Files
insar-management-system-v2/experiments/isce2_sbas_timeseries/scripts/phase0_todo.md
T

4.4 KiB

Phase 0 Practical TODO

Immediate

  • Run scripts/check_env_ubuntu2404.sh inside Ubuntu-24.04.
  • Use scripts/scan_lt1_stack_candidates.py to keep one baseline sample stack manifest current.
  • Treat E123.3_N46.1 as the first tile-level smoke-test sample unless a better sample appears.
  • Confirm ISCE2 stack-processing scripts are present.
  • Confirm the official helper scripts do not advertise LT-1/LUTAN1 stack prep.
  • Record required orbit, DEM, and metadata adaptations.
  • Run scripts/build_lt1_stack_prep.py to keep the dry-run stack workspace current.

Before first end-to-end run

  • Implement an LT-1 scene materializer that creates YYYYMMDD.slc, YYYYMMDD.slc.xml, and data.
  • Materialize the remaining acquisitions for E123.3_N46.1 under scratch/.../SLC/.
  • Smoke-test the materializer on the reference date 20250510.
  • Run the generated run_stripmap_stack_dryrun.sh preflight and then stackStripMap.py --nofocus.
  • Inspect the produced baseline/, configs/, and run_files/ outputs.
  • Prepare a stack-local DEM to avoid global-DEM bbox behavior during createWaterMask.
  • Add a reproducible synthetic waterMask fallback for run_01_reference when Earthdata credentials are unavailable. Working rule: DEM is already local and sufficient; do not download SWBD during this experiment stage.
  • Extract shared LT-1 input preparation helper for DEM/orbit resolution. Compatibility rule: original D-InSAR entry logic remains in place; only the duplicated input-prep internals were consolidated.
  • Decide MintPy installation strategy after stack generation is stable. Decision: default to a dedicated WSL conda env named mintpy so the working isce2 processing env stays unchanged on the development machine.
  • Freeze the first smoke-test command chain. Frozen chain: run_01_reference -> run_02_focus_split -> run_03_geo2rdr_coarseResamp -> run_04_refineSecondaryTiming -> run_05_invertMisreg -> run_06_fineResamp -> run_07_grid_baseline
  • Execute run_01_reference through the WSL wrapper and verify the fallback-recovered geometry outputs.
  • Execute run_02 to run_07 and record LT-1-specific failures if they appear. Result: all stages exited 0 in Ubuntu-24.04. run_04_refineSecondaryTiming logs still contain Bad match at level 1 and correlation error, but pair-level misreg, date-level misreg, merged SLC, and merged baseline products were all generated successfully.

Next Focus

  • Run scripts/install_mintpy_runtime_ubuntu2404.sh in Ubuntu-24.04 and verify the new env. Result: dedicated WSL env mintpy was created successfully and smallbaselineApp.py / prep_isce.py are available.
  • Validate MintPy ingestion against the current stack_work/merged/ outputs. Result: build_lt1_stack_prep.py --workflow interferogram plus run_08_igram produced Igrams/*/filt_*_snaphu.unw, and prep_isce.py completed successfully after bridging the working isce2 Python package into the mintpy env.
  • Draft the first smallbaselineApp.cfg for the LT-1 sample stack. Result: configs/sample_smallbaseline_lt1_e123p3_n46p1.cfg now records the first runnable LT-1 stripmapStack -> MintPy SBAS contract.
  • Execute the first MintPy workflow steps after prep_isce.py. Result: the repo-local smoke-test chain now reaches radar-coordinate timeseries.h5 and velocity.h5 in stack_work/mintpy_sbas_v5/. Current helper chain:
    • scripts/run_mintpy_with_isce_ubuntu2404.sh
    • scripts/create_mintpy_all_ifgram_mask.py
    • scripts/run_smallbaselineApp_patched.py
    • scripts/run_mintpy_sbas_smoketest_ubuntu2404.sh
  • Draft the first production-side SBAS artifact manifest and publish contract. Result:
    • configs/sample_psinsar_manifest_lt1_e123p3_n46p1.json
    • docs/ISCE2_SBAS_TIMESERIES_DESIGN.md

New Follow-up

  • Decide whether production should keep the repo-local patched MintPy launcher or pin an upstream-fixed MintPy version.
  • Add the geocode/export stage needed for publishable SBAS rasters and previews. Result: experiment-layer publish export now succeeds into publish/mintpy_sbas_v5/ with geocoded HDF5, GeoTIFF, preview PNG, and manifest.json.
  • Wire the validated SBAS runtime chain into backend workflow submission and artifact publishing.
  • Run a separate unified-environment experiment by cloning the current WSL isce2 env and installing MintPy directly inside it.