"left" and "right" sides of glacier boundaries

[Bruce Raup]

All,

First, thanks for a good discussion, and for sending your ideas directly to
the list.

I sense that we're all understanding the issue in the same way now.  In
summary, we're talking about the definition of "left" and "right" when
specifying material or feature attributes for a segment that is part of a
polygonal glacier boundary or internal rock boundary.  The conventions that
seem appropriate here are the ones cited in previous messages related to
Oracle's, GML's, ECS's, and FGDC's handling of geospatial polygons.

Given all the discussion, I'd now like to (re-)propose we go with the
following convention for glacier polygons in GLIMS.  After that I'll point
out some of the implications.

Proposed convention:

- Glacier outline polygons are to be closed (this is old news)

- Internal rock boundary polygons are to be closed (still old news)

- Glacier and rock polygons can be composed of different segments that
  carry different attributes, specifying what materials or features are on
  the "left" and "right" sides. (The new part comes next.)

- Glacier polygons will be stored in the database such that their vertices
  will be in counterclockwise (or anticlockwise, if you prefer) order, as
  you look down from above.

- Internal rock polygons will be stored in the database such that their
  vertices will be in clockwise order, as you look down from above.

See attached figure for an illustration.  It's also at
http://cires.colorado.edu/~braup/GLIMS_temp/polygon_direction.png

Implications:

- As Graham Cogley pointed out, area calculation for the ice only (glacier
  minus internal rocks) becomes a matter of applying the same formula to
  all "glac_bound" and "intrnl_rock" polygons for a given glacier and
  summing, since anticlockwise polygons have positive area while clockwise
  ones have negative area.

- Given a rock-ice boundary, ice will always be on the left, and rock will
  always be on the right.  (Anyone got a counter-example?  Debris cover
  doesn't count as "rock".)

The RCs and Stewards don't need to worry about putting the segments or
vertices in the proper order, since my ingest software does this already.
Specifically, it strings segments of the same type (e.g. "glac_bound" or
"intrnl_rock") with the same glacier ID together in the right order,
reversing the direction of segments as needed (which might have been
digitized in the other direction compared to the rest of the polygon), and
detects gaps or open polygons.  If everyone agrees to this convention, the
software will soon make sure glac_bound polygons are counterclockwise and
intrnl_rock polygons are clockwise as well.

What RCs and Stewards *do* need to pay attention to is the definition of
"left" and "right".  It doesn't matter what direction a segment is
digitized in, but the attributes of what materials are on the "left" and
"right" must be assigned according to this definition.  This might affect
work already done, so please let me know if so.

Assuming the second implication above is always correct, I can make the
ingest software swap "ice" and "rock" for ice-rock boundaries if they don't
make sense.

Please post to the list what you think.

Bruce

-- 
Bruce Raup                                               Phone:  303-492-8814
National Snow and Ice Data Center, U. of Colorado, 449 UCB, Boulder, CO 80309
http://cires.colorado.edu/~braup/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: polygon_direction.png
Type: image/png
Size: 73750 bytes
Desc: 
URL: <https://nsidc.org/pipermail/glims/attachments/20050303/e5b1b389/attachment.png>