Make the Synthetic Storm Surge Example "Interesting"

[mandli]

Currently the synthetic storm surge example works but does not have a lot going on. It would be nice to make this example a bit more interesting by doing things such as

1. Adding gauges
2. Adding examples as to how to create a synthetic track and intensity for the storm
3. Create synthetic bathymetry
   These are just suggestions of course.

As this example has been removed this issue now either needs to add this example back or should move what was here back into the apps repository.

[rjleveque]

@mandli, I was remaking the galleries for v5.8.0 and discovered that this example now dies with "too many dt reductions" at t = 5638.52. I see you recently changed it from order 1 to 2, and even with the new rpt2_geoclaw.f routine it dies for me (but still runs with order 1). Did you get this to run with order 2?

[mandli]

Odd, I thought it was working before with order 2 when I was running it before but I agree, it seems to not be working although I don't see why right now. I will work on this a bit but if I don't see the issue right away we should roll back the order to 1 I suppose.

[mandli]

Strangely when trying to debug this I increased the number of output and everything ran fine. More puzzling, when I reduced the number of output it also ran fine.

[rjleveque]

Yes, increasing the number of output times seems to make it run, not sure why. Also it takes hours to run, unlike most most things in the examples directories, and doesn't seem to illustrate anything in particular so I'm not sure what the goal is with this example. For now I'm in favor of a quick fix so we can release 5.8.0, if possible.

[mandli]

Hours to run? For me it takes a couple of minutes, which admittedly is still longer than the other examples. More importantly though is the stability of the problem either with more or less output steps.

As to what it was supposed to demonstrate, I was not aware that the example had the gauges removed. It was supposed to show how to setup an example synthetic test case, which is something I constantly get asked about. Right now this is pretty basic, which is why this issue was opened. In principle this could be more a demo with a Jupyter notebook added with some more explanation.

[rjleveque]

Am I looking at the right example? In geoclaw/examples/storm-surge/synthetic the oldest commit is a6d4f1e when it was renamed from square-basin in 2019, and there weren't any gauges in that version either. It's a 3-level run from -12 hours to +3 days.

[mandli]

That's the right one. This is the timing I have with 3 outputs:

============================== Timing Data ==============================

Integration Time (stepgrid + BC + overhead)
Level           Wall Time (seconds)    CPU Time (seconds)   Total Cell Updates
  1                     8.402                 32.233            0.291E+08
  2                     9.964                 11.532            0.853E+07
  3                    93.587                632.387            0.526E+09
total                 111.952                676.152            0.564E+09

All levels:
stepgrid              105.570                651.894
BC/ghost cells          6.020                 23.690
Regridding              2.121                  6.543
Output (valout)         0.005                  0.005

Total time:           115.536                685.049
Using  8 thread(s)

Note: The CPU times are summed over all threads.
      Total time includes more than the subroutines listed above
Note: timings are also recorded for each output step
      in the file timing.csv.

=========================================================================

[rjleveque]

We decided to remove this example for now since it's giving strange behavior. When I run it with order=2 and more output times, something funny happens at time = 0 days that causes it to refine the entire domain to the finest level, leading to the slow run time. Perhaps something about the synthetic storm being used?

It also needs some gauges and other enhancements to better illustrate use of the storm surge code, and then perhaps put in the apps repository?

Let's leave this issue open as a reminder to follow up on it.