insar-management-system-v2/experiments/isce2_sbas_timeseries/notes/PHASE4_UNIFIED_ENV_DECISION.md

Phase 4 Unified-Environment Decision

Updated: 2026-04-06

Decision

Current experiment preference:

• prefer the unified WSL environment for SBAS experiment work

Current production safety rule:

• do not replace or mutate the existing D-InSAR production environment
• keep pair-oriented D-InSAR on the existing WSL isce2 env
• keep the current backend/public D-InSAR entry unchanged

Current environment split:

• D-InSAR production baseline:
  • isce2
• SBAS experiment preferred runtime:
  • isce2_mintpy_v1

Why This Decision Is Reasonable

The unified env has now completed the full current experiment chain:

• MintPy command invocation without the isce bridge
• load_data
• strict-mask generation
• modify_network -> velocity
• publish export to geocoded HDF5, GeoTIFF, preview, and manifest.json

This makes the unified env a valid experiment baseline.

At the same time, the existing pair-oriented D-InSAR route is already working and should not be destabilized just to simplify the SBAS experiment runtime.

The safest rule is therefore:

• let SBAS experiments move forward in a separate unified env
• do not touch the current isce2 production env used by D-InSAR

Evidence

Successful unified env:

• /home/administrator/miniconda3/envs/isce2_mintpy_v1

Successful unified SBAS work directory:

• /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/stack_work/mintpy_sbas_unified_v1

Successful unified publish directory:

• /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/publish/mintpy_sbas_unified_v1

Matched quality indicators:

• maskAllValid:
  • 1219001 / 4076199
  • 29.91%
• maskTempCoh:
  • 62987 / 4076199
  • 1.55%

Key package difference:

• current isce2 env does not provide the MintPy-side package set needed for this SBAS route
• current isce2_mintpy_v1 env includes:
  • mintpy
  • cartopy
  • pyaps3
  • pykml
  • cvxopt

What This Decision Does Not Mean

It does not mean:

• the backend should immediately switch to unified-env execution
• the bridge route must be deleted now
• the current D-InSAR runtime should be modified

It only means:

• for the next experiment steps, unified env is the preferred path
• for current production safety, isce2 remains untouched

Required Guardrails

For the next phase, keep these rules:

• do not install MintPy into the existing isce2 env
• do not redirect current D-InSAR scripts to isce2_mintpy_v1
• do not remove the bridge-based helpers yet
• keep all SBAS work in experiment scripts, notes, and scratch directories

Reproducibility

Environment snapshots should be exported and kept with the experiment record.

Current snapshot command:

wsl -d Ubuntu-24.04 bash /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scripts/export_phase4_env_snapshots_ubuntu2404.sh /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/configs/env_snapshots/20260406

Next Stable Experiment Steps

Before any system integration work:

1. Keep the unified env as the default SBAS experiment runtime.
2. Preserve environment snapshots for both isce2 and isce2_mintpy_v1.
3. Write one comparison note focused on:
   • runtime simplicity
   • reproducibility
   • remaining workarounds
   • risk to D-InSAR production
4. Optionally repeat the chain on one more LT-1 sample stack.
5. Only after the experiment is stable, design the system integration boundary.