Initial full-river tests

During the past week I have run some initial 'unconstrained' simulations of the entire 59-mile MNRR segment, using the circumferential speed method (which is the unmodified 'sigma' in JP), Johannesson-Parker implementation 3, and two prototype variable-width methods. The second variable-width method, based on sediment uptake and loss from water flowing above or below a critical speed, did not perform well in the short time I devoted to it, and is not shown here. Nor did I try a braided-river meandering method, also due to time constraints. I would like to save the variable-width and braided meandering methods for next month, if time permits, or for the followup project, if it occurs. Although these are both interesting and worthwhile lines of pursuit, they are not strictly germane to the current project. For the remainder of this project, I suggest that we restrict ourselves to fixed-width meandering methods, using either an average-width river as input (in the case of JP), or the actual-width river, in the case of circumferential speed. The examples shown below should make this recommendation clearer.

Here is the initial average-width (747.84 m) river:

(Click for larger image)

Here is the initial actual-width (186.79 to 1647.97 m) river:

(Click for larger image)

Here are the end results of 100 yr simulations (dt = 0.1) using several methods and input parameters. Migration rates were adjusted so that the initial erosion was on the order of 1 ha/yr/km (as quoted in EJ2006). All plots are shown at the same scale:

method = circumferential speed (sigma)
width = 747.84
flow = 899.33
depth = 3.429
ldist = 0.5 widths
sdist = 1.0 width
mrate = 200.0
mexp = 1.0

(Click for larger image)

method = circumferential speed (sigma)
width = actual
flow = 899.33
depth = 3.429
ldist = 0.5 widths
sdist = 1.0 width
mrate = 200.0
mexp = 1.0

(Click for larger image)

method = JP3
width = 747.84
flow = 899.33
depth = 3.429
slope = 1.918e-04
diam = 0.001
mrate = 2.0
mexp = 1.0

(Click for larger image)

method = JP3
width = 400.0
flow = 899.33
depth = 3.429
slope = 1.918e-04
diam = 0.001
mrate = 20.0
mexp = 1.0

(Click for larger image)

method = JP3
width = 200.0
flow = 899.33
depth = 3.429
slope = 1.918e-04
diam = 0.001
mrate = 40.0
mexp = 1.0

(Click for larger image)

method = variable width speed (sigma)
width = 747.84
flow = 899.33
depth = 3.429
ldist = 0.5
sdist = 1.0
mrate = 140.0
mexp = 1.5

(Click for larger image)

method = variable width speed (sigma)
width = actual
flow = 899.33
depth = 3.429
ldist = 0.5
sdist = 1.0
mrate = 100.0
mexp = 1.5

(Click for larger image)

As you can see, there is quite a bit of variation in the results, although it is variation 'on a theme'. Even in 100 years, the simulation does not become 'unbounded'. The planform is clearly finite, and the river stays fairly close to its original position. And yet none of these simulations captured the essence of what I had hoped to achieve: the dynamic 'wildness' of the unmanaged river of over 100 years ago. I was hoping, in the unconstrained simulations, to get something that looked like the 1895 maps:

(Click for larger image)

including the long, open loops near Vermillion and Sioux City. So far, the 'wildest' simulation I have done is the following:

method = circumferential speed (sigma)
width = actual
flow = 899.33
depth = 3.429
ldist = 1.0 widths
sdist = 2.0 widths
mrate = 1000.0
mexp = 1.5

(Click for larger image)

My schedule for the next several months is:

July

• Convert the left and right valley boundary lon/lat coordinates (and 1200' lines) to {x, y}.
• Run 'constrained' simulations using valley boundaries and fixed river endpoints.
• Continue to look for parameter combinations that produce representative MNRR behavior.
• Accumulate quantitative statistics for erosion rate, total area eroded, maximum area of planform, and occupation frequency for {x, y} points in the valley, in preparation for accumulating similar probability statistics (e.g. surfaces) for multiple simulations. i.e.: Begin to define concrete quantities which delineate the '100 yr migration corridor'.
• Overlay simulation runs with lon/lat maps, satellite images, of area.
• If time permits, try a variable-width braided simulation (I think one of these needs to be included in the final report for this project, as it will be used in the next (?) project).

August

• Create framework for performing multiple (e.g. 1000) river simulations and automatically accumulating planform/migration statistics.
• Pick 1-2 meandering methods and parameter centroids for production runs.
• Choose variation in parameters (e.g. 10 x 10 x 10 combinations) to yield ~1000 slightly different simulations for each method.
• Test multi-simulation framework with methods and parameter sets.

September-October

• Perform production runs: ~1000 simulations.
• Accumulate planform/migration probability statistics and display as surfaces overlaying lat/lon maps and satellite images.

November-December

• Write PDF report.
• Create website of results and software.
• Finish everything else.