coordinate systems used in GLIMS

[Bruce Raup]

Hello GLIMS-related folks.  I submit the following for your feedback.  I am
sending it to the entire community because 1) the issue touches the whole
community, and 2) this list has too little traffic on it.  :-)
Happy New Year.

Bruce

Issue:  coordinate systems used in GLIMS

Previous idea:

  Multiple coordinate systems would be supported by (stored in) the GLIMS
  database.  Examples include:  lat/long, locally defined
  northings/eastings, UTM.

Benefit of previous idea:

  GLIMS data suppliers (Regional Centers) could work in their preferred
  coordinate systems and submit the data without conversion.

Problems with previous idea:

  - End users of the GLIMS database who query the database over potentially
    large regions should be able to expect the returned data to be in a
    uniform coordinate system, not a hodge-podge of different projections.

  - Storing many coordinate systems complicates the design of the database.




Proposal:

  The only coordinate system stored in the GLIMS database will be lon/lat
  on the WGS84 datum.

Benefits of current proposal:

  Having a single coordinate system for the entire GLIMS database would
  greatly simplify its use, as well as its implementation.

Potential problem with current proposal:

  Regional Centers must translate their output data into lon/lat/WGS84.
  (This is thought not to be a big problem.)

Notes:

  - Absolute and relative positional uncertainties will still be tracked
    separately.

  - We acknowledge that many RCs do or will work in local coordinate
    systems (e.g.  UTM/NAD83, Polar Stereographic), but we assume it
    wouldn't be a big burden to request that all new analysis results be
    submitted in lon/lat/WGS84.  Is this assumption correct?

  - For older data, and perhaps special cases with newer data, NSIDC may be
    able to do the coordinate conversion at time of data ingest.  But my
    sense is that most RCs should be able to do the translation easily with
    whatever package(s) they use to generate the results.

  - Commercial packages like ESRI's should be able to handle coordinate
    transformations easily.  I also know of a few free software packages
    that will do coordinate transformation:

      GMT    (Generic Mapping Tools (http://gmt.soest.hawaii.edu/))
      Proj.4 (Cartographic Projections Library; does datum shifts
              as well.  (http://www.remotesensing.org/proj/))
      GRASS  (actually uses Proj.4 internally; http://www.baylor.edu/grass/)

  - Distances and areas are generally calculated in projected coordinates.
    Obviously, this can still be done at Regional Centers.  There is,
    however, the question of how individual points (polygon vertices) are
    assumed to be connected:  along great circles, or as straight lines in
    some projection.  A quick test has shown (see below) that this
    difference will not be a problem, assuming 1) suitable projections are
    used by both data supplier and end user, and 2) points that are
    connected aren't more than 100 km or so apart.

In the below, "Error" is the distance between the midpoint of the segment
as calculated on the great circle connection versus that calculated on the
straight-line connection in the projection.  Lengths (last two columns) are
great circle lengths.

Projection:  South polar stereographic, true scale at -71 degrees)

Input coordinates                                             Length  Error
(lon,lat,lon,lat)       Note                                  (km)    (km)
---------------------------------------------------------------------------
-60  -75 -50     -77    Ronne Ice Shelf front                 349.700 0.290
-165 -78 -180    -78    Ross Ice Shelf front                  347.335 0.249
-165 -78 -166    -78    Ross Ice Shelf front                   23.219 0.001
-165 -78 -169.33 -78    100 km segment, east-west             100.516 0.021
-165 -78 -165    -78.9  100 km segment, north-south           100.484 0.019


Projection:  (UTM zone 13, over Colorado, USA)

Input coordinates                                             Length  Error
(lon,lat,lon,lat)                   Note                      (km)    (km)
---------------------------------------------------------------------------
-105.6505 40.0233 -105.6490 40.0251 Arapaho Glacier             0.237 0.000
-105      40      -105      40.9    100 km segment north-south 99.939 0.002
-105      40      -106.175  40      100 km segment east-west  100.337 0.002
-108      40      -108      40.9    same N-S, at edge of zone  99.939 0.008
-108      40      -109.175  40      same E-W, at edge of zone 100.337 0.010


-- 
Bruce Raup
National Snow and Ice Data Center                     Phone:  303-492-8814
University of Colorado, 449 UCB                       Fax:    303-492-2468
Boulder, CO  80309-0449                                    braup at nsidc.org