VGRID: a generic, dynamic HDF5 storage model for georeferenced grid data

Describes the Variable resolution GRID (VGRID) storage model designed to support the storage and retrieval of bathymetric data collected through the Precision Underwater Mapping (PUMA) System using the Tactical Environmental Data Server (TEDS) and the Naval Oceanographic Office's (NAVOCEANO) Digital Bathymetric Data Base-Variable (DBDB-V) Resolution product. Sponsored by the Space and Naval Warfare Systems Command (SPAWAR, PMW-155), PUMA-TEDS represents a significant advancement in the collection and assimilation of environmental data at global, regional or local levels. Although VGRID has been developed for PUMA bathymetry, its generic implementation makes it suitable for use with any type of environmental data grid through the definition of a product specification. Built on NCSA's Hierarchical Data Format version 5 (HDF5), the VGRID model inherits the HDF5 file format and library implementation that is optimized for large-scale scientific data storage. The VGRID model provides a hierarchy of environmental storage objects: files, constituents, and grids. A VGRID file can contain VGRID constituents enabling multiparameter data storage. VGRID constituents can contain VGRID grids that are identified by resolutions and have grid increments specified in arc minutes, metres, or polar stereographic grid units. The grid interface supports the storage of geographic, polar stereographic, Universal Transverse Mercator (UTM), and Universal Polar Stereographic (UPS) projected grids. Behind the scenes of the VGRID API, a tile scheme is applied to data written to the VGRID file. When VGRID grids are created, compression options can be set for all tiles created in the resolution. The VGRID tile scheme provides the framework for a robust tile caching mechanism, which minimizes the time required to read data from a VGRID file. The VGRID API uses a "bounce" algorithm to search each resolution and extract the highest resolution data for a point query. In addition, three interpolation options are available for point queries: nearest neighbor, bilinear and minimum curvature spline. The minimum curvature spline algorithm provides a "feathering" capability that effectively reduces the artifacts that often occur at the resolution boundaries of multiple resolution datasets.