Hello. Could we please ask three questions about data sources and fault types involved in identifying the new test faults? We’d like to have better insight into how they were identified, so we can choose appropriate input data and a sensible prediction density. Specifically:
Data sources. Which data did the NLR and USGS experts use to identify the new faults? For example: GeoDAWN lidar, the 1 m 3DEP DEM, GeoDAWN magnetics or radiometrics, imagery, field mapping, or existing geologic maps.
Fault types. Are the new faults mainly surface features, such as scarps and lineaments visible in topography? Or do they also include buried or inferred structures, interpreted from geophysics under basin fill?
Coverage. Was the whole GeoDAWN region reviewed at a similar level of detail? Or were the new faults identified in selected parts of it?
This is a great question. I’m trying to figure out what kind of interpretation the ground truth represents. A fault mapped from lidar will usually have a visible scarp, while one picked mostly from geophysics might be subtle at the surface, with something deeper to blame. I don’t want to take what’s visible at the surface for granite if the labels were actually based on several different sources.
I suppose because i cant figure out the right questions to ask it. The conductance gave me something to work with, so did area grids, theres some interestingly useful stuff buried in all download data.
We’re not sharing details about the data sources, fault types, or coverage behind the test faults beyond what’s in the problem description.
Note that the largest prize pool (Phase 2) will use a test set that is updated by expert review of all Phase 1 submissions, so your fault predictions have an impact on final evaluation even if they are not the most performant in Phase 1.