ISRO-BAH-26¶

Subsurface water-ice mapping and rover traverse planning at the lunar south pole, using Chandrayaan-2 DFSAR radar together with public LRO data. Built by team Nakshatra for the ISRO Bharatiya Antariksh Hackathon 2026.

Technical report (PDF): architecture, every formula used in the code with its real constants, results and limitations.
Pitch deck (PDF): the hackathon presentation.
Source code

What it does¶

The study area is a 60 × 60 km window around Faustini crater on the LOLA 20 m south-polar DEM.

Data. download_and_clip.py streams the DEM crop. ice_data.py resamples the LOLA PSR map, Diviner maximum temperature and depth to ice stability, Mini-RF S-band CPR, LAMP Lyman-α ratio and the DFSAR L-band Level-3C mosaics onto the same grid.
Polarimetry. dfsar.py computes CPR and DOP from compact-pol or full-pol data.
Ice probability. ice_model.py fuses roughness-corrected radar anomalies, LAMP and PSR evidence in log-odds space, multiplies by a thermal-stability term (σ((110 K − T_max)/5 K)) and propagates weight uncertainty by Monte Carlo.
Depth to ice. The L/S CPR excess ratio is inverted for the depth of the ice top with a two-layer attenuation model (e-folding depths 0.5 m for S, 1.5 m for L), calibrated on cold pixels and accepted only if it passes a cold-versus-warm enrichment test.
Volume. Expected ice volume in the top 5 m, split into the directly sensed 0–3 m and the extrapolated 3–5 m, for ice fractions of 1, 5 and 10 %.
Traverse. cost_map.py builds a slope cost map and path_planner.py runs 8-connected A* with an octile heuristic from the Faustini rim to the floor.

Results (Faustini, 20 m)¶

Metric	Value
Mean P(ice) inside / outside PSR	0.52 / 0.06
Area with P(ice) > 0.5	603 km²
Expected ice volume, 0–3 m (f = 1 / 5 / 10 %)	2.1 × 10⁷ / 1.1 × 10⁸ / 2.1 × 10⁸ m³
Expected ice volume, 3–5 m, extrapolated	1.4 × 10⁷ / 7.1 × 10⁷ / 1.4 × 10⁸ m³
L/S depth signal significant	No (Diviner depth to stability used)

Ice probability evidence layers

DFSAR L/S depth classes

Status¶

The ice model, depth inversion, volume estimate and A* planner are implemented; the polarimetry and depth inversion are covered by synthetic tests. The weighted multi-term cost map, PPO agent, rover–relay–orbiter simulation and OHRC hazards are planned; see the limitations section of the report.

Citing¶

If you use this work, please cite it using the metadata in CITATION.cff.