MNRR Project -- 03/21/08

More Meandering Tests

To make a long story short, I spent this week fixing the problems encountered last week (plus some others). Although, rather than say 'fixing', I should say 'patching' or more accurately 'finding workarounds for'. The practical end result is the same: progress. Rather than belabor the issue(s), I'll just present some working results, and concentrate on the success of the effort instead of its gory details.

To test the code, I've manually set the channel width to 10.0 and the depth to 2.5 for all meandering nodes. The actual meandering calculations are now performed by an external Python program which is called (via popen()) from within meander_() in meander.cpp. All input and output arrays are passed to/from this program as text, which probably isn't the fastest way to do it, but makes the testing much easier to debug at present.

Here is an excerpt from a single meandering 'transaction' for the null case (delta x and y are always set to zero):

time = 0, stations = 96, stnserod = 96

    i              x              y             xs           dels          flow          rerody        lerody           slope          width          depth           diam
----- -------------- -------------- -------------- -------------- -------------- -------------- -------------- -------------- -------------- -------------- -------------- 
    0     484.974226     960.000000       0.000000      10.000001       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    1     484.970962     950.000000      10.000001      10.000001       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    2     484.974293     940.000000      20.000001      10.000000       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    3     484.976040     930.000000      30.000001      10.000000       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    4     484.974226     920.000000      40.000001      10.000001       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    5     484.970859     910.000000      50.000002      10.000001       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    6     484.975120     900.000000      60.000003      10.000000       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    7     484.976321     890.000000      70.000003      10.000000       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    8     484.974226     880.000000      80.000003      10.000000       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
    9     484.974239     870.000000      90.000003      10.000000       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
...
   86     484.970384     100.000000     860.000045      10.000000       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
   87     484.967932      90.000000     870.000045      10.000002       0.031688       0.010000       0.010000       0.010000      10.000000       2.500000       0.010000
   88     484.974226      80.000000     880.000047      10.000000       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000
   89     484.974203      70.000000     890.000047      10.000000       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000
   90     484.976472      60.000000     900.000047      10.000000       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000
   91     484.974729      50.000000     910.000047      10.000000       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000
   92     484.974226      40.000000     920.000047      10.000001       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000
   93     484.970844      30.000000     930.000048      10.000001       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000
   94     484.966139      20.000000     940.000049      10.000000       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000
   95     484.967670      10.000000     950.000049      10.000002       0.031688       0.010000       0.010000       0.009999      10.000000       2.500000       0.010000

    i           delx           dely     rightdepth      leftdepth         lambda
----- -------------- -------------- -------------- -------------- -------------- 
    0       0.000000       0.000000       2.500000       2.500000       0.500000
    1       0.000000       0.000000       2.500000       2.500000       0.500000
    2       0.000000       0.000000       2.500000       2.500000       0.500000
    3       0.000000       0.000000       2.500000       2.500000       0.500000
    4       0.000000       0.000000       2.500000       2.500000       0.500000
    5       0.000000       0.000000       2.500000       2.500000       0.500000
    6       0.000000       0.000000       2.500000       2.500000       0.500000
    7       0.000000       0.000000       2.500000       2.500000       0.500000
    8       0.000000       0.000000       2.500000       2.500000       0.500000
    9       0.000000       0.000000       2.500000       2.500000       0.500000
..
   86       0.000000       0.000000       2.500000       2.500000       0.500000
   87       0.000000       0.000000       2.500000       2.500000       0.500000
   88       0.000000       0.000000       2.500000       2.500000       0.500000
   89       0.000000       0.000000       2.500000       2.500000       0.500000
   90       0.000000       0.000000       2.500000       2.500000       0.500000
   91       0.000000       0.000000       2.500000       2.500000       0.500000
   92       0.000000       0.000000       2.500000       2.500000       0.500000
   93       0.000000       0.000000       2.500000       2.500000       0.500000
   94       0.000000       0.000000       2.500000       2.500000       0.500000
   95       0.000000       0.000000       2.500000       2.500000       0.500000

Here are graphics for running 1000 iterations of this case:

(Click for larger image)

This is (finally!) as expected: no change. Here is detail of the final output:

(Click for larger image)

The second set of nodes on the left bank was added by AddChanBorder() at 1/2 the width plus 1cm.

The meandering output can be set to something more interesting than zero; in this case delta x is set to the sine of the y coordinate:

    i           delx           dely     rightdepth      leftdepth         lambda
----- -------------- -------------- -------------- -------------- -------------- 
    0       0.000000       0.000000       2.500000       2.500000       0.500000
    1       0.013218       0.000000       2.500000       2.500000       0.500000
    2       0.026378       0.000000       2.500000       2.500000       0.500000
    3       0.039423       0.000000       2.500000       2.500000       0.500000
    4       0.052296       0.000000       2.500000       2.500000       0.500000
    5       0.064940       0.000000       2.500000       2.500000       0.500000
    6       0.077300       0.000000       2.500000       2.500000       0.500000
    7       0.089322       0.000000       2.500000       2.500000       0.500000
    8       0.100953       0.000000       2.500000       2.500000       0.500000
    9       0.112143       0.000000       2.500000       2.500000       0.500000
   10       0.122843       0.000000       2.500000       2.500000       0.500000
   11       0.133005       0.000000       2.500000       2.500000       0.500000
   12       0.142586       0.000000       2.500000       2.500000       0.500000
   13       0.151543       0.000000       2.500000       2.500000       0.500000
   14       0.159838       0.000000       2.500000       2.500000       0.500000
...
   81      -0.159838       0.000000       2.500000       2.500000       0.500000
   82      -0.151543       0.000000       2.500000       2.500000       0.500000
   83      -0.142586       0.000000       2.500000       2.500000       0.500000
   84      -0.133005       0.000000       2.500000       2.500000       0.500000
   85      -0.122843       0.000000       2.500000       2.500000       0.500000
   86      -0.112143       0.000000       2.500000       2.500000       0.500000
   87      -0.100953       0.000000       2.500000       2.500000       0.500000
   88      -0.089322       0.000000       2.500000       2.500000       0.500000
   89      -0.077300       0.000000       2.500000       2.500000       0.500000
   90      -0.064940       0.000000       2.500000       2.500000       0.500000
   91      -0.052296       0.000000       2.500000       2.500000       0.500000
   92      -0.039423       0.000000       2.500000       2.500000       0.500000
   93      -0.026378       0.000000       2.500000       2.500000       0.500000
   94      -0.013218       0.000000       2.500000       2.500000       0.500000
   95       0.000000       0.000000       2.500000       2.500000       0.500000

Here are graphics for 1000 iterations:

(Click for larger image)

Here is detail of the final output:

(Click for larger image)

One abberation is that AddChanBorder() seems to duplicate nodes along with their drainage areas, although I don't think these are sent to the next meandering cycle:

(Click for larger image)

More detail of the final output showing the potential floodplain:

(Click for larger image)

Here are results for:

w = (960.0 - y) / 950.0
delx = sin(w * 2.0 * pi) / 5.0 + sin(w * 6.0 * pi) / 10.0 + sin(w * 10.0 * pi) / 15.0 + sin(w * 14.0 * pi) / 20.0

(Click for larger image)

More detail of the final output:

(Click for larger image)

This is pretty much where I wanted to be last week (sigh). However, I still have 10 days remaining in March to begin testing the JP code. I expect to discover additional problems, but am fairly confident that I can find workarounds as required.

Some advantages of using the hybrid external testbed include:

Very fast turnaround in testing code edits.
Ability to test the meandering code independently of Child.
Reasonably fast running of the hybrid system (<< 5 minutes for 1000 iterations). I expect much better performance if I switch to a binary interface and keep an open connection to the external program.
In addition to calculating just the meandering outputs, the external program can be passed additional simulation data and can return other important values (e.g. channel width, critical flow, criteria for channel interpolation and node additions or removals, etc..) in a dynamic manner. If necessary, it could also handle some of the dynamic remeshing operations.
In short, a 'meandering protocol' considerably more advanced and useful than the current 'Fortran-like' interface can be evolved so that, in a real sense, the Child code becomes scriptable from outside its C++ bounds. This may be essential for the MNRR project, in which it is envisioned that many many sequential simulations with slightly different static and dynamic parameters will be required.