Aetrova · SouthWatch

At the Moon's pole the Sun skims the horizon, so terrain, not latitude, decides whether a site has power and a line to Earth. Space trackers show what's around the Moon; SouthWatch shows whether one spot works, when, and how sure we are. My rule for it: 5% real accuracy beats 100% fake accuracy.

• Sunlight at a polar site depends on crater rims and massifs tens of kilometres away.
• Earth contact depends on libration and on which ground station can see the Moon, not on Earth's centre.
• The raw elevation data is gigabytes, far too big for a browser.
• Scrubbing through time has to answer within one frame.

1. 1

   Offline Python pipeline

   NASA LOLA elevation (80°S at 80 m, 60°S at 120 m, plus the global model) becomes small Float32 heightmaps and horizon masks sampled every 0.25°.

2. 2

   Astronomy in the browser

   astronomy-engine gives Sun, Earth and Moon positions; a frame chain turns them into elevation and azimuth on the local horizon of the chosen site.

3. 3

   Physics layer

   The Sun as a limb-darkened disk, eclipses by Earth's shadow, NASA Deep Space Network stations with a 10° keep-out, and live signal light time.

4. 4

   Web Worker

   Year-long illumination and Earth-contact sweeps run off the main thread, so the time scrubber never drops frames.

5. 5

   three.js scene

   Real terrain draped with NASA Moon Trek imagery: pole, 60°S and whole-Moon views in 3D, 2D and from the ground.

• Stress-test the physics before writing it

  Why: I reviewed the plan from a physics angle and a mission-operations angle and logged 25 gaps, 6 of them critical (horizon reach, azimuth resolution, ground stations vs Earth's centre, eclipses, validation pairing, a region vs a point). They were fixed in the formulas first.

  Trade-off: Weeks of preparation before any feature work.

• Heavy terrain work offline, time-dependent maths in the browser

  Why: What never changes is computed once in Python; anything tied to the time slider runs live, so scrubbing answers within one frame.

  Trade-off: A data pipeline to maintain alongside the app.

• No server at all

  Why: A static app can't go down during judging, and there's no user data to store.

  Trade-off: Click-anywhere horizon calculations are limited to what the browser can do.

• Every number carries a MEASURED / COMPUTED / MODELED tag

  Why: A planner can see exactly how much to trust each value, and every source is credited on screen.

  Trade-off: More UI, and some values shown with honest caveats instead of false precision.

• Aetrova wordmark and a black-glass interface with one teal accent reserved for state.
• Each value has an icon, a trust tag and an info tooltip, so every simulation explains itself.

• Built and running: real terrain, 39 NASA data layers, Sun, Earth, eclipse and ground-station physics, 108 tests passing.
• Entering NASA Space Apps Challenge 2026 (14–15 November).

What's next

• Golden tests against NASA SPICE reference values.
• Validation against NASA's published illumination maps.