Resources and Data Usage Attribution

Primary Data

Primary data consists of butterfly images and their associated metadata. These datasets are courtesy of museum providers. We downloaded the metadata from data aggregators (mainly GBIF, iDigBio, Ecdysis, and SCANBUGS). The images were downloaded directly from the institutions through the links provided in the metadata.

Taxonomy Matching

The scientific names in museum records can be outdated, misspelled, or since synonymized. We update the recorded names to match the most current taxonomy in a Catalogue of Life release. This is done using automated methods, with no manual curation step. We reserve the original recorded name and display it alongside the accepted name when they differ, so you can compare them and judge for yourself.

The pipeline normalized the names first: underscores become spaces, whitespace is collapsed, case is folded, and parenthetical subgenera are dropped. We then attempt to match the normalized name against the Catalogue of Life release and use the categories below to describe the outcome of the match.

Matching Categories

Each name ends in one of three states, shown as a badge on the specimen:

Outcome	Meaning
Matched	One accepted taxon was resolved with sufficient evidence.
Ambiguous	Candidates were found, but the evidence did not identify one accepted taxon.
Unmatched	No eligible accepted taxon was found, or the input was invalid or unsupported.

How a match is made

Candidates are identified using seven methods below and ranked by score. Spelling or fuzzy matches are accepted only when unique or clearly outperform the runner-up. Otherwise, the result is ambiguous.

Method	Rule
Exact accepted	The normalized input name exactly matched an accepted name usage.
Exact synonym	The normalized input name exactly matched a synonym of the accepted taxon.
Exact canonical	The parsed genus and epithet exactly matched an accepted canonical binomial.
Unique family epithet	One accepted taxon matched the input family and specific epithet.
Spelling genus	The family and epithet matched while genus spelling similarity resolved the taxon.
Spelling epithet	The family and genus matched while epithet edit distance resolved the taxon.
Fuzzy typo	Family-restricted spelling similarity resolved both genus and epithet.

The full specification — normalization, candidate methods, scoring, the rank cascade, and the ambiguity rules — is published with the col-taxonomy project.

Taxonomy is reconciled against the Catalogue of Life (v2026-08-26 XR). Each recorded name is matched to an accepted Catalogue of Life usage.

Coordinate Matching

A specimen's origin are usually georeferenced in two ways: as a written locality (country, state or province, and locality) and geographic coordinates (latitude and longitude). These independent sources of information may disagree because of transcription errors, reversed coordinate signs, or coordinates assigned later using a different gazetteer. We develop a method to automatically check every coordinate against its associated written locality.

Coordinates are validated and compared against GADM administrative boundaries. The identified region is compared with the recorded country and state/province. A region is assigned only when exactly one boundary matches. Coordinates outside all boundaries or overlapping multiple boundaries fall into other categories described below.

Verdict	Meaning
No coordinate	No coordinate was recorded.
Out of range	The coordinate is outside the range of valid latitudes or longitudes.
Zero coordinate	The coordinate is 0, 0.
No region matched	The coordinate falls outside every known boundary.
Ambiguous region	The coordinate falls where several regions meet.
Country mismatch	The coordinate falls in a different country than the one recorded.
State mismatch	The coordinate falls in a different state or province than the one recorded.
Valid	The coordinate matches the recorded locality.

The component checks behind each category are described in the tables below. Click to expand.

Coordinate

Check	Meaning
Missing coordinate	Latitude or longitude was missing, non-finite, or could not be parsed.
Latitude out of range	Latitude was outside -90 to 90 degrees.
Longitude out of range	Longitude was outside -180 to 180 degrees.
Zero coordinate	Both latitude and longitude were zero.
Valid coordinate	Both coordinates were finite and in range.

Country

Check	Meaning
Country match	Recorded and coordinate-derived countries agree.
Country mismatch	Recorded and coordinate-derived countries differ.
Country not provided	No recorded country was supplied.
No reference match	The coordinate intersected no reference region.
Ambiguous reference	The coordinate intersected multiple distinct reference regions.
Not evaluated	Country was not checked because coordinates were invalid.

State or province

Check	Meaning
Adm1 match	Recorded and coordinate-derived ADM1 values agree.
Adm1 mismatch	Recorded and coordinate-derived ADM1 values differ.
Adm1 not provided	No recorded ADM1 value was supplied.
No adm1 reference match	The reference region had no ADM1 name.
No reference match	The coordinate intersected no reference region.
Ambiguous reference	The coordinate intersected multiple distinct reference regions.
Not evaluated	ADM1 was not checked because coordinates were invalid.

Coordinate validation uses fixed geographic boundaries and may produce mismatches due to border proximity, administrative changes, or differences in locality precision. A mismatch warrants review but does not necessarily indicate an error. The original locality and coordinates are preserved alongside the region identified by the automated check.

Source: GADM

Real-Time Occurrence Data

In addition to the primary collection data, real-time occurrence records are displayed in the species overview section. These records are retrieved directly from GBIF using the species name.

The occurrence data is sourced from the Global Biodiversity Information Facility (GBIF).

Trait Datasets

Genetic Data

Genetic data is sourced from the United States National Center for Biotechnology Information (NCBI). BioCosmos queries realtime data using NCBI Datasets API. It includes only genes that have been sequenced for the queried species.

Wikipedia

We use Wikipedia as a supplementary source of species information. Its content is subject to Wikipedia's licensing requirements.

Literature

The literature data is obtained from CrossRef. BioCosmos filter the data to get the most relevant information for literature lists.