Different kinds of observations feature different strengths, e.g. visible-infrared imagery for clouds and radar for precipitation, and when integrated better constrain scientific models and hypotheses. Even critical, fundamental operations such as cross-calibrations of related sensors operating on different platforms or orbits, e.g. spacecraft and aircraft, are integrative analyses. The great variety of Earth Science data types and the spatiotemporal irregularity of important low-level (ungridded) data has so far made their integration a customized, tedious process which scales in neither variety nor volume. Generic, higher-level (gridded) data products are easier to use, at the cost of being farther from the original observations and having to settle with grids, interpolation assumptions, and uncertainties that limit their applicability. The root cause of the difficulty in scalably bringing together diverse data is the current rectilinear geo-partitioning of Earth Science data into conventional arrays indexed using consecutive integer indices and then packaged into files. Such indices suffice for archival, search, and retrieval, but lack a common geospatial semantics, which is mitigated by adding on floating-point encoded longitude-latitude (lon-lat) information for registration. An alternative to floating-point lon-lat, the SpatioTemporal Adaptive Resolution Encoding (STARE) provides an encoding for geo-spatiotemporal location and neighborhood that transcends the use of files and native array indexing allowing diverse data to be organized on scalable, distributed computing and storage platforms.
Keywords: Big Data; DGGS; STARE; data fusion; geolocation; integration.