Ranks 24 Earth sites by how closely they match the conditions of a permanent Moon base or Mars base, and shows the working behind every score.
Live demo: https://cosmo-code-mayer-doa.vercel.app/
Built for the NASA Space Apps Challenge 2026 challenge Identify Earth Locations that Analog the Permanent Moon Base Locations and Mars. Mission teams use Earth sites to test equipment, train crews and study isolation before flying. This tool makes the choice of site explicit and checkable: each score is a transparent weighted sum that can be traced to its inputs and recomputed with your own priorities.
- Interactive map of all 24 sites, with markers coloured by score and dashed outlines for low-confidence sites. Street, satellite and topographic layers.
- Moon base and Mars base profiles, plus a custom profile builder: change any weight and the ranking recomputes instantly. Presets (Lunar geology, Mars arid, Mars aqueous) show how priorities change the order.
- Per-site breakdown across 13 parameters: value, target, tolerance, similarity, weight and contribution, with each parameter labelled measured, derived or estimated.
- Structural mismatch reporting. Axes that no Earth site can match (for example the lunar radiation dose) are set aside and shown separately, not averaged into every score.
- Separate confidence rating per site, so a well-scoring but poorly resolved site is visible.
- Monte Carlo sensitivity range per site, so you can see how far a score moves when the tolerances and weights are varied.
- Search, filter, sort, and CSV export.
Requires Node.js 18.17 or newer. No API keys or accounts are needed.
npm install
npm run dev # http://localhost:3000| Command | What it does |
|---|---|
npm run build |
Production build |
npm run lint |
ESLint |
npm run typecheck |
TypeScript check |
npm run validate |
Re-runs the scoring in the terminal and checks the invariants |
npm run sensitivity |
Weight-perturbation test of how stable the rankings are |
npm run uncertainty |
Monte Carlo score range for every site under both profiles |
The scoring and data work offline. Only the map tiles need a network connection.
similarity = clamp( 1 - |value - target| / tolerance , 0 , 1 )
score = 100 * sum( weight * similarity ) / sum( weight ) over scorable axes only
confidence = completeness * resolutionTier * 100
A score is a comparative similarity to a stated profile, on the axes an Earth site can reach. It is
not a go/no-go engineering criterion and should not be read as "percent similar to the Moon".
Every target, tolerance and weight is in src/data/config.json. See
METHODOLOGY for the full model and a worked example.
Of the 13 parameters, 9 are measured, 3 are derived by calculation, and 1 is an estimate
(aqueous geochemistry). The measured values can be regenerated with the scripts in scripts/.
| Source | Used for |
|---|---|
| NASA POWER | Temperature, temperature range, pressure, humidity, precipitation, dust |
| SRTM 30 m / ASTER 30 m | Elevation and slope |
| ISRIC SoilGrids (CC BY 4.0) | Soil texture |
| GeoNames | Isolation |
| LRO CRaTER, Curiosity RAD | Moon and Mars radiation targets |
python3 scripts/fetch_power.py
python3 scripts/fetch_dem.py
python3 scripts/fetch_extra.py
python3 scripts/fetch_soil.py # needs: pip install rasterio pyprojOr run the whole pipeline in one command, which also applies the elevation correction and
rewrites src/data/sites.json:
python3 scripts/refresh_data.pyFull details, including the elevation correction applied to the NASA POWER grid, are in PROVENANCE.
- The app ranks a fixed list of 24 curated sites. It does not search for new analog locations.
- The Moon profile is an equatorial / mare global-mean reference. Lunar south-pole targets are not modelled.
- Weights and tolerances are the team's judgement.
npm run sensitivitymeasures how much the rankings move when the weights change. - Aqueous geochemistry is an estimate, and three Antarctic sites use estimated soil values because the soil dataset has no Antarctic coverage.
- Validation is internal consistency and weight sensitivity. It is not a comparison against an independent reference.
- The Monte Carlo range covers tolerances and weights only. It is a sensitivity, not a measurement error on the input data.
The full list is in METHODOLOGY.
src/
app/ page.tsx (main view), layout.tsx, docs/ and api/docs/ (documentation pages)
components/ map, filters, ranked table, site detail, profile builder
lib/ scoring.ts (the model), data.ts (loading, filtering, CSV), types.ts
data/ config.json (parameters, profiles, presets), sites.json (24 sites)
scripts/ validate.mjs, sensitivity.mjs, uncertainty.mjs, scoring.mjs,
refresh_data.py, fetch_*.py (data retrieval)
docs/ METHODOLOGY.md, PROVENANCE.md, VALIDATION.md, DATA_DICTIONARY.md, FAQ.md
- Methodology: the scoring model, assumptions and limitations
- Data provenance: where each number comes from
- Validation: automated checks, sensitivity and score ranges
- Data dictionary: every field in the two data files
- FAQ: common questions about the score and the estimates
The same pages are served by the app at /docs.
- Code: AI tools assisted with bug fixes and code reviews.
- Submission video: AI tools assisted with video editing.
- Audio: Some sounds in the submission video were AI-generated.
Map tiles: OpenStreetMap, OpenTopoMap and Esri World Imagery. Data sources are listed above and credited in PROVENANCE. This project is not affiliated with or endorsed by NASA.
MIT. Data from third-party sources remains under its own license.
Team CosmoCode.