Page History

Vision for Legacy HDF (.he4), SMAP and ICESat-2 migration to the Cloud

Possible Feature Roadmap - Highly subjective, Order/Content subject to change.
(Essentially this is a "gap" analysis - what needs to be provided in cloud services to cover what is available today in on-prem services).

• General enable/configure/test Main service per Datasets
  • Trajectory Subsetter - Available as a cloud data service and mostly pre-configured for SMAP & ICESat-2.  Should only require metadata connections.
  • HOSS - Minor updates for SMAP, ICESat-2 special cases (e.g., TRT-290 - HOSS support for SMAP L3)
    • Missing CF for SMAP (configuration files already accommodate, mostly)
    • Note e.g., HOSS (OPeNDAP) produces NetCDF-4 output, even for HDF-5 inputs, but .nc4 is built upon .h5.  Likely we will end up with .nc4 as default output format, and as a preferred format over "plain" .h5 in all cases. "Plain" .h5 (native) for processed outputs is not available.
    • Lower-Left Orientation - is a known issue, not currently supported in HOSS.
    • Edge aligned dimension scales - also a known issue.
  • SMAP-to-GeoTIFF - Special case in Trajectory Subsetter for SMAP data,
    • Could be - just needs enabling existing code for In-Cloud deployment
    • However, it is likely we will recommend a revised implementation - SMAP to Gridded HDF-5 format + Common HDF-5 to GeoTIFF/other (see following) - SMAP_L2_Grid.
• Harmony-GDAL-Formatter - GeoTIFF/COG outputs - ICESat-2 & SMAP (L3/L4, .h5 Format) inputs
  
  • Refactor HGA - split archive, 2 or 3 separate services - HGF, HGA, HGAgg, (TRT-213 - Images for GITC?)
  • COG formatter - GeoTIFF source (1st?), NetCDF source (2nd?)
  • See HGA Roadmap
• L1/L2 Formats (mostly custom - port on-prem code to in-cloud):
  • Shapefile (Vector data, ESRI-based file format, Custom code exists, GDAL?)
  • NetCDF (Custom upgrade .h5 to .nc4 with CF support, L1/L2, port to in-cloud)
  • ASCII (CSV, Tabular, other? - Custom code exists, GDAL?)
  • kmz, kml (GDAL + custom front-end) - requirements?
• (Possible Trajectory Subsetter rewrite - Python - as a maintenance issue)
• L3/L4 Reprojection - PyResample and/or GDAL based?
• HGA-Reprojection? (SMAP focus, currently GDAL based, but not clear if GDAL specific option is best path)
  • See HGA Roadmap
• Projection Gridding for Trajectory Data (based on Swath-Projector, e.g. for ICESat-2) ?
  • Minor fix for 1D vs 2D inputs
  • Not specifically an on-prem feature, but from use-case analysis - an opportunity
• Legacy HDF-4 file handling (HDF-4 to NetCDF/CF converter as pre-step, OPeNDAP?)
  
  

Feature Summary:

• L1/L2 - Trajectory Subsetter - SMAP special case - Gridded outputs
  • Enable in existing code base for Cloud deployment
  • Gridded outputs to .h5 (Cropped grid, Revive un-merged branch?  Not-GeoTIFF as on-prem)
  • GeoTIFF formatter (COG, HGA) - chaining.  (Alternative is enabling existing on-prem Gridding to GeoTIFF functionality).
  • Reprojection special case? (HGA?, chaining)
    
    
• L1/L2 - Trajectory Subsetter + general chaining for formats (ICESat-2, SMAP, other?)
  • Projection - use Swath-Projector (minor updates likely, 1D inputs)?
    
    • Potential use cases for Oblique Equal-Area, Oblique Mercator, or Oblique EquiDistance projections
    • Note SMAP is spiraling trajectory and is already "gridded" with CEA row/col data
    • Not an existing on-prem feature
  • Reformat: Vector Data: Shapefile, KML, NetCDF, multi-file ASCII
    - as micro services, lifted mostly from on-prem implementations (or port/rewrite)
    • (Note size constraints, e.g., ICESat-2 files)
    • NetCDF: CF, metadata upgrade from .h5 (on-prem implementation)
    • ShapeFile, KML format output (.shp, .kml, .kmz) 
    • ASCII format output? - rewrite to Python, chaining - single file per variable
      • CSV
        • Choice of delimiters? (Comma, tab)
        • “Row of vars” per time-stamp? (see Tabular ASCII)
        • “Time, Name, Value” format
          or “Time, Value” per variable, separate files (variable name as filename?)
        • GDAL standard?
      • “Tabular-ASCII” - lifted from on-prem, or rewritten to Python
        • Textual layout
        • “Row of vars” per time-stamp
        • Blocked per group and/or frequency (count)
      • OPeNDAP / GDAL (HGA): various ASCII format choices ?
        
        
• L3/L4 - HOSS + chaining (ICESat-2, SMAP other?) √ (some reconfig, perhaps tweaks to code?)
  (Note - will be .nc4 output,  not original source .h5 format, but this hopefully is not too significant)
  • • ICESat-2 has some notorious quirkiness - Lower-Left orientation, Edge-Aligned dimension scales
    • Possible support for missing dimension-scales - use CF-ish config entry for grid-parameters (SMAP)
      - see: GeoTagging for a Geo-Located Grid Array
  • Reprojection - TBD (I suspect not HGA, rather PyResample based, Data Reprojection Wizard)
  • Reformat - GeoTIFF/COG (HGA?), other standard Cloud formats, Zarr

• HDF-EOS-4 legacy data - multiple divergent data products
• HDF4 - NetCDF-4 pre-processor.
• OPeNDAP functionality?
• LADDS GeoLoco/Band-Subsetter

Some preliminary thoughts:

• Separate support for Projection (Swath-Projection-Gridding) ?
  • Vs. Regridding (same projection)
  • Vs. Reprojection (Equal-Area to Conformal, Cartesian to Polar, others)
• Feature expansion in Cloud Services to match on-prem
  • Less any no-longer-relevant featues (H5-EOS)
  • “Main Services”: HOSS for L3/L4, Trajectory Subsetter for L1/L2
    • For Variable and Spatial/Temporal Subsetting 
    • BBox and Shape/Polygon Region
    • HDF-5, NetCDF-4
  • + Formats, + Projection/Reprojection
  • Alternative for Legacy (HDF4, HDF-EOS-2)?
    • or updates to OPeNDAP ?
• Back-Port to on-prem - Harmony-on-prem within SDPS ?
  • To potentially retire legacy code base for data not yet migrated
  • Open port on EIL for a localized Harmony access?
  • EDSC/CMR reconfiguration

...