RealClimate deleted my comment.
Back on March 21st I made a comment at Realclimate about the first of my series of posts concerning Vermeer’s and Rahmstorf’s 2009 attempt (referred to as VR2009 for the rest of this post) at predicting sea level rise for the 21st century. Not surprisingly, they did not allow the comment to go beyond the “moderation” phase. That’s OK I suppose – it’s their blog. But it has come to my attention that there have been several comments by “EFS_Junior” and Phillip Machanick about my series over at Realclimate to which I should respond. And since RealClimate likes to delete my comments, I will respond here for the record.
When I entered my comment back in March I had the foresight to make a screen print of how it appeared while awaiting moderation (otherwise known as deletion at RealClimate). Here it is (click to enlarge)…
Philip Machanick calls me a lair and a fraud.
Philip Machanick transforms his misunderstanding of the fit parameters derived by VR2009 into slander against me.
VR2009 provided an equation relating sea level rise rate to temperature (see equation 2, here). This equation had parameters (a, b, and T0) that were determined by finding the best fit to sea level and temperature time series form 1880 to 2000. Philip Machanick accused me of “fibbing” (a wimpy way of calling me a liar) about the values that VR2009 determined for these parameters.
In comment #107 Philip Machanick (who says one of his interests is “science ethics“) accuses me of the “fib.” He starts right in with this…
“[W]hy is it reasonable … to set dH/dt to a constant? Note that the author’s [ClimateSanity’s] logic requires a temperature trend growing as exp(-at/b). This is a growth rate that rapidly converges to zero, unless you reverse the sign by making one of the multiplied constants (a or b) negative. Big surprise: it results in constant sea level rise when you plug it into the equation given in the paper covered here. Oh and the author fibs slightly about using the same constants as in the paper in his own calculations. In the paper, b = 2.5″
I suggest that Mr. Machanick actually read the blog post where he wrote a comment accusing me of a “fib.” This post was written by a fellow named Martin Vermeer, the principal author of VR2009. If he just scrolled up a little bit from his own comment he could read Vermeer’s own words concerning a negative b…
“I downloaded Stefan’s [Rahmstorf’s] script, modified it, did the first computations with the same real tide gauge and temperature data Stefan had used — surprise: negative b. Hmmm, strange. That was for real data from the real Earth”
Mr. Machanick states that “In the paper, b=2.5.” Well, yes Mr. Machanick, 2.5 was used when, as VR2009 states on its first page…
“The parameters are fitted to the global temperature and sea-level output from the climate model for 1880–2000 ” [emphasis added by Climatesanity]
At the risk of being redundant, for the actual temperature and sea level records Vermeer says…
“negative b. Hmmm, strange. That was for real data from the real Earth” [emphasis added by ClimateSanity]
In fact, for the real data (GISS global temperature, and the reservoir corrected sea level data from Chao) VR2009 found b to be a whopping negative 4.9. Perhaps you would have understood this if you had managed to read as far as the third page. I would think that knowing the facts would be the ethical thing before blindly accusing me of a “fib.” Would you agree Mr. Machanick?
You asked “[W]hy is it reasonable … to set dH/dt to a constant?” As I have explained elsewhere, setting dH/dt to a constant is a mere convenience. I could have set it to a time evolving function, but doing so would make the solution the resulting differential equation more complicated, but not more illuminating. The important point is that exp(-at/b) gives an increasing temperature for my hypothetical temperature scenario precisely because b is negative. With the other variables that I defined in part 2 of the series chosen properly, this exponential can yield realistic temperature scenarios which the VR2009 model render into unrealistic sea level scenarios, thus calling the VR2009 model in to doubt.
Mr. Machanick accuses me of fraud
I was confident enough of my conclusions to provide spreadsheets in part 3 of my series with my hypothetical temperature scenarios applied to VR2009’s equation relating sea level rise rate to temperature. In another demonstration of his superior ethics, Mr Machanick accuses me of fraud in comment #121, saying…
“Your colleague should download the spreadsheet if he (or she) has any numeracy at all and will find that the whole thing is a total fraud: the formulae for the graphs on the site bear no relationship to the calculations in the spreadsheet.”
Really? I urge all interested readers to download the spreadsheets here. They are not that difficult to comprehend. It will be easy for any person of average intelligence and high school level calculus to see that the “calculations in the spreadsheet” are accurate representations of “the formulae for the graphs on the site.” Claiming that “the whole thing is a total fraud” is a serious charge, Mr. Machanick. Perhaps you will would like to elaborate on the nature of the fraud – I will give you unlimited space on this blog to burn me with your searing logic. I look forward to your reply.
EFS_Junior blows smoke
Later on, in comment #178, Mr. EFS_Junior takes up the charge were Mr. Machanick’s superior ethics laid off. Much to my relief, EFS_Junior found “Upon further investigation” that my spreadsheet renderings of VR2009’s equation relating sea level rise (SLR) to my hypothetical temperature scenarios “will be correct…if you use a very small dt ~ 0.0001 years.” But Mr. EFS_Junior still finds fault. Let him explain in his own words…
“Upon further investigation the integration will be correct for SLR [sea level rise] and SLR (sic) if you use a very small dt ~ 0.0001 years, but still the assumed temperature curves are ad hoc, and produce ludicrous values for temperature before even the end of the 20th century…Also the backward finite difference isn’t calculated correctly (small error introduced), they should use a central diffenence with any 3- 5- 7- or 9-point stencel. I used a 9-point stencel with dt = one year (same as spreadsheets), but you don’t need to do this. Just calculate a*(Toffset – To) to arrive at the annual SLR slope per year.”
What a bunch of jibberish, Mr EFS_Junior. First point: Your supposedly important “backward finite difference” and “3-5-9 or 9-point stencels (sic)” are irrelevant (but it sure makes you sound smart when you mention them). Stencils are used for numerical approximation algorithms for differential equations. But the differential equations we are dealing with here have analytic solutions that require no numerical approximation for a solution. Those analytic solutions are rather simple, require no numerical techniques, and are laid out very clearly in part 2 of my series.
Next: It does not make any difference how small you choose dt (that is, the discrete time interval) in the spreadsheet. To demonstrate this point I have re-worked test 4, Hypothetical temperature #1, form part 3 of my series on VR2009 using dt = 0.001 years and dt = 1 year for the hypothetical part of the temperature scenario. This makes for a very large spreadsheet with over 32,000 rows. But it was worth it to make a point. You can download the spreadsheet, but be advised, it is 6MB and may respond very slowly on your computer. Or, you can just take my word for it and look at the comparison of the dt = 0.001 years and dt = 1 year versions below (click on image to enlarge)…
I guess the size of dt doesn’t make such a big difference after all, does it Mr. EFS_Junior?
Third point: Mr. EFS_said…
“But the ad hoc temperature curves really don’t do anything if you look at the spreadsheet(s) and their value for gamma (there are now 5 parts at climatesanity, but I’m just discussing the first three parts) is equal to one, this means that the first term of their eq 4 (with gamma = one as per the spreadsheets); is (1 – gamma) * and desired temperature profile (temperature profile doesn’t matter as (1 – gamma) = (1 – 1) = 0!”
Wow! Congratulations on figuring out that I was using γ = 1. I see that I can’t pull the wool over your eyes. What was it that gave me away? Was it my explicit statement in part 2 where I said…
From equation 4 we see that when γ = 1, dH/dt = a(Toffset – To ), which is a constant (d2H/dt2=0). That is, as long as γ = 1 the sea level rise rate will not change, no matter how high the temperature goes. But if γ does not equal one, then the sea level rise rate will change with time. If γ<1, then the sea level rise rate will increase with time (d2H/dt2>0). But if γ>1, then the sea level rise rate will decrease with time (d2H/dt2<0).
Or maybe it was my graphs…
Or perhaps I didn’t disguise it well enough in my spreadsheets…
But seriously, you miss the obvious when you say…
“temperature profile doesn’t matter as (1 – gamma) = (1 – 1) = 0!”
The temperature profile does indeed matter when gamma = 1. Here is the temperature profile from part 2, equation 3…
Gamma = 1 was chosen specifically to make a simple example. It has the bizarre result of keeping the sea level rise rate the same while constantly increasing the temperature. That is the reason I chose it, as explained multiple times. If gamma = 1 bothers you, then please have a look at “Part 5: Variation of Gamma”. I think you will find the VR2009’s model will make things even more bizarre when gamma does not equal 1.
Conclusion
I am not a regular follower of Realclimate, so I may miss Mr. Machanick’s and Mr. EFS_Junior’s further criticisms unless they post them in my comments. Their comments are welcome here. And if they do post them here, I promise I won’t delete them. I am running out now to buy sunglasses in anticipation of their brilliance.
Climate Sanity 