Atlas Botánico del Perú

Problem

Peru's plant and fungal diversity is scattered across GBIF, WCVP and APG IV with no single, synonym-resolved view. Raw occurrence counts overstate diversity because they never collapse different names for the same accepted species before counting.

Approach

A reproducible ETL over GBIF, WCVP and APG IV → static marts → a backend-free site. Synonyms get resolved before anything gets counted — which is why the figures come out lower, and more defensible, than sources that count raw names.

The pipeline runs ten reproducible stages over roughly 1.3 million raw GBIF occurrence records (1.29 million for Plantae alone): download, taxonomic matching against WCVP’s accepted-name backbone for plants and Index Fungorum’s backbone for fungi, modeling, aggregation, analysis, export and a data-quality profile. Every download is DOI-stamped, so any figure on the site traces back to an exact, citable snapshot rather than a live query that could silently change.

Matching against an accepted-name backbone is the actual work: raw GBIF records include synonyms, misspellings and outdated names, and none of those should be counted as a distinct species. After matching, 99.9% of plant occurrences and 98.75% of fungal occurrences resolve to an accepted name — the unresolved remainder is reported rather than silently dropped or folded into the total.

The eight resulting marts (richness by department, family composition, lifeform spectrum, described-per-year, and more) are plain JSON and Parquet files read at build time with a direct readFileSync, so the published site issues zero runtime fetches and needs no backend or database server to serve a navigable taxonomic tree and department choropleth.