Skip to main content

Phase 3 — Simulation & Digital Twin

Duration: ~10 working days
Critical-path predecessor: Phase 2 (locked)
Status: in progress

Deliverables (§5.5)​

  1. iSnobal multi-layer snowpack — energy + mass balance, Kojima compaction
  2. 1-D Saint-Venant — HLL Riemann solver for valley routing, CFL ≤ 0.5
  3. 2-D Boussinesq shallow-water — Lax-Friedrichs + semi-implicit friction (PyTorch, GPU-ready)
  4. Voellmy-Salm avalanche — release-zone probability, depth-averaged friction closure
  5. Quantum Amplitude Estimation — Qiskit Aer IAE with classical MC fallback
  6. NVIDIA Earth-2 client — FourCastNet + CorrDiff (NIM), staging-mode synthetic
  7. Buckingham Π similitude — automated dimensional-invariance test suite
  8. Risk aggregator — fuses snow / hydro / avalanche outputs → HazardAssessment

Acceptance Gate​

72-h probabilistic forecast with documented confidence intervals across all sub-basins.

cd snow-ir-backend
python -m snow_ir.orchestration.forecast_pipeline
pytest tests/simulation -v
curl http://localhost:8000/forecast/latest | jq '.[0]'

Frontend shows the ForecastPanel with severity badges, probability bars, and a critical-count badge in the operational header.