  quality standard.  R2011 has determined that Church’s and  White’s 2006 sea level data (CW06)[5] is better than Church’s and White’s 2011 data, despite the fact that Church and White obviously think their updated 2011 data is better.

It comes down to To

Why does R2011 think the 2006 sea level data is better than the improved 2011 sea level data?  Well, I have already explained that – the 2006 Church and White sea level data gives the results that R2011 wants – higher sea level rise projections for the 21st century!

But they can’t really say that.  Instead they say that the 2011 Church and White data leads to a baseline temperature, To, that they insist is too low.  To is the steady-state temperature deviation from the 1950-1980 average temperature at which Rahmstorf’s model says the sea level would be unchanging.

Look at the right side of figure 1.  It shows the baseline temperature that R2011 derived with the various sets of input data.  The values of To that meet with R2011’s approval average out to about -0.43 degrees.  But those based on CW11 average out to about -0.62 degrees C.  A difference of less than two tenths of a degree.

If you were to ask the authors of R2011 what other evidence do they have that To must be about -0.43 degrees, they will refer you to “Climate related sea-level variations over the past two millennia[6],” which used evidence from two salt marshes in North Carolina to corroborate this global value.  And they have great confidence in this independent confirmation (because two out of three of the R2011 authors were also authors on this paper).  Hmmm.

I will have more to say about R2011’s preference for To in a later post.

A few input combinations that R2011 did not show you

R2011 implies that it has tried some vast universe of input sea level and temperature data combinations in their model. They say “We then compare projections of all these different model versions (over 30)…”  Wow! Count them – over 30!

But there are many more possible combinations than that.  R2011 has picked a few cherries from a very prolific tree.

In figures 2 and 3, below, I have run several temperature and sea level input data sets in my implementation of Rahmstorf’s model.  In some cases my input combinations are the same as some found in figure 1.  In some cases they are different.  I have arranged the input combinations in chronological order, with older versions of input data on the bottom.  Notice a trend?  Figure 2 and figure 3 give projections based on the RCP45  and RCP85 emission scenarios, respectively.

FIGURE 2. Sea level rise projections for the 21st century based on my implementation of Rahmstorf’s model under the RCP45 emissions scenario (Moss, 2010) for various temperature and sea level input data sets.
FIGURE 3. Sea level rise projections for the 21st century based on my implementation of Rahmstorf’s model under the RCP85 emissions scenario (Moss, 2010) for various temperature and sea level input data sets.

As you can see, newer sea level data (whether it is actually sea level (CW06 vs CH11, or reservoir storage (RS) or ground water depletion (GWD)  modifiers) tends to lead to lower 21st century projections when inserted into Rahmstorf’s model.

Which projection do I endorse? None of them.  Make no mistake – the Rahmstorf model is bogus, no matter what the inputs are.  I am just playing games with it.  The Rahmstorf model is an illusion that hooks you with a simple truth: It is a pretty good bet that higher temperatures lead to higher sea levels.  But the Rahmstorf model is not much better than a Ouija board for quantifying how much.

There is much to be said about the results in figures 2 and 3.  The 48 files below give the long story that is summarized in figures 2 and 3.

Much more to come in later posts

Sea level data: Church and White 2006
Reservoir storage: Chao 2oo8
Ground water depletion: none
Result files…
Summary: vr-summary-120923-091214.doc
Inputs: vr-input-image-120923-091214.png
Fit: vr-fit-image-120923-091214.png
Projections: vr-projections-image-120923-091214.png

Sea level data: Church and White 2006
Reservoir storage: Chao 2oo8
Ground water depletion: Wada 2010 extrapolated to 1880
Result files…
Summary: vr-summary-120923-091326.doc
Inputs: vr-input-image-120923-091326.png
Fit: vr-fit-image-120923-091326.png
Projections: vr-projections-image-120923-091326.png

Sea level data: Church and White 2006
Reservoir storage: Chao 2oo8
Ground water depletion: Wada 2010
Result files…
Summary: vr-summary-120923-091413.doc
Inputs: vr-input-image-120923-091413.png
Fit: vr-fit-image-120923-091413.png
Projections: vr-projections-image-120923-091413.png

Sea level data: Church and White 2006
Reservoir storage: Chao 2oo8
Ground water depletion: Wada 2012
Result files…
Summary: vr-summary-120923-091517.doc
Inputs: vr-input-image-120923-091517.png
Fit: vr-fit-image-120923-091517.png
Projections: vr-projections-image-120923-091517.png

Sea level data: Church and White 2006
Reservoir storage: Pokhrel 2012 extrapolated back to 1900
Ground water depletion: Pokhrel 2012 extrapolated back to 1900
Result files…
Summary: vr-summary-120923-091643.doc
Inputs: vr-input-image-120923-091643.png
Fit: vr-fit-image-120923-091643.png
Projections: vr-projections-image-120923-091643.png

Sea level data: Church and White 2006
Reservoir storage: Pokhrel 2012
Ground water depletion: Pokhrel 2012
Result files…
Summary: vr-summary-120923-091727.doc
Inputs: vr-input-image-120923-091727.png
Fit: vr-fit-image-120923-091727.png
Projections: vr-projections-image-120923-091727.png

Sea level data: Church and White 2011
Reservoir storage: Chao 2008
Ground water depletion: none
Result files…
Summary: vr-summary-120923-091904.doc
Inputs: vr-input-image-120923-091904.png
Fit: vr-fit-image-120923-091904.png
Projections: vr-projections-image-120923-091904.png

Sea level data: Church and White 2011
Reservoir storage: Chao 2008
Ground water depletion: Wada 2010 extrapolated to 1880
Result files…
Summary: vr-summary-120923-091956.doc
Inputs: vr-input-image-120923-091956.png
Fit: vr-fit-image-120923-091956.png
Projections: vr-projections-image-120923-091956.png

Sea level data: Church and White 2011
Reservoir storage: Chao 2008
Ground water depletion: Wada 2010
Result files…
Summary: vr-summary-120923-092105.doc
Inputs: vr-input-image-120923-092105.png
Fit: vr-fit-image-120923-092105.png
Projections: vr-projections-image-120923-092105.png

Sea level data: Church and White 2011
Reservoir storage: Chao 2008
Ground water depletion: Wada 2012
Result files…
Summary: vr-summary-120923-092202.doc
Inputs: vr-input-image-120923-092202.png
Fit: vr-fit-image-120923-092202.png
Projections: vr-projections-image-120923-092202.png

Sea level data: Church and White 2011
Reservoir storage: Pokhrel 2012 extrapolated to 1900
Ground water depletion: Pokhrel 2012 extrapolated to 1900
Result files…
Summary: vr-summary-120923-092330.doc
Inputs: vr-input-image-120923-092330.png
Fit: vr-fit-image-120923-092330.png
Projections: vr-projections-image-120923-092330.png

Sea level data: Church and White 2011
Reservoir storage: Pokhrel 2012
Ground water depletion: Pokhrel 2012
Result files…
Summary: vr-summary-120923-094501.doc
Inputs: vr-input-image-120923-094501.png
Fit: vr-fit-image-120923-094501.png
Projections: vr-projections-image-120923-094501.png

_________________________________

[1]  Rahmstorf, S., Perrette, M., and Vermeer, M., “Testing the robustness of semi-empirical sea level projections” Climate Dynamics, 2011

[2] Vermeer, M., Rahmstorf, S., “Global sea level linked to global temperature,” PNAS, 2009

[3] Moss, et. al., “The next generation of scenarios for climate change research and assessment,” Nature, 463, 2010

[4] Church, J. A. and N.J. White, “Sea-level rise from the late 19th to  the early 21st Century“, Surveys in Geophysics, 2011

[5] Church, J. A., and N. J. White, “A 20th century acceleration in global sea-level rise“,  Geophys. Res. Lett., 33, 2006

[6] Kemp, Horton, Donnelly, Mann, Vermeer & Rahmstorf,  “Climate related sea-level variations over the past two millennia,” PNAS, 2011

h1

Rahmstorf (2009): (part 9): Applying three corrections

November 17, 2010

This is part 9 of a series on Vermeer’s and Rahmstorf’s 2009 PNAS paper, “Global sea level linked to global temperature“  (referred to as “VR2009″ in this series of posts).

Naturally, Vermeer’s and Rahmstorf’s  conclusions were scary: oceans rising by as much as 1.8 meters by 2100.  Their results, with the imprimatur or the National Academy of Sciences, have been gleefully touted by those who crave the authority to reshape the economy of the planet to fit their more highly evolved ideals.  A google search for the title of their paper, “Global sea level linked to global temperature” yields thousands of hits.

But they were wrong.

The basic model

VR2009 proposed a model linking sea level rise to global temperature based on the following equation…

where

H is the sea level

T is the temperature

T0 is a constant “equilibrium temperature”

t is the time

a and b are constants 

VR2009 used Church’s and White’s 2006 sea level data  modified with Chao’s correction for artificial reservoir storage for sea level, H.  For temperature, T, they used the GISS global temperature .  They inserted them into the above model equation and found the values of a, b and T0 that yield the best fit.  Then they inserted their values of a, b and T0 back into the model equation and used IPCC temperature scenarios for the 21st century to determine the sea level rise for the 21st century. 

It turns out that the sea level data that VR2009 used was profoundly flawed.

Church and White sea level data update

About the same time that the National Academy of Sciences published VR2009, Church and White updated their sea level data.  The 2009 version of Church’s and White’s sea level data extended the data out to the year 2007, but more importantly, it also incorporated corrections that drastically changed the sea level versus time for the previous 100 years.  I have searched high and low for some acknowledgment of the updated Church and White data by Vermeer or Rahmstorf, but I have found nothing.

Groundwater depletion

VR2009 also gave short shrift to question of groundwater depletion.

VR20009 included the Chao artificial reservoir correction to compensate for water that would have been added to ocean depth but has instead been stored in artificial reservoirs.  They were happy to add this correction to the Church and White sea level data.  I was critical of  Chao for not including the inverse effect of artificial reservoir impoundment: groundwater depletion.  A correction for groundwater depletion would have to be subtracted from the Church and White data.    I have also been critical of VR2009 for brushing this point aside by saying   “No time series of this is available” for groundwater depletion.  It turns out that I was right – in the last part of the 20th century groundwater depletion dominated artificial reservoir impoundment.  And now a time series IS available from 1960 to 2000.

A new Geophysical Research Letters paper (Wada, Y., L. P.H. van Beek, C. M. van Kempen, J. W.T.M. Reckman, S. Vasak, and M.F.P. Bierkens (2010), Global depletion of groundwater resources, Geophysical Research Letters) provides the necessary information.  Wada provides groundwater depletion data covering 1960 to 2000.  That data fits an exponential very nicely, so I have extrapolated it backward and forward along the exponential (see here for details).

Making the corrections

Correcting for either the updated Church and White sea level data or the Wada groundwater depletion data drastically changes the outcome of the VR2009 model.  Taken together they destroy it.

In this post I will use the updated Church and White data,  a groundwater depletion correction based on Wada’s data, and the Chao reservoir correction used by VR2009 to create a superior time series for the sea level.  This more accurate time series will be used to  re-calculate the values for a, b and T0 for the VR2009 model equation.  Figure 1 shows the components of the sea level.

Figure 1. Sea level components.

Figure 2  is an overlay of the sea level data that VR2009 used, and the new, more accurate version created by combining the updated Church and White sea level data, the Wada groundwater depletion correction and the Chao reservoir correction shown in figure 1.

Figure 2. The VR2009 version of sea level data compated to the more accurate version using the updated Church and White data and the Wada groundwater depletion correction.

Look at the difference.  The VR2009 version of the sea level data starts with a lower slope than the more accurate version, but it ends up with a larger slope than the more accurate version.  In fact, the slope for the VR2009 version increases by nearly a factor of 3, while the more realistic version increases by about a factor of 1.6 (see figure 3).

Figure 3. Beginning and ending slopes for VR2009 version of sea level data and the more accurate version used in this post.

VR2009 smoothed their sea level and temperature data with a 15 year smoothing period.  I will smooth them with a 15 year FWHM gaussian filter with end reflection.  The smoothed sea level data is shown in figure 4.

Figure 4. Improved sea level data with 15 year FWHM gaussian smoothing.

Turning the crank

In a previous post I demonstrated that I could reproduce VR2009’s results with my own implementation of their model and the same data sources.  Using the same, less accurate sea level data, my results for the model fit parameters a, b and T0 were nearly identical to VR2009’s results, and easily within their margins of error.  The point is that I have accurately implemented their model, and to gain credibility when I when I make further claims about it.

Vermeer and Rahmstorf found

a = 5.6 ± 0.5 mm/year/K

b= -49 ± 10 mm/K

To = -0.41 ± 0.03 K

I found

a = 5.6  mm/year/K

b= -52 mm/K

To = -0.42 K

What happens when VR2009 is applied to the more accurate sea level data?

The new values for a, b and T0  are

a = 3.1  mm/year/K

b= -52 mm/K

To = -0.71 K

What do these numbers mean?

Everything.  This is huge.  When these numbers are inserted into Vermeer’s and Rahmstorf’s model equation, and 21st century IPCC temperature scenarios are applied, the resulting  sea level predictions are half of what Vermeer and Rahmstorf claimed.  It is just that simple. 

More details coming soon.

Martin and Stefan, I still have a lot more cards to play.  All in good time.

Follow

Get every new post delivered to your Inbox.

Join 49 other followers