DEMOGORGN

DEMOGORGN is a 100-member ensemble of geostatistical simulations of subglacial bed topography of Antarctica. 

 

These bed estimates are constrained by a variety of geophysical data sources including ice-penetrating radar measurements of ice thickness and bed elevation. These bed measurements are sparse and therefore require interpolation. Other bed mapping efforts such as BedMachine and Bedmap have approached this interpolation problem deterministically: they find a single best fit solution. While this approach maximizes accuracy, it comes at the cost of the textural realism of the topography, and it makes it difficult to propagate topographic uncertainty into ice-sheet model uncertainty.

​

With DEMOGORGN we take a different approach. Instead of optimizing accuracy in our interpolations, we prioritize geological realism, or realistic looking topography. We do this by stochastically simulating multiple topographic realizations using geostatistical simulation. Each ensemble member is equally probably. The ensemble quantifies uncertainty, and the realizations can be used to propagate uncertainty into ice-sheet investigations. 

​

DEMOGORGN conserves ice mass, which means that when an ice flow model is applied, ice mass is not created or destroyed. We enforce this constraint by applying a geostatistical Markov Chain Monte Carlo mass conservation inversion procedure. Like BedMachine, we can account for this constrain in fast-flowing regions of the ice-sheet.

​

DEMOGORGN is a living data product, which means we are constantly working to make improvements and integrate new data. Check our blog for updates, and feel free to reach out if you have feedback, suggestions, or data we should include.

 

Data repository: https://source.coop/englacial/demogorgn

GitHub and data access scripts:   https://github.com/GatorGlaciology/DEMOGORGN-Antarctica