Rough digitization of MNRR and valley

During the past week I have added new features to the coordinate input program, and performed a rough test digitization of the MNRR banks and valley boundary.

New program features include:

1. Unlimited undo and redo. The following operations automatically save a copy of the current data and settings in memory:
   
   
   1. Adding a point,
   2. Deleting points,
   3. Moving points,
   4. Joining paths,
   5. Splitting a path,
   6. Reversing a path,
   7. Moving the background and points,
   8. Zooming in or out,
   9. Loading a new background.
   
   Only the first of several contiguous additions, moves, or zooms is saved. In addition, the current state can be manually saved with a button at any time. Previous states can be visited by clicking another button, which in turn saves 'future' states for redoing, if desired. A finite number of states can be accumulated, or an unlimited number. By saving all intermediate states, one can undo back to the beginning of the session, and then redo up to the end again. Past and future states are saved with the session file, as well.
   
   
2. Path interpolation. Path points can be resampled at a given resolution using circular arcs. After interpolation, points are evenly spaced and are placed on piecewise circular arcs (although distance is still calculated linearly). This has the effect of smoothing the curve, of ensuring that every point lies on a tangent with a well-defined radius of curvature, and of maintaining a more continuous change of curvature along a path. Here is an example of before and after interpolation:
   
   
   (Click for larger image)
   
   
3. Curve attributes (name, colors, visibility, etc...) can be edited:
   
   
   
   
4. The display can be zoomed in and out (maximum magnification = 1.00). The background image is rescaled accordingly.
   
   
5. A different background image can be loaded (see below for example).

I've used this program to perform a first-pass digitization of the MNRR and surrounding valley from USGS topographic maps from 1957-1994. A mosaic using 24 maps was created yielding a 10551 x 7337 image at a scale of 8.46 meters per screen pixel at center. This was sufficient to digitize the river banks at about every 125 m, which was then interpolated to 100 m. The valley boundary was digitized at about half this resolution, and then was also interpolated to 100 m. Here is the result at a magnification of 0.125:

(Click for larger image)

Here is detail in the mid-section at 50% showing the distribution of points:

(Click for larger image)

Note that:

• The river banks were digitized from Gavins Point Dam to just below the city of Ponca, where the river begins to be bounded by walls and groynes. The resulting lengths of the right and left banks were 98.748 km and 101.541 km respectively.
  
  
• The valley 'walls' were digitized wherever obvious. Otherwise, I followed the 1200' topographic line (although not in as much detail as the river banks). I generally cut across canyons, rather than followed into them, although they were at a lower elevation.
  
  
• The valley is fairly flat throughout. The river falls from about 1170' at Gavins Point to about 1100' below Ponca. Mid-river is at about 1140'. So, a 1200' boundary is probably a good first guess. If meandering simulations exceed this boundary, it can be raised.
  
  
• I have attempted to capture a 'reasonable' rather than 'exact' facsimile of the river from these maps. For example, I have generally not included chutes, but have included islands:
  
  
  
  Ideally a later version of the meandering program can properly take these into account. I have also tried to combine conflicting data from maps made at different times (most notably the Burbank/Ponca maps from 1994 and the Elk Point map from 1963):
  
  
  
  In these cases I have not tried to favor one time period over another, but simply to render a reasonably-looking whole. At some point we may want to perform different simulations based on exact homogeneous historical data, using either maps or photographic images.
  
  
• It is interesting to superimpose the digitized banks over a mosaic made from recent Google Earth satellite images:
  
  
  (Click for larger image at 12.5%)
  
  Although the general correspondence is good (especially at each end, which is man-made), there are several notable discrepancies. I'm not sure when the satellite photos were taken, but probably within the last few years. Several features present in the maps no longer exist, and the river has clearly meandered:
  
  
  (Click for larger image at 50%)

Still to do are:

• Allow combining multiple discrete sessions into a single mosaic. I'm not sure this is required, but it would allow for digitizing the banks at a much higher resolution from multiple images rather than from a single image mosaic. Multiple landmarks per image would allow the segmented coordinates to be rectified into longer paths. For example, there are Google Earth images for most of the river at truly phenomenal resolutions. Using these images, one could easily accumulate 10,000 points or more per bank at 10 meter resolution or better. However, this would be pretty time-consuming (and the meandering simulations would be much slower), and I'm not sure we need the extra resolution right away.
  
  
• Calculate river centerline and width from banks.
  
  
• Add support to signify bank fortifications and other man-made alterations to erosion rate.
  
  
• Gather data for river depth, flow, and bed composition.