Unlicensed Engineers, Part 1
By Hendrik Tennekes
In the series of Weblogs I am now starting, I will lay the foundations for a theory of climate software development. I am of the opinion that most scientists engaged in the design, development, and tuning of climate models are in fact software engineers. They are unlicensed, hence unqualified to sell their products to society. In all regular engineering professions, there exists a licensing authority. If such an authority existed in climate research, I contend, the vast majority of climate modelers would vainly attempt certification. Also, they would be unable to obtain insurance against professional liability.
I am an unlicensed engineer. I was an engineering professor for many years, but I never needed to be certified as a Professional Engineer. One advantage of my University position at Penn State was that I did not need to purchase any liability insurance, which would have absorbed a sizable chunk of my modest salary.
I am also an unlicensed model builder. I have many years of experience as the builder of model sailboats based on wooden shoes, and I also contributed fresh insights in the scaling laws of sailboats, but those feats by themselves are not enough to qualify for certification. My saving grace is that model sailboats are toys, with little or no impact on the welfare of society.
Two of my favorite clog sailboats are the heroes of my story today. I renamed the black sloop with red sails two weeks ago. I had christened her Hella, in part because of the hell I’ve gone through since I was forced into early retirement. But now her name is Dallas, in honor of Roger Pielke’s superb webmaster, Dallas Jean Staley. The white schooner is called the Flying Dutchman, in memory of the threemaster skipper Willem VanderDecken, who swore at Cape Horn that he would fight the fierce headwinds he encountered until Doomsday if necessary, rather than seek refuge in a nearby bay.
“Dallas”
“The Flying Dutchman”
Both clogs have a waterline length of one foot, unsurprisingly so because human feet slip into them. Their masts are about a foot long, too. Their sail area is on the order of 90 square inches, and their weight is about 2 pounds. Their “immersion ratio�, that is, their water displacement divided by the cube of their waterline length, is four times as high as in full-scale sailing yachts. This is the major fudge needed to obtain acceptable performance in the water. As a fudge, it is comparable to the excessive numerical viscosity needed in climate models in order to suppress unwanted instabilities.
My fudge consists of one pound of roofer’s lead wrapped around a length of carbon piping. Unlike full-scale flat-bottoms, my sailing clogs need this heavy keel to make them self-righting. A gust cannot topple them, no matter how strong it is. Their sails may get wet, but that does not deter them. Not all bystanders are convinced when they see me carry one of my sweethearts, but a quick experimental demonstration makes them believers. I just throw the clog upside down in the water. That quiets them. The proof is in the throwing.
As to the tuning of my models, the major problem there is that very few people understand the difference between a stabilizer and a rudder. A rudder does not provide directional stability to a ship unless it is locked in a fixed position. In the town I grew up in, fathers converted worn-down clogs into toys for their kids, and took pride in constructing a plywood rudder hung off hook-and-eye hinges. Kids liked that, naturally, because they could play helmsman in their daydreams. But this doesn’t work at all. I dare any climate modeler in this audience to provide a scientific explanation why.
My models need to be tuned. I am an engineer; I do that experimentally. I take them one by one to a large pond, a canal, or a quiet river arm, and try them out there. In earlier years, it occurred often enough that I had to sew a new set of sails, but these days I merely have to lengthen or shorten one or two sheets. For those of you not familiar with yachting jargon, sheets are the ropes by which a sailor trims his sails.
Look, I am not perfect. Occasionally, a newly-finished model on its maiden trip escapes me in an unexpected direction. Once or twice I had to come to the rescue by skinny-dipping and swimming in order to retrieve them. The laws of hydrodynamics are in my favor then: the “hull speed� of surface vessels is proportional to the square root of the waterline length, so I can swim faster than they sail. Not much faster, but fast enough.
In the second part of this series, I will analyze one of the cornerstones of engineering. It is, in Karl Popper’s unforgettable words:
“We can learn from our mistakes.�
Yes, we can, but do we?
And will climate modelers ever learn?
Those of you who are eager to prepare for Part 2, do get hold of a copy of one of the books by Henry Petroski. My favorites are “Success Through Failure – The Paradox of Design� and “Invention by Design – How Engineers Get From Thought to Thing�.
Dr Tennekes,
Quite a funny and illustrative warm-up (sort of) for your part II.
Here in France, we (try to) make model boats with bottles of wine instead of shoes. Much harder enginering work. As an aside, computer simulations show that Cognac bottle-boats are aquaeronautically outperformed by Bordeaux or Bourgogne ones.
Comment by Demesure — February 28, 2007 @ 8:30 am
I am looking forward to part II and subsequent parts. Considering the importance of these models, we all should know more about them. And monster tales are always fun.
Comment by Bob Hawkins — February 28, 2007 @ 10:32 am
Not software engineers that would get a job anywhere I have worked, if the GCM code that I have looked at is anything to go by…
Comment by Chris H — February 28, 2007 @ 10:36 am
Reading this, I am having a “using a toy catapult to teach Design of Experiments” moment. Bravo! Eagerly awaiting Part 2 ….
Comment by Steve Sadlov — February 28, 2007 @ 2:51 pm
Comment by Jaye — February 28, 2007 @ 11:37 pm
But we don’t want to license people.
We just want to appreciate and communicate the utter idiocy of the models and the silliness of the conclusions derived from them.
Comment by Graeme Bird — March 1, 2007 @ 3:00 am
Bob Hawkins, thanks for your interest. It isn’t all that difficult to get a grip on these clogs. The three key parameters are waterline length L, in feet, the sail area S, in square feet, and the dispacement D, in cubic feet. The only nondimensional performance parameter that can be made with them is D/SL.
But what does it mean? For that, you’ll have to delve into the theory of surface waves in deep water. Any textbook on hydrodynamics will help you out. A shortcut is to guess that the speed of surface waves is governed by gravity (g) and their wave length Y. The only way in which these parameters can be put together is to make the wave speed proportional to the square root of gY. In the picture of the Dallas clog you see that she’s making waves half a foot long. She feels comfortable there, because there is quite a dip in the drag curve at that speed. Consult any textbook, and you’ll discover that she runs at about one knot. If she were 36 times as large, she would run at about 6 knots, indeed typical of 36-foot yachts in a fair breeze.
Now, if the square root of gL is representative of the speeds a displacement hull can reach in view of its wave-making resistance, then the square root of gD/S must be representative of some relevant speed, too (remember, D/SL is nondimensional). Think this over. I’ll do the second half of this summary lecture tomorrow.
Comment by Henk Tennekes — March 1, 2007 @ 6:10 am
The clog models are great! My grand children would have loved them, but they are too old now to appreciate them, and in time they will be old enough to appreciate them.
I’m a modeler too. Also an RPE but I wouldn’t bet somebody else’s money on the predictions of my models. Somehow, some modelers have convinced some politicians that it is a good thing to wager 1-20% of the US GDP on their model predictions. Maybe the reason I am so poor - I am too conservative in my approach to monetary risk, but it is easier to play the high roller when it is someone else’s money.
Any climate modeler must know that any representation of the most rudimentary features of atmospheric dynamics admits chaos. I don’t think it is possible for any 2 or 3 dimensional dynamical system admitting the nonlinear coupling of momentum convection not to show chaos within all admissible sets of initial data. If that is true, then any final state can be derived from initial data that are sufficiently “near”. This is the ergotic property of chaos. Climate modelers are therefore ascribing 90% certainty or better to the solutions of equations for which virtually any end state could be ascribed the same probability.
Wagering someone else money is getting much easier for me to contemplate now
Comment by Brian — March 1, 2007 @ 10:59 am
Back to the insubstantial concerns of a clog modeler. Since D/SL is nondimensional and the square root of gL a speed, the square root of gD/S must be a speed as well. If you make a quick analysis of the heeling caused by aerodynamic forces on the sails, you’ll find that the square root of (d’/d”)gD/S is a measure for the wind speed that can be handled without taking in sails. Here, d’ is the density of water and d” the density of air. I am making that explicit here because the square root of d’/d” is about 30, quite a large number.
What do we obtain? The maximum comfortable wind speed V’ and the boat speed V are given by:
V’ = 30x root(gD/S),
V = root(gL).
I am ignoring coefficients of order one here. The ratio between the two is:
V’/V = 30x root(D/SL).
Ocean-going yachts typically have D/SL = 0.01, which allows them to sail without danger in winds three times as strong as their maximum wave-making speed (which is called the “hull speed”). This checks with the facts: hull speeds of some 7 knots allow wind speeds of some 20 knots. That’s 10 m/s or Beaufort 5.
Now for the models. If you insist on isometric scaling, you would get a one-foot model that can sail no faster than 1.4 knots and cannot stand winds stronger than 4.2 knots (2m/s). That’s no more than a gentle breeze! It would be a featherweight, though, weighing in at half a pound.
This is why I need a heavy keel under my clogs. I make them four times as heavy as the “real” thing (by now, it isn’t clear anymore what is real and what not). That way, they can handle winds twice as strong, not 4.2 but 8.4 knots. In marine handbooks, that’s on the boundary between moderate and fresh breezes, quite acceptable for joy rides on a pond or lake.
I wouldn’t dare writing all of this if I couldn’t vow that each of these findings is substantiated by actual experiments. No computer simulations were involved.
This much is needed to understand the modeling of mere toys. If I try to look at climate modeling from this perspective, I’m almost fainting.
Comment by Henk Tennekes — March 2, 2007 @ 2:23 am
Sorry, I tend to forget that I ought to do thorough proofreading before hitting the transmit button.
Where it says: that’s no more than a gentle breeze,
it should read: that’s no more than a LIGHT breeze.
And where I wrote: In marine handbooks, that’s on the boundary between moderate and fresh breezes,
it should read: (…), that’s a GENTLE breeze, quite acceptable for joy rides on a pond or lake.
It wasn’t just the conversion from m/s to knots that did me in, but also the experimental evidence. Many of my clogs still plough on in 15-knot winds, heeling as much as sixty degrees but refusing to give up. They do deserve to be named Flying Dutchmen!
Comment by Henk Tennekes — March 2, 2007 @ 3:23 am
Professor Tennekes: As one of your slower students, this lecture was passing me by. This all changed after studying the picture of the proud “Dallas”, which her magnificant red sail. Her wake is impressive as she makes mighty headway at around 1.5-2 knots. No matter that her keel is not to scale, if you had not confided this, no one would know it anyway!
If a picture is worth a thousand words, this one says it all. Thanks for holding class.
P.S. You mentioned in a previous blog that you have specialization in aeronautical engineering. Question for you:
Since computer simulations are now highly regarded as being extremely accurate, why do we still need test pilots to fly newly designed aircraft? Can’t we just build the new aircraft with engineering code and first principles, and put the test pilot in a simulator without actally going up? It would seem that with all the computing power available, the simulation might be more accurate than the real thing.
Comment by Bryan Sralla — March 2, 2007 @ 3:19 pm
Re; #9: “Climate modelers are therefore ascribing 90% certainty or better to the solutions of equations for which virtually any end state could be ascribed the same probability.”
Brian, the modelers try to get around this by posing climate prediction as a boundary values problem. It is a good idea, but does not appear to provide skill in trying to predict multi-decadal regional climate change. The conundrum in climate modeling is that as the AOGCMs get better (more closely resemble real climate), their multi-decadal predictive skill may not improve, but rather reveal that multi-decadal climate prediction problems are not in fact well-posed.
Comment by Bryan Sralla — March 2, 2007 @ 3:54 pm
Is one problem with the “boundary value” assumption that a state can be approached from either colder or warmer - meaning flora regimes could be vastly different?
Comment by Tim Clear — March 2, 2007 @ 5:53 pm
Bryan (#12), wait for parts 2 and 3. Better yet, borrow or buy a copy of the book Twenty-First Century Jet - the Making and Marketing of the Boeing 777, by Karl Sabbagh. The “Triple Seven” was designed in large part in (not: by) a large computer system, but still certain crucial tests were done in engineering reality. Read, for example, how one wing was subjected to a “static test’, where it flexed 24 feet before it snapped. Would you rather trust a computer simulation?
Comment by Henk Tennekes — March 2, 2007 @ 6:36 pm
Bryan (#11), let me add a few more thoughts. Yes, much of the test program flying is done on a simulator. Also, except for taxiing through the maze of taxiways to and from runways, all routine flying is done by computers. But would you dare fly an airliner if the pilot left the plane before take-off? And on reaching destination another pilot would climb in and roll the plane to a waiting gate? Let me tell the story of a female KLM 747 co-pilot who flew through the remnants of a volcano dust cloud over Alaska at 35,000 feet and was suddenly confronted by flame-out in all four engines. She immediately pushed the plane into a steep dive, with the air brakes almost shaking the plane apart, pulled out of the dive at 10,000 feet and managed to restart the engines. Her supervisor took over then, and made a precautionary landing at Anchorage. Apart from scratched compressor and turbine blades, no other substantial damage was found. Can you program a computer for ALL conceivable emergencies? No, you want a professional on board, who can cope with UNFORESEEN incidents.
Comment by Henk Tennekes — March 3, 2007 @ 10:36 am
Re #15: Good story. I spend a good amount of time in Alaska researching rocks, so the story of the Augustine volcano (in lower Cook Inlet) is one I can relate to. Speaking of Anchorage, I had the unique experience earlier this week of taking the stick of a brand new A-Star helicoptor flying through the Chugach Mountains east of Anchorage. I am not particularly qualified to fly an A-Star, but I am pretty darn good flying my kid’s little micro-helicoptor around our living room. My co-worker in the back seat did not particularly feel comfortable that my many hours flying the microhelicoptor put his life in good hands. I had a very trusty bush pilot shadowing my every move however. He routinely safely lands me on the sides of mountains during summer field expeditions, so I felt pretty confident he could quickly correct any of my mistakes.
Comment by Bryan Sralla — March 3, 2007 @ 5:47 pm
Anyone who questions the current CO2-driven climate change belief is going to be called a “tool of the coal industry” or “a tool of the oil industry.”
How do I know that? Well, I check the searches that bring people to my site and guess what one of them was?
hendrik tennekes oil connections
Henk, get read for the fun.
Comment by Bruce Hall — March 5, 2007 @ 3:25 pm
RE: #15 - I was riding in a 737 one evening. We experienced a near miss - so near, that the pilot took us through a barrel roll to evade it. I seriously doubt that a computer would have done it quite that way. Maybe there was another way that would have been equally effective - or - not. I was darn glad we had a guy who was probably an ex military jet jockey at the helm that night.
Comment by Steve Sadlov — March 5, 2007 @ 3:38 pm
Steve (#18), that reminds me of Tex Johnston’s 1955 barrel roll with the Dash-80 over Lake Washington in Seattle. Boeing’s test pilot, he felt so comfortable with the prototype-707 that he decided to pull that stunt while 200,000 people were watching the hydroplane races on the lake. He was almost fired, but insisted that a barrel roll is a one-G maneuver, with the airplane not knowing it is doing anything but flying straight and level, though with aileron deflection. Tex knew.
Comment by Henk Tennekes — March 5, 2007 @ 11:34 pm
RE: #19 - 737 is even more of a stout little animal for pulling it off. RE: “He was almost fired, but insisted that a barrel roll is a one-G maneuver, with the airplane not knowing it is doing anything but flying straight and level, though with aileron deflection.” In the case of my incident, I’d say probably 5% max, of the passengers specifically knew we were doing a barrel roll. Another 15% suspected there was something unusual. The rest were completely clueless.
Comment by Steve Sadlov — March 6, 2007 @ 11:08 am
Re Steve at 18: How many died in the collision? A “near miss” is a hit. But I love the stories.
Comment by Darwin — March 7, 2007 @ 4:11 pm
RE: #21 - I won’t argue whether that colloquialism is gramatically correct or not, but in the US, that is the term used even by the media to describe a narrowly averted collision … it is amazing / frightening to consider all the “almost disasters” one never hears about …
Comment by Steve Sadlov — March 8, 2007 @ 12:26 pm
[...] A senior climatologist suggests that those scientists engaged in building climate models are in fact merely software engineers [...]
Pingback by OpenMarket » Global warming latest — March 13, 2007 @ 11:47 am
Merely software engineers? Licensing?
If one needed a PE to sell software, we’d probably still be in the stone age. I don’t think too many people who work(ed) on DOS, Windows, UNIX, or MacOS have PEs.
Licensing SW engineers would be futile. First, just how would you test them? Arcane language syntax? Make them implement a traveling salesman algowithm?
Having a license is like having a degree: guarantees very little. I know some brilliant engineers and I know some real duds. Not much correlation between their skills and degrees.
Streets and highways are designed by PEs… serial red lights; suicide on ramps; blind curves; unmarked exit-only lanes; new intersections sited at RR crossings…
Comment by Robert — March 13, 2007 @ 7:41 pm
I don’t think they’re really software “engineers”. Modelers are more like software “technicians” and old (relatively) computer gamers.
Comment by Tim Clear — March 13, 2007 @ 11:08 pm
RE: #24 - written like a truely arrogant code writer … if I were to rank the level of quality development in most software writing enterprises on a scale of 1 - 10, it would be about a 3. Step one to understanding how to make software into a normal product, subject to proper specs, standards, testing and guard banding, is to get some structured SW development training. Additional steps to comprehend software quality best practices are avaiable. And yes, I have run SW Quality orgs in profit making enterprises.
Comment by Steve Sadlov — March 14, 2007 @ 11:08 am
Economists still remain employed when their future “projections” prove to be wrong. Ordinary software engineers would be out of business if their “projections” were always unsuccessful.
Perhaps computer modellists could be considered comparable to economists.
Comment by Vincent Gray — August 14, 2007 @ 8:54 pm
Our author writes,
“They are unlicensed, hence unqualified to sell their products to society.”
This is easily the silliest sentence I have read this month. It is absurd that humans need permission (license) to trade with one another. Such a stupid premise leads to an infinite regress. Who is to license (permit) those who license to license others? Our author does not understand liberty and the human right to trade. By implication, he worships authority.
Comment by John Howard — August 22, 2007 @ 8:37 am
[...] But the credibility of these computer model predictions took a significant hit in June 2007 when Dr. Jim Renwick, a top UN IPCC scientist, admitted that climate models do not account for half the variability in nature and thus are not reliable. (LINK)  In addition, Dr. Hendrik Tennekes, former CEO and director of research for the Netherlands Royal National Meteorological Institute, recently compared scientists who promote computer models predicting future climate doom to unlicensed “software engineers” who were “unqualified to sell their products to society.” (LINK) [...]
Pingback by Newsource21 News — September 2, 2007 @ 3:49 pm