Ticket 066: Stochastic Propagation GeoJSON Export

Goal

Add --format geojson and --format kml output modes to the propagate CLI command. The export renders the mean particle trajectory as a LineString coloured by per-step reserve-violation probability, making it possible to load the propagation result directly into QGroundControl, QGIS, or Google Earth.

Motivation

propagate already computes a per-step timeline with lat_mean, lon_mean, and p_reserve_violation. All the data needed for a map layer is present; it is locked inside a JSON envelope that requires bespoke parsing. A GeoJSON adapter — roughly 80 lines — makes the stochastic result visible without any new computation. A drone operations team can immediately overlay the probability-of-reserve-violation heatmap on the planned route and spot the flight phase where mid-flight risk peaks.

Output Format

A single GeoJSON FeatureCollection with two named layers.

trajectory layer — mean particle path

One LineString over all timeline points, plus per-point properties as a parallel feature array (or embedded in a MultiPoint feature).

{
  "type": "Feature",
  "geometry": {
    "type": "LineString",
    "coordinates": [[lon0, lat0], [lon1, lat1], ...]
  },
  "properties": {
    "layer": "trajectory",
    "propagation_id": "my-propagation",
    "feasibility_rate": 0.92,
    "sample_count": 100,
    "dt_s": 2.0
  }
}

timeline layer — one Point per timeline step

{
  "type": "Feature",
  "geometry": { "type": "Point", "coordinates": [lon, lat] },
  "properties": {
    "layer": "timeline",
    "elapsed_time_s": 42.0,
    "energy_remaining_wh_mean": 812.3,
    "energy_remaining_wh_p5": 740.0,
    "energy_remaining_wh_p95": 890.0,
    "p_reserve_violation": 0.04
  }
}

p_reserve_violation drives point colour: green < 0.05, amber 0.05–0.20, red > 0.20. Standard GeoJSON renderers in QGroundControl and QGIS can apply this styling directly.

KML variant

KML mirrors the GeoJSON structure: one Placemark per timeline step as a Point, plus one Placemark LineString for the trajectory. Colour is encoded as AABBGGRR in the color element, mapped from p_reserve_violation.

Acceptance Criteria

1. propagate --format geojson produces valid GeoJSON that passes geojson.is_valid().
2. The trajectory Feature's LineString has one coordinate pair per timeline point in the same order.
3. Each timeline Feature includes elapsed_time_s, energy_remaining_wh_mean, p_reserve_violation.
4. propagate --format kml produces a KML document parseable by standard XML parsers.
5. Output can be saved with --output and read by QGIS / QGC without modification.
6. Tests: valid GeoJSON structure, coordinate count matches timeline length, KML root tag present.

Scope

• adapters/stochastic_geojson.py — new GeoJSON builder for stochastic results
• adapters/stochastic_kml.py — new KML builder for stochastic results
• adapters/cli.py — extend propagate command's format enum to include geojson and kml; wire new builders
• adapters/cli_support.py — extend _render_stochastic_output with geojson/kml branches
• tests/test_stochastic_cli.py — geojson and kml format tests

Notes

• Reuse the coordinate ordering convention from adapters/geojson_export.py (lon, lat, not lat, lon) to match GeoJSON spec.
• When sample_count == 0 (all particles infeasible), the trajectory is an empty LineString and the timeline layer is empty — output must still be valid GeoJSON.
• estimation_error_timeline and cross_track_timeline are out of scope for this ticket; they can be added later as separate layers.
• Priority: immediately after Ticket 063 (RTH reserve) and Ticket 064 (batch scenario/propagate).