What we mean when we refuse a column

When we cannot interpret something in your file, we say so and stop — we never guess, because a wrong interpretation is impossible to tell from a right one once it is published. Every refusal carries one of the codes below.

Most files produce a handful of these, not hundreds: each one is a decision about a whole column rather than about a row. A refusal is not a rejection of your data — it is the part of the report that tells you what still needs a human.

63 codes in 21 groups. This page is generated from the same registry the checker emits from, so it cannot drift from what you were actually told.

Opening the file

file.bad_container: the .gz/.zip wrapper itself would not open
file.empty: readable, and there is nothing in it
file.unreadable: no table could be produced from these bytes at all

Character encoding

encoding.not_utf8: the bytes are not UTF-8 and the fallback is a guess

Zip archives

zip.ambiguous_members: several tables and no declaration of which is meant
zip.no_tabular_member: a zip with nothing we read inside it
zip.too_large_uncompressed: the archive expands past the parser's bound — a COMPRESSED byte cap does not bound this

Excel workbooks

xlsx.multi_sheet_first_taken: several sheets with DIFFERENT headers; counts describe the first only
xlsx.no_worksheet: an .xlsx with no worksheet (probably .xlsb or a macro workbook)
xlsx.parallel_sheets: one table split across tabs — the subject is the SHEET NAME
xlsx.shared_strings_unreadable: text cells will be empty
xlsx.styles_unreadable: dates cannot be told from numbers

JSON structure

json.nested_not_flattened: arrays/objects below the flatten depth were not exposed

PDFs

pdf.ambiguous_tables: several DIFFERENT tables in one PDF — which one is the data is the depositor's to say, not ours to guess
pdf.encrypted: encrypted — we hold no key and will not attempt one
pdf.invisible_text: render-mode-3 text (the OCR-layer tell)
pdf.not_a_pdf: the signature is not a PDF
pdf.partial_page_coverage: the table was read from some of the PDF's pages, not all of them — a short read that must never look like a whole one
pdf.prose_not_table: running prose, not a table
pdf.rotated_text: rotated/sheared text, not projected flat
pdf.scanned: a page image, not text — OCR is a DECLARED rung we have not built
pdf.signals_contradict: ruled lines and whitespace disagree about the columns
pdf.too_few_lines: fewer lines than a header plus a record
pdf.unmappable_glyphs: runs whose font encoding we could not map, dropped rather than emitted

Darwin Core Archives you upload

dwca.core_is_not_occurrence: the declared core is an EVENT (or other) table — the species are in an extension
dwca.core_read_extensions_dropped: a Darwin Core Archive: we read its declared core and set the extensions aside

Finding the header row

header.duplicate: the same header appears more than once
header.not_full_width: no row is wide enough to be the header
header.preamble_rows: title/blank rows above the real header

Individual rows

row.field_unparsed: a single row's value did not parse for its term — a ROW event, not a column one
row.not_landable: the row lacks the minimum (coordinate + date) to be placeable
row.ragged: rows disagree with the header on column count

Matching your columns to Darwin Core

term.ambiguous_header: several columns token-match and none matches exactly

Values that contradict their header

value.atomised_date_conflict: year/month/day contradicts eventDate — GBIF flags the record twice and trusts neither
value.column_empty: the column exists but holds no values at all — ABSENT, which is not the same as invalid and must never be reported as one
value.fused_terms: one cell holding several Darwin Core terms at once — REFUSED, never parsed, because separating them means inventing which number belongs to which description
value.insufficient: too few values to decide — NOT evidence of absence
value.not_count: a count column that is not non-negative numeric
value.not_date: a date column nothing in which parsed
value.not_in_vocabulary: values outside the term's controlled vocabulary
value.not_integer: an integer-typed DwC term carrying a non-integer — GBIF drops the field rather than rounding it
value.numeric_not_prose: a prose term whose values are entirely numeric
value.out_of_range: the column's values contradict the term its header names
value.refused_other: a value check refused it and the reason is in the message
value.uncorroborated_count: a token-only `count` match whose header carries an unexplained modifier; the values cannot corroborate either reading
value.wrong_cardinality: an identifier whose repetition contradicts the term
value.zero_coordinate: a 0,0 coordinate pair — a missing-value sentinel emitted as a position

Species names

taxon.authorship_veto: a taxon column carrying nomenclatural authorship lost to a cleaner one
taxon.rank_refinement_applied: two taxon columns; the finer one was chosen

Grid references

gridref.malformed: not a British National Grid reference at all — 100 km letters outside the grid, the letter I (which the grid does not use), or an odd number of digits (a ref splits evenly into easting and northing)
gridref.recognised_not_decoded: a coordinate system we DO decode whose values did not decode — almost always a projected CRS in metres, not degrees
gridref.unsupported_form: a well-formed British grid reference in a form we deliberately do not decode — a DINTY tetrad (SK98A), a quadrant (SK98NW), or more than 10 digits. REFUSED rather than widened to the enclosing square, which would discard the precision the recorder wrote

Packed coordinates

packed.unparseable: a packed coordinate column nothing in which parsed as a pair

Controlled vocabularies

vocabulary.source_unavailable: a SOURCED vocabulary artifact we depend on was not available, so the mouth ran without it — this is OUR degradation, not a defect in your file, and nothing you can fix

Answers you gave us

choice.unknown_header: the depositor answered a disambiguation with a column the file lacks

Publication metadata

eml.metadata_incomplete: the archive cannot be PUBLISHED until facts only the depositor has (creator, description, licence) are supplied

Building the archive

archive.metadata_not_an_object: depositor metadata arrived as something other than a JSON object
archive.metadata_unparseable: the depositor metadata was not valid JSON — REFUSED rather than ignored, so our parse failure cannot read as their gap
archive.no_data_from_file: the file was read and NOTHING in it reached the archive — every emitted value would be one we minted or defaulted ourselves
archive.too_many_rows: above §B.3's automated ceiling — this size is QUOTED, and slicing it silently would hand back a partial archive that looks complete

Payment

payment.required: the archive for this submission has not been paid for (§B.3)

Upload limits

upload.refused: refused before parsing, by the abuse budget

Uncoded — a defect on our side

UNCODED: ⚠ emitted without a code — a site that has not been migrated; this is a DEFECT