And they ended up having to spray the south leg with fire hoses to cool it down, because it had expanded too much and they couldn’t fit the last section.
I reverse engineered all the curves on the Atlantic Road in one of my road-design classes!.
What does that mean, to reverse engineer the curves?
Interestingly enough, I have read that they had to do about the same thing when the steel arches of the Eads Bridge over the Mississippi River at St. Louis (finished in 1874) were being closed. Although they used ice, instead of a fire hose.
Basic answer:
For me it means “designing” a road to match some kind of existing conditions. So if you have survey data for an existing roadway, you would be figuring out the horizontal alignment, vertical alignment, and cross slopes that would best approximate it.
TLDR:
Horizontal alignment would consist of the plan view curves and the tangents that connect them. There may be spiral transitions between the tangents and the curves - common for rail design but less so for roadways in my experience.
Vertical alignment, or profile, is made up of tangents (constant grade/slope) and the vertical curves that connect them. Vertical curves are almost always parabolic. For small changes in grade (such as one half to one percent depending on design speed) there sometimes won’t be a vertical curve.
There are two common ways I’m familiar with to convey how cross slope varies along the length of the roadway. One way is to have one or more additional profiles. For example on a residential street there might be profiles associated with the centerline of the roadway, the left top of curb, and the right top of curb.
The other way is to describe where along the roadway the cross slope is constant and where it is in transition. So for a two lane highway, you would indicate that along a straight section of roadway from point A to point B, the pavement slopes down away from the centerline at some constant figure, commonly 2%.
The following section, from point B to C, approaching a left curve for example you would indicate that the cross slope along that length is 2% down at point B and transitions to 4% up (or whatever) at point C. Through the curve, from point C to D, the cross slope would again be constant, this time staying at 4% up.
The other side of the roadway (for vehicles traveling in the other direction and going around a curve to the right) would transition from 2% down to 4% down but not necessarily begin transitioning at the same points along the roadway.
The “banking” of these curves is called superelevation.
How you actually develop these roadway parameters, store them, manipulate them, etc. varies depending on the software being used of course.
Most of the time I “reverse engineer” a roadway, it is just a short section and it would be used for figuring out how a proposed roadway would tie into an existing roadway. Unless there’s a compelling reason to do otherwise, this will often be along a tangent section as it’s somewhat easier to design and build that way. Modern design software, survey technology, and construction methods being what they are, it is actually not all that much more difficult to tie in along a curve.
I sometimes reverse engineer long stretches of roadway as well. A good example is a roadway being re-constructed in place, possibly as part of a widening project. In this case you want to design or “reverse engineer” a description of the roadway along the entire length of interest so that the new pavement will match the old roadway as close as possible.
Even if you have a copy of the design plans from the existing roadway, they will probably only be of limited help as a reference. They are often many decades old and of dubious accuracy/quality plus the road that got built may have been slightly different than the design. “As-built” plans are sometimes available, but the reverse engineered alignment is still likely to be slightly different. They should be close but they won’t be identical.
They also cooled down final segments of the Gateway Arch in St. Louis, which is very close to the Eads Bridge, for the same reasons.
Thanks for the full explanation, and this particular part sounds like what was done for the replacement eastern span of the San Francisco-Oakland Bay Bridge that opened in September 2013.
Wow! Thanks for the links! Did you watch the video in the second one? Around 4 minutes in it starts getting mighty wet. That’s an amazing road, and riding it would be fantastic, and somewhat crazy 
Thanks for the glimmer of insight… Civil engineering fascinates me, especially when Crew A plops down a set of abutments or piers, and months or even years later, Crew B starts unrolling a new road across the countryside and everything meets up smoothly without any last-moment bumps or joggles.