Poster #186, SCEC Community Models (CXM)

Geomodelgrids: A storage scheme and API for multiscale Earth models

Poster Image: 

Poster Presentation

2020 SCEC Annual Meeting, Poster #186, SCEC Contribution #10352
Geomodelgrids provides a self-describing storage scheme and API for querying multiscale Earth models with the goal of standardizing storage and access of gridded community models. The storage scheme and API are agnostic to the model data values, thereby supporting many different community models, such as seismic velocity models, stress models, and thermal models. The models are stored as a stack of grids, each with a different resolution, along with topography if it is included. The grids are stored in an HDF5 file (, which is widely used in the scientific community. The HDF5 file includes sufficient metadata to completely describe the model contents..., including values, units, and coordinate systems. Geomodelgrids builds on the IRIS Earth Model Collection (EMC) by including the same metadata (e.g., description and attribution information) while providing support for multiresolution models and scalability. Users can use various HDF5 tools (for example, h5py and the HDF5 C API) to directly access the model data, but I expect most users will prefer the high-level Geomodelgrids API. The Geomodelgrids API leverages the HDF5 file structure to implement high-performance queries of model values while only loading a portion of the model into memory and the Proj library ( for conversion between georeferenced coordinate systems. An API for parallel access is planned, which will allow querying seismic velocity models in ground-motion simulations while running on large computer clusters. The USGS San Francisco Bay Area 3D Seismic Velocity Model, including the 2008 version and a new update, will use the Geomodelgrids storage scheme and API.