insar-management-system-v2/experiments/isce2_sbas_timeseries/notes/PHASE3_PUBLISH_EXPORT_SMOKETEST.md

Phase 3 Publish Export Smoke Test

Updated: 2026-04-06

Goal

Validate that the successful MintPy SBAS experiment can be converted into a publish-style artifact bundle without touching the main system.

Inputs

Source MintPy work directory:

• /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/stack_work/mintpy_sbas_v5

Source stack:

• group key:
  • LT1A|STRIP1|HH|DESCENDING|E123.3_N46.1
• dates:
  • 20250118
  • 20250315
  • 20250510
  • 20250705
  • 20250830

Successful Command

wsl -d Ubuntu-24.04 bash /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scripts/export_mintpy_publish_products_ubuntu2404.sh /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/stack_work/mintpy_sbas_v5 /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/publish/mintpy_sbas_v5

Current Export Scripts

• scripts/export_mintpy_publish_products_ubuntu2404.sh
  • geocode core MintPy outputs
  • convert selected outputs to GeoTIFF
  • copy preview and metadata files
  • build root manifest.json
• scripts/build_mintpy_publish_bundle.py
  • generate preview/velocity_preview.png
  • generate metadata/source_quality_summary.json
  • generate publish-style root manifest.json

Geocode Contract

Current experiment settings:

• lookup source:
  • inputs/geometryRadar.h5
• output pixel size:
  • latitude step:
    • -0.000185185
  • longitude step:
    • 0.000185185
• interpolation:
  • nearest

Observed geocoded grid:

• extent:
  • south:
    • 45.80391
  • north:
    • 46.42206
  • west:
    • 122.930244
  • east:
    • 123.76654
• shape:
  • rows:
    • 3338
  • columns:
    • 4516

Generated Publish Bundle

Successful publish-style directory:

• /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/publish/mintpy_sbas_v5

Main outputs:

• manifest.json
• assets/geo_timeseries.h5
• assets/geo_velocity.h5
• assets/geo_temporalCoherence.h5
• assets/geo_maskTempCoh.h5
• assets/velocity.tif
• assets/temporalCoherence.tif
• assets/maskTempCoh.tif
• preview/velocity_preview.png
• preview/numTriNonzeroIntAmbiguity.png
• metadata/smallbaselineApp.cfg
• metadata/source_quality_summary.json

Current Interpretation

This confirms:

• the experiment now supports radar-coordinate MintPy inversion
• the experiment also supports geocoded HDF5 exports
• the experiment can produce publish-style GeoTIFF outputs
• the experiment can build a stable manifest-driven bundle outside the MintPy work directory

This does not yet confirm:

• direct backend catalog registration
• frontend rendering against the real system APIs
• whether EPSG:4326 should remain the final publish CRS decision

Important Notes

1. Current CRS assumption

save_gdal.py warned that no explicit EPSG metadata was found and assumed:

• EPSG:4326

Current judgment:

• acceptable for this experiment because the geocoded outputs are in latitude/longitude grids
• should still be checked when formalizing the production publish contract

2. Group key in the generated manifest

Because PowerShell treats | specially, passing group keys on the command line is awkward from Windows.

Current practical rule:

• the sample manifest stored in git remains the clean reference:
  • configs/sample_psinsar_manifest_lt1_e123p3_n46p1.json
• the generated publish bundle manifest can be post-filled or generated from backend metadata later

System Embedding Implication

At this point the experiment-layer chain is split cleanly into three stages:

1. stripmapStack preprocessing
2. MintPy SBAS inversion
3. geocode + publish-bundle export

That means the future system workflow can wire them as separate workflow steps without changing the validated experiment logic first.