GLIMS Glacier Database: Geographic Control

NOTE: This document is obsolete. The GLIMS Glacier Database is designed to contain spatial entities only in geodetic coordinates (lat/lon/WGS84). We hope to have a replacement document soon that discusses how measurement uncertainties are tracked.

The locations of all geographic entities in the GLIMS database, such as glacier boundary polygons, are stored as northing/easting coordinates (offsets) from a reference point which is identifiable in imagery. Reference points will typically be mountain peaks, sharp bends in rivers or lake boundaries, or other stable and easily identifiable features. These coordinates are related to the line/sample and latitude/longitude coordinate systems via standard transformations. In scenes that do not contain any appropriate immobile features, such as within ice sheets away from all rock, an artificial reference point point will be used.

Figure 1: Relation between difference coordinate systems.
G: glacier; A: ablation zone; I: immobile region.

There will be three basic coordinate system types used in this processing and in the database:

1. Local image coordinate system (abbreviated L/S). Units are line/sample, and are referenced to a corner of the image.
2. Local ground coordinate system (abbreviated N/E). Units are meters north and meters east of some reference point on the ground that is probably, but not necessarily, contained in the image. Typically, there will be one reference point per glacier, although a single reference point can be shared by many glaciers.
3. Global coordinate system (abbreviated L/L). Units are geographic longitude and latitude (degrees).

The Figure 1 above depicts the three interrelated coordinate systems.

Figure 2: Coordinate Transformations

Polygons, vectors, and the like will be stored in the database expressed in the local ground coordinate system (N/E). Also stored will be the transformations to go between L/S and N/E (Transform 1 in Figure 2), and between N/E and L/L (Transform 2 in Figure 2). We will not know the precise geolocation of most images, thus all location information should be stored in a relative sense. As new information is obtained relating the N/E reference point to L/L, only that latitude/longitude coordinate need be updated, rather than having to update every location associated with that reference point, as would be the case if we stored everything in L/L to start with. In this scheme, L/L coordinates are computed every time they are desired, making it easy to update geographic coordinates for image-derived data. This flexibility represents little computational burden for modern computers, for which the time to read L/L coordinates from a disk is longer than required to do a N/E to L/L calculation. Note that the reference point for the N/E coordinate system is simply the origin for that coordinate system. It is not to be confused with tie-points used to co-register two images, or ground control points (GCP's) used to refine absolute locations.

The alignment between the image coordinate system and the true L/L grid is determined by the angle the orbit track makes with the meridian lines (a known function) and the yaw of the spacecraft, typically known to better than 10 arc-seconds (48 microradians). This amounts to 3 m over the 60 km width of an ASTER image, and should thus not be a problem . The angular relation between the N/E system and the L/L grid varies slightly over a spacecraft scene and depends upon the cartographic projection used. Any well-defined cartographic projection may be used; GLIMS will probably use Transverse Mercator with a local central meridian for individual scenes, and UTM for mosaics.

Each glacier will be assigned an ID automatically the first time it is entered into the GLIMS system. This ID will be a combination of a representative longitude and latitude for the glacier, rounded to the nearest 0.001 degree, which corresponds to 111 m at the equator. Because the uncertainty of dead-reckoned positions (based upon engineering information alone) in current spacecraft imagery systems is typically about 200 m (e.g., RADARSAT; expected to be 185 m for ASTER), even where no ground control, ID assignments should rarely, if ever, need to be modified.

Transform 1 can consist of a rotation, translation, magnification, the location of the reference point (L/S) in the image, topographic correction, and any geometric distortion internal to an imaging system. It does not depend upon absolute geodetic locations. Transform 2 simply consists of the longitude and latitude, together with their uncertainties, of the ground reference point, together with standard mapping transforms for converting northings and eastings to L/L. Thus, it does not depend upon the imaging system, only upon the geodetic control.

A given image will have associated with it as many ``Transform 1''s as there are reference points associated with it. Pointers to ``Transform 2''s will be stored with the glacier information (pointers because several glaciers might share the same reference point, and the information about the reference points should be stored only once).

Since glacier IDs will contain location information in the L/L coordinate system, searching for glaciers based on latitude and longitude will be straight forward.