Data Sources & Methodology

Where every field comes from, how often it is refreshed, how conflicts between sources are resolved, and what the data cannot tell you.

What feeds each field

• Accepted names, authorship, publication year — Plants of the World Online (Kew) and Tropicos (Missouri Botanical Garden); re-checked quarterly.
• Occurrence records and country distribution — GBIF, restricted to records determined to species rank with usable coordinates; refreshed monthly.
• DNA sequences and locus assignment — NCBI GenBank, matched back to species names; refreshed weekly.
• Conservation status — IUCN Red List assessments, matched by species name; refreshed quarterly.
• Literature records — Crossref metadata plus manual curation of taxonomic revisions; refreshed monthly.
• Images — GBIF media records and contributing herbaria, each linking to its original record and licence.

How the pipeline runs

1. Ingest. Source records are downloaded and stored with their original identifiers.
2. Reconcile names. Names are matched against accepted names in POWO and Tropicos. Names that cannot be resolved keep their source spelling and are flagged rather than silently rewritten.
3. Derive. Country lists, elevation ranges and genus roll-ups are computed from the underlying records and labelled as derived figures.
4. Publish. Pages without enough independent data to answer a question are kept browsable but excluded from search engines.

When sources disagree

• Taxonomy — POWO is treated as the authority for accepted names, Tropicos for protologue details. Conflicts are shown, not hidden.
• Occurrences — duplicates at the same coordinates and date are collapsed; records with a coordinate uncertainty above 50 km are excluded from country lists.
• Sequences — attributed only from the organism field of the GenBank record, never inferred from an associated publication.

Known limits

• Occurrence maps inherit herbarium collection bias; sampling density reflects where botanists worked.
• Absence of a record is not evidence of absence.
• Elevation ranges are indicative, not physiological limits.
• Sequence coverage is uneven, with well-studied genera far better represented than rare ones.

Corrections

Send correction reports to contact@melastomataceae.org. The procedure is documented in the editorial and corrections policy.