Tool 04 · Cross-mission reasoning

Discovery Engine

Links observations of the same quantity from different missions and instruments, finds excursions and disagreements, then tests each one: is it replicated, does it follow a processing version, is it confounded by when it was observed? Every finding ends with a specific question to investigate next.

Catalog

About 700 synthetic Mars observations from seven made-up instruments on orbiters, a lander and a rover. Hidden inside: a real dust storm, a real warm spot, a calibration change, a single bad frame and an unexplained methane signal.

One row per observation. Required columns:

id,instrument,quantity,value,sigma,lat,lon,day,local_time,proc_version,unit
OBS-1,ORB-A TIR,surface_temp_K,203.4,1.2,12.1,-31.8,214,3,v1,K

Engine settings

Link distance

Excursion threshold

Disagreement threshold

Excursions use robust z-scores against the same instrument's own baseline: its site history for landers and rovers, or a robust latitude-trend fit for orbiters. Disagreements use the paired difference divided by the combined measurement error.

How a verdict is reached

Likely real when independent instruments see it, it holds across processing versions, and it spans several observations.
Likely artifact when it tracks a processing version, shows the same offset everywhere, or rests on one point that another instrument contradicts.
Unresolved when the instruments that could confirm it never observed under the same conditions.

Cross-mission links

Each cell counts observation pairs of the same quantity within the link distance and time window. Color shows how well they agree: median |Δ| / combined error.

Findings

Questions to investigate next

Catalog preview

Real archives to build a catalog from: the NASA Planetary Data System and its search API, and ESA's Planetary Science Archive. The engine runs entirely in your browser; imported files are never uploaded.