Earth Analog Finder

This is a frontend prototype. It runs on a curated sample dataset bundled with the app: eight of thirteen parameters are measured and reproducible, five are our estimates. A live NASA data pipeline is planned.

We rank 24 real Earth sites against the conditions of a permanent Moon base or Mars base, across 13 parameters. Every score is shown with the working behind it, so you can check it or change it.

Profile:Equatorial/mare global-mean reference for a permanent lunar base. South-pole (e.g. Artemis) targets are future work.

No Earth site can match these

Temperature range, Radiation dose, Daylight score zero for all 24 sites, not because the data is missing but because Earth cannot do them: the atmosphere and magnetic field shield the surface from the dose rate, and no Earth day lasts 336 hours. That is 19% of the total weight, so it is left out of the weighted mean and reported here instead. Averaging it in would only scale every score by the same constant and hide the finding.

Search site or countryMinimum score: 0Sort by

Showing 24 of 24 sites

Ranked sites (Moon base)

#	Site	Score	Conf.
1	Atacama Desert Chile	48.7	81
2	McMurdo Dry Valleys Antarctica	46	85
3	Devon Island Canada	41.4	85
4	East Antarctic Plateau Antarctica	39.6	80
5	Erta Ale Ethiopia	38.9	63
6	Craters of the Moon United States	37.9	85
7	Namib Desert Namibia	37.3	68
8	Mauna Kea United States	35.3	56
9	Deception Island Antarctica	35.1	69
10	Pilbara Australia	34.9	80
11	Wadi Rum Jordan	34.5	85
12	Svalbard Norway	33.8	80
13	Mojave Desert United States	33.2	81
14	Ladakh India	33	60
15	Negev Desert Israel	32.5	85
16	Teide Spain	31.4	56
17	Lanzarote Spain	31.1	85
18	Kilauea United States	29.1	77
19	Mount Etna Italy	27.8	60
20	Sonoran Desert United States	27	80
21	Danakil Depression Ethiopia	26.6	55
22	Death Valley United States	25.5	81
23	Mono Lake United States	23	85
24	Rio Tinto Spain	16.1	90

Custom profile

Move any weight and the ranking recomputes. Presets can move a target too, and the panel always shows which numbers changed.

Edit all 13 weights

Mean temperature (°C)1.20Temperature range (°C)0.60Radiation dose (µSv/day)0.80Terrain slope (°)0.70Soil composition (index)1.00Atmospheric pressure (kPa)0.80Daylight (h/day)0.40Precipitation (mm/yr)1.10Relative humidity (%)0.60Dust activity (index)0.70Isolation (index)0.80Elevation (m)0.30Aqueous geochemistry (index)0.50

Site detail

Pick a marker on the map or a row in the table to see the full breakdown.

Methodology at a glance

For each parameter, similarity is 1 minus the distance to the target divided by the tolerance, clamped between 0 and 1. The score is 100 times the weighted sum of those similarities divided by the sum of the weights, taken over the axes a site can actually score on. Confidence is completeness times resolution tier times 100, and it is reported separately. Targets, tolerances and weights all sit in src/data/config.json, and scripts/validate.mjs re-runs the whole thing in the terminal.

What you can do next

Read where every number comes from in the provenance doc.
Re-run the scoring yourself with npm run validate.
Edit a weight in src/data/config.json and watch the ranking move.
Export the current ranking as CSV and check it in a spreadsheet.