For engineers, getting the right answer is not the
problem, communicating it is.
Take Global Warming or Y2K for example of bad
communication. Using the right tools, models and equations,
assumptions and data is key. Communicating,
teaching and selling the proper problem and solution to the various clients,
regulators and political powers-that-be is possibly the most important part of
the process.
This includes the media,
lawyers, politicians and the public that think we fake elections and moon
landings.
We have all heard the medical doctor learning process
for procedures. Learn it, do it, teach
it. I contend that our learning process should
consist of learning models, doing models, teaching models and finally defending
a model in a court of law where everyone throws stones at you to create
reasonable doubt and throw your entire model out. I had such an opportunity.
I worked on a dam failure
trial for 5 years, because of Covid, and learned some good lessons. After my client’s failure I did the field
forensics and had a fair idea of what had happened but convinced myself and the
client that I had to do a lot of expensive HEHMS and RAS 2D modeling to prove
it. They agreed to an unlimited budget
(since they all think they will win, and the other side will have to pay for
it). Not only would we prove that this
flood was an Act of God or a ‘Force Majeure’ – something unavoidable - we would
also satisfy the ‘but – for’ defense that the results would have happened ‘but
– for’ (with or without) or our contribution.
I was stoked and modeled my
face off and did some of the best work of my career modeling and calibrating
the model and communicating my results to our legal team.
It was like playing with a slot machine or
Turbo tax where the numbers go up and down all day long.
I can almost hear the bells ringing.
It is my concept of Apparent Computer Veracity
where everything looks true on computers.
I can make a model say anything.
Liars model and models lie.
Sometimes
we approach models with a preconceived notion of what we want to prove and
justify with Apparent Computer Veracity. Sometimes we can back into the
preconceived solutions that we want. Sometimes we adjust reality to
fit the model. Everything must be tied
to reality. Regulators are savvy and
resistant to this backing-in process and owners usually
think it is too fancy and will cost more money. What we
must strive for is the truth, no matter how surprising or inconvenient. Garbage in equals garbage out.
I spent years recreating the storm and rain-on-snowpack
event in data free NE Nevada. Big
country, no precipitation or flow gages.
I calibrated everything to hypothetical and to real events, data,
measurements. I used contiguous and
equivalent data employing normalized regressions, considering proximity, area,
altitude, elevation angle and aspect. I
had 1000 photos with circles and arrows and a paragraph on the back of each one
explaining where it is at.
I even learned and used the latest versions of the
models since my lawyers said my 1974 HEC-I punch card version would be laughed
out of court, even if the algorithms were the same. I surveyed cross sections and highwater
marks and calibrated my 50 miles of routing through canyons and across an
alluvial plane and over several dams and bridge crossings. I balanced grid size and time intervals for
reasonable convergence and results over a million acres and ten days. I ran sensitivity
analysis and a Monte Carlo distribution of solutions to land on the one most
probable, most likely and the most defendable.
When I finished, I noticed that everything in the
system worked well but the answer always depended on one undersized culvert
under a highway that obviously backed up enough attenuated dam breach water to
cause a second, larger breach that was responsible for the damage
downstream. The answer was simple and
elegant, but the robust modeling led me there and helped me communicate the
results to our esteemed team of lawyers.
I could not help but wonder if the solution could have been more easily
formulated with better engineering and imagination. Did I model it just because I can? As a farmer rancher
once cautioned me about overthinking a project – “We ain’t building pianos
here.” Simplify.
When we presented all the technical results at a mock
trial and watched the resulting jury deliberations, we realized that we had
missed the mark.
Even as I patiently and
respectfully, but not condescendingly explained units of CFS and ACFT, they
were more concerned with my tie and haircut than they were about my
infiltration rates or Manning Coefficients.
I felt like screaming “You can’t handle the math’ but my team told me
not to get mad or try to be funny.
That
would happen naturally.
They threw 99%
of the technical models out the door and said it was a big flood and no one was
responsible.
It was an Act of God.
Force Majeure.
It would have happened anyway.
They were not wrong.
When the trial came, we threw all the technical stuff
out the window and after getting the jury to like me, (they have to like you) I
looked them in the eye and told them, “ You live here and were here for this
storm and you know that was not a 20 year storm as proposed by the plaintiff,
or a 100 year storm as required by the state, it was an Act of God.” They liked my casual Bola tie and my professorial
ponytail and agreed with me. After two
hours of deliberation, they completely exonerated the defense of any
wrongdoing. The other side had to pay
the legal fees.
The moral of the story is
consistent communication and calibration of your client and your audience as
well as your data and model is the key to elegant yet robust problem solving. Start with what answer you want to find and
who and how you want to communicate it.
Then back up and perform a minimal tool analysis, like they do for dam
work in wilderness areas. Sometimes
pack-horses are more efficient than helicopters when fixing dams in high
places. What is the least complicated
way of arriving at the desired solution.
What are the requirements, what are the limitations, what are the
desired results. Often the most elegant
solution is the simplest – Occam’s Razor.
Robust models have their place, but make sure it is necessary, fitting
and proper. Don’t use a bazooka to kill
a mosquito, even if you want to, and even if you can.