From a reference to a scientific record

A before-and-after investigation begins with a deceptively simple request: find observations around an event. A defensible result also needs the query scope, the archive state, temporal rules, acquisition metadata, exclusions, computational method, and interpretation limits.

LunarTrace makes those transitions explicit. It does not replace an archive, a photogrammetry system, or a planetary image-processing environment. It connects discovery and inspection to a reproducible study framework.

The research chain

reference → bounded query → observations → temporal roles → pairs → screening → evidence inspection → scientific record

Stage	Input	Output	Important boundary
Reference	A location, time, convention, and uncertainty	A research anchor	A user-defined point is not independently established event localization
Discovery	Source scope and place/time query	Captured response population	Retrieval closure and archive freshness are separate
Normalization	Source metadata	Identified observations with explicit missing fields	Missing values cannot become zero or guessed measurements
Partition	Acquisitions and reference interval	Before, after, and excluded/overlap groups	PRE and POST depend on this investigation
Pair construction	Before and after observations	A bounded Cartesian population	A pair is a candidate relation, not comparable imagery
Screening	Candidate metrics and declared objectives	Evaluable and nondominated sets	A frontier is relative to the analyzed population
Inspection	Source identity and available evidence	Human understanding and next decisions	Viewing does not establish registration or change
Record	Canonical computation, dependencies, and review	Traceable claims and evidence	Integrity, interpretation, and admission remain distinct

Three deliberately different outputs

Editable Lab investigation

The Lab stores the user’s current intent and a bounded source capture. It supports interactive selection, filtering, comparison exploration, and export. These are useful research operations, but they do not make a canonical CompiledStudy or automatically admit claims.

Canonical Python candidate

The Python compiler consumes typed event, protocol, snapshot, and observation objects. It validates their relationships and produces a structured CompiledStudy. Exported bundles retain dependencies for verification and replay where supported.

A candidate produced by code is not equivalent to a scientifically endorsed interpretation. Human review and the proof ceiling still matter.

Historical aggregate

The retained Chang’e 6 case has a different evidentiary shape. It preserves historical counts, protocol definitions, reference identities, and two real browse assets, but not the original observation/pair rows. The appropriate result is a traceable historical record, not a manufactured replay.

Why the separation matters

A polished interface can accidentally make a weak relationship appear strong: a marker suggests precise geography, a pair of images suggests alignment, a green check suggests truth, or a complete-looking table suggests complete retrieval.

LunarTrace instead asks each representation to carry only the claim its source supports. This does not require a passive interface. New source-backed discovery can be rich and interactive precisely because its actual rows, times, metadata, and receipts exist.

Interoperability rather than enclosure

Useful work should leave the browser. Draft URLs support communication of intent. Captured exports support inspection of source-response bytes and research choices. The Python package supports explicit acquisition, compilation, artifact export, and verification.

For tasks such as geometric registration, calibrated photometry, detailed terrain analysis, or a scientific change claim, use the appropriate external processing and evidence. The current Lab does not silently claim those capabilities.