The GLIMS Glacier Database was originally designed to contain rich data on not only glacier outlines, but the physiographic context of different parts of each outline. Outlines were to be composed of segments that when joined together would form a complete outline, and these segments could have associated attributes unique to them, such as what materials (rock, vegetation, snow, ice, …) were on the left and right sides. Similarly, nunataks (rock outcrops internal to a glacier) were represented with separate polygons so that they could carry attributes separately from the glacier outline.

As the GLIMS community and database grew, actual patterns of data production and use diverged from this initial design. Researchers provided glacier outlines not as collections of segments, but as complete outlines, with “holes” (inner rings) representing nunataks. The Randolph Glacier Inventory (RGI) used this same simpler data model. These differences were accommodated by on-the-fly data transformations during both data ingest (holes-to-polygons for nunataks) and data download (nunatak polygons-to-holes).

To align the data model with actual use patterns, we completed a project to simplify the database schema and associated software (input and output).

The database contents are almost completely the same in the new data model. Representation of image metadata, people and their affiliations, literature references, and so on, remain unchanged. The following changes resulted from the new simpler data model:

1. Nunataks (rock outcrops internal to a glacier) are represented by holes in the glacier boundary polygon rather than separate polygons. The old download interface offered a choice between these two representations, but most people downloaded data in the “holes” format. A query sent to the GLIMS mailing list in late 2024 about the possible loss of the "separate polygon" option elicited no complaints.
2. The vast majority of the small topological errors in some glacier outlines have been fixed. These tend to be small self-intersections that can be fixed with minimal modification to the outline.
3. Multi-polygons, as well as multiple outlines that share a glacier ID, have been converted to single outlines, each with its own identity. Thus, the glacier IDs have changed for some entities.
4. Many tables have been eliminated, such as those that stored individual segments of an outline.
5. A few fields have been removed, such as the "local" and "global" uncertainty fields. This is because those values were fairly arbitrary in practice, and one can know these uncertainties from knowledge of the imagery used.
6. The download interface has become more streamlined, since the data model (number 1 above) is now "holes" only. Also, the experimental "RGI-on-demand" feature has been removed for now.