Earth observation (EO) satellite files can encode the geolocation of their
observations in a variety of ways. This often depends on the processing
level of information. Gridded rasters can be described by concise map
projection information while Level 1 and 2 data that are in native satellite
coordinates require more detail. Often these files encode the pixel level
geolocation information as multi dimension variables internal to the file.
In the past there have been example implementations of storing geolocation
in an external file (early NASA MODIS) or sub sampling geolocation
information (early NASA SeaWiFS) that did not work out very well for
various reasons. Storing the geolocation data or map projection references
in each file (granule) has many advantages the most important is playing
"nicely" with tools and services and software, and promoting
interoperability. However, the geolocation data for L1/L2 EO files are
often the storage heaviest individual component as its precision requires at
least float data types (its information cannot be elegantly "packed") so it
is worthwhile to revisit ideas and methodologies for reducing its
footprint. How best could geolocation information be shared across
different variables, different files from the same sensor, or even
different sensors on the same platform. Furthermore, in the age of cloud
and database tiled storage of satellite information how is geo-location (and
other) information best packaged and utilized to improve data access and
processing. In this session we will look at this problem and potential
solution space via a number of presentations, historical lessons learned and
dynamic discussion.
Presentations:
- Introduction: Ed Armstrong and Alexsander Jelenak
- Kwo-Sen Kuo: STARE and data packaging
- Robert Wolf: MODIS Experience with External Geolocation
- Ed Armstrong; Reducing geolocation storage: a CF success story, and introducting a novel (and complicated) Eumetsat L2 data model
- Alexsandar Jelenak: HDF5 and external references
Find and access all slides: https://doi.org/10.6084/m9.figshare.8986325Session recording here.
