Satellite constellations for disaster response

43.9% shorter mean revisit gap than the standard 48-hour single-satellite benchmark

Best simulated orbits over eight disaster regions, against a fixed 48 h figure. Held-out R² 0.892.

Sweeping toward California · 0 passes over California · mission clock 0 h 00 m

Region

Wildfires/Earthquakes · 15.6 h predicted revisit gap

Predicted gap, California: 15.6 hR² 0.892 held out
Fixed 48 h single-satellite benchmarkBest simulated 22.8 h · upper bound 90.8 h · LOW confidence

Time compressed: 1 s ≈ 30 min at 1×. The sweep is an illustration; the gaps are the measured and predicted values. Path lines join computed points.

Predicted revisit gap by region, with the held-out R² 0.892
RegionPredicted gapChange
15.6 h+31.7%
13.6 h+31.5%
12.1 h+30.0%
11.4 h+20.8%
35.4 h+8.7%
37.5 h+7.9%
34.0 h+2.8%
27.9 h-4.2%

48 h is a fixed benchmark figure, not a simulation. Predictions, not flown orbits. LOW confidence in every region, with wide upper bounds.

How this works

Hours-long orbital simulations were replaced with a learned model, then satellite layouts were searched for faster disaster coverage. Across eight disaster regions the best simulated orbits average 26.9 h between passes, 43.9% shorter than the fixed 48-hour single-satellite benchmark.

The globe shows four satellites on representative 800 km sun-synchronous orbits, the benchmark orbit described in the validation script. Positions come from real SGP4 propagation; the layout is illustrative. The gap figures are the measured and predicted values from the result files.

Limits

  • The 48 h benchmark is a fixed figure, not simulated here.
  • Three regions' best gaps may come from generated rows.
  • Predictions, not simulator runs of the optimised orbits.
  • LOW confidence in every region, with a wide upper bound.
  • No orbit flown.

Numbers

  • Mean best simulated gap: 26.9 h over eight regions.
  • Headline: 43.9% shorter than 48 h.
  • Held-out R² 0.892.
  • Rows: 7,000 simulated + 4,200 generated = 11,200.