To translate Darwin Core occurrence notes quickly, separate observational prose from the data that identifies the biological record. Preserve occurrenceID, event dates, scientific names, classification identifiers, coordinates, controlled terms and the relationship between an observation and its sampling event. Translate explanatory notes only within an approved language-aware presentation or publication workflow, then check that the translated description still supports the original observation rather than inventing a new sighting or changing what an observer actually saw.
This article addresses occurrence-level documentation in biodiversity datasets, not the general problem of translating species names. A research volunteer records a bird or tree, writes a short field note, and later shares the record through a Darwin Core data pipeline. The note is written in one language, while readers may prefer another. The translator’s responsibility is to communicate the note accurately while respecting the observation’s identity, uncertainty, evidence quality and sensitive-location rules. That boundary matters more than the fluency of a single sentence.
Darwin Core is maintained through Biodiversity Information Standards, also known as TDWG. The official list of terms had a version issued on 13 October 2026; individual terms have their own definitions and version histories. GBIF provides distinct data-publishing requirements and recommendations for occurrence datasets. These sources are complementary but not identical. A translation workflow must preserve the precise vocabulary used by its publisher and recognize that an illustrative CSV file is not, by itself, a validated Darwin Core Archive or a guarantee of GBIF acceptance.
Begin with an evidence record, not a paragraph
Imagine a fictional school biodiversity survey beside a pond. A student reports a bird, takes a photograph, notes that a leaf partly obscured the view, and records an approximate time. Later, the school prepares bilingual descriptions for a public learning site. It is tempting to translate the entire row because most of its cells look like words. That approach risks modifying the observation ID, the observer’s original wording, the date format or the taxonomic identification as if they were interchangeable phrases.
The first task is to establish what each cell represents. An occurrence identifier is a persistent connection to evidence and version history. A scientific name is a taxonomic assertion, not a phrase to rewrite for style. An occurrence remark is a note about the occurrence, while an event remark concerns the broader collecting or observation event. A human-readable common name may have culturally appropriate alternatives, but it is not a substitute for a scientific taxon record. Map these roles before assigning strings to translators.
Write a publication acceptance statement that a field biologist can understand: a reader of the translated note should be able to reconstruct what the observer reported, what remains uncertain, and which record supplies the supporting evidence. The translation must not imply a more precise identification, larger population, different location or stronger certainty than the source. If the original note is ambiguous, the correct action is to preserve or resolve that ambiguity with the evidence owner, not to conceal it under polished educational language.
Distinguish an occurrence from a taxon and event
Darwin Core can describe organisms, occurrences, events, locations and identifications as related concepts. A taxon entry describes a classification concept or name; an occurrence gives evidence of an organism at a particular place or time. A sampling event describes the observation or collection activity that produced evidence. Translation workflows that collapse these categories can turn a single survey visit into several apparent records or attach notes to the wrong species. The distinctions are conceptual, not merely the spelling of column headers.
For a school field trip, the event may be a pond survey conducted on a specific day. Several occurrence records can refer to that event: a flowering plant, a bird and an insect. Their occurrence remarks should not all inherit a remark that applies only to the survey weather. Keep eventID links stable when the dataset uses them. Conversely, do not move a species-specific injury observation into the event note simply because translators want fewer separate fields to handle.
The published dataset may have one row per occurrence or use an Event core with related Occurrence extensions. Its actual archive structure determines where information belongs. Review the archive mapping and identifier links before translation. If an event record has a parent event, keep that hierarchy intact. A translated public summary can combine information from multiple fields for reading convenience, but the underlying evidence should remain separately retrievable and auditable so future researchers know what came from which source.
Protect occurrenceID through all language versions
The identifier occurrenceID lets an occurrence be recognized across dataset versions and downstream reuse. A translation that creates a new record with a new identifier for each language can make one sighting appear to be multiple independent observations. A translation that changes punctuation inside an ID may break joins to photographs, citations or annotations. Treat the ID as an exact machine value and keep its original spelling, case and separators unless the dataset steward has explicitly corrected an identifier error.
Suppose the source record uses a fictional identifier urn:school-survey:2026:pond:017. Its Malay and Chinese descriptions should point back to that same occurrence. If the publication needs separate display objects for each language, those display objects can have their own presentation identifiers while preserving the original occurrence reference. Do not introduce multiple biological occurrence records solely to hold translated strings. A display-language change is not a new ecological encounter.
Test identifier stability with an exact comparison before and after the translation pipeline. A row-count comparison alone cannot detect a split if a translator has accidentally duplicated one record and omitted another. Build a mapping by immutable identifier, compare its source and release frequency, and inspect any unexpected many-to-one or one-to-many relationships. Record what changed at the language layer so that a downstream researcher can distinguish a revised description from an amended biological observation.
Handle scientific names as controlled evidence
A scientific name such as a binomial is not ordinary English prose even though its letters use a Latin script. It conveys a taxonomic hypothesis or classification and often carries authorship information. Do not translate it into an informal common name inside scientificName, remove an authority for readability, or substitute an accepted synonym without a taxonomic review. Those actions alter the identification claim, which must be governed separately from language editing.
Sometimes a field record contains an uncertain identification or a qualifier. The appropriate Darwin Core terms and guidance should represent that uncertainty rather than hiding it by changing the spelling of scientificName. A translator can explain the uncertainty in a public note, provided that the source record and permitted editorial representation agree. A confident target statement such as “This was definitely species X” is unacceptable when the observer recorded only a possible identification. The scientific limit is part of the meaning.
Test a species lookup separately from a translation. Verify that the record still resolves to its original taxon reference and that a localized common name has not been written back into a machine field. If the team discovers a genuine taxonomic correction, open a data-curation change with its own date and evidence. Keep the translation revision log distinct. Otherwise a future user will be unable to tell whether the organism was reidentified or the description simply became available in another language.
Give vernacular names the right status
Vernacular or common names can vary by language, region and community, and a single taxon may have more than one widely used name. These names are valuable for learners but can be misleading if presented as exact equivalents of a scientific identification. A target-language name should be selected from an appropriate authoritative or locally reviewed resource, and the interface should make the scientific name available where it helps resolve ambiguity.
A field note saying “large tree beside the drain” is not a taxon diagnosis. Translating it as the name of a particular species would replace an observational description with an identification. Likewise, calling every small bird a sparrow because the local word sounds familiar can introduce a false classification. Keep descriptive nouns and taxonomic assertions separate. A glossary can state when a local name is broad, colloquial or contested without claiming there is one universal translation.
When records support more than one audience, consider separate display labels with language tags in an application layer or another explicitly documented schema. A simple Darwin Core CSV column called vernacularName does not automatically offer a language map or provide language metadata for each distinct audience. Do not append invented language suffixes to standard Darwin Core term names and assume aggregators will understand them. Document extensions and check the consuming platform’s rules before production delivery.
Read field notes at the level of actual observation
An occurrence remark can preserve information that does not fit a controlled field: damaged feathers, unusual movement, partial obstruction, or a note that an individual was seen only briefly. Such remarks often carry the limits of observation. A translator should retain qualifying words such as “possibly,” “appeared,” “approximately” and “not confirmed.” Removing them may make the sentence smoother while substantially overstating the evidence.
Consider the source note “Possibly two juveniles; partly hidden behind reeds.” A literal target that preserves the uncertainty might feel less elegant than “Two young birds behind the reeds,” but the shortened version erases both the possible count and the limited visibility. The repair is to find natural uncertainty markers in the target language, not to make the data more definite. If the language needs a longer expression, provide display space rather than deleting scientific caution.
Make an evidence ledger with fields for source expression, intended observation, uncertainty markers, numeric quantities, translator choice and verifier notes. Pair the translated sentence with its photograph or original field notebook where permission allows. A reviewer should be able to point to evidence supporting the wording. Do not demand that translators infer details from a tiny image when no qualified observer has validated it; an unclear photograph is a reason to escalate the identification, not to invent a confident translation.
Preserve the source’s original recorded wording
The term verbatimLocality is especially important because it captures locality information as originally described. An original label may use a historical place name, a collector’s spelling or a phrase that is meaningful only in its time. Replacing that field with an updated translation erases evidence about what the observer actually wrote. A modern explanatory location label can be provided separately, but the archival source should remain available and unchanged.
The distinction also applies to dates, source names and handwritten transcription. A translated display may explain that a former place name refers to a modern district, but it should mark the explanation as editorial context rather than silently editing the source. If a transcription error has been confirmed, data stewardship may allow a corrected interpretation alongside the original. That is a different workflow from ordinary translation and should have a traceable note explaining the change.
For museum specimens, old locality names may be historically useful even when no longer accepted in modern maps. Translators should avoid imposing an assumed contemporary jurisdiction that could be wrong. Consult curators, geographic gazetteers and original labels when available. The publication can explain uncertainty plainly: “Locality recorded as X; current location not established.” That phrase is more valuable than an unjustified precise map pin, because future researchers can revisit the underlying evidence.
Treat coordinates and geographic uncertainty as measurements
Decimal latitude and longitude identify a position in a coordinate system. They are not numbers to localize by changing their mathematical values or swapping them for the formatting preferences of a reader’s language. Changing a decimal separator inside a machine-readable CSV can turn one coordinate into an invalid value or split it across columns. A translator may help produce a localized display, but the source numeric fields and their interpretation must remain stable.
Check geodetic datum, coordinate uncertainty and any information indicating that a place has been generalized. A public location may represent a broad area rather than the exact site of a rare organism. If a translator turns “generalized to district level” into “recorded at this exact point,” the translation invites false precision and may undermine conservation safeguards. The limits of geographic evidence are as important as the position itself.
Use synthetic coordinates in tests rather than exposing protected collection sites. Compare the numeric fields before and after translation and verify the display against the intended map projection. A plausible-looking map pin is not an integrity test: the coordinates may have been reversed or rounded excessively. Record whether the target sentence describes an observed place, an approximate area or a deliberately generalized location. Those are different kinds of geographical claims.
Respect sensitive species and withheld information
Some occurrence records contain location or collector information that should not be disclosed at full precision. Darwin Core includes terms such as informationWithheld and dataGeneralizations that can communicate withheld or generalized information. Their presence must be interpreted according to the publisher’s release policy. Translating a public explanation does not authorize access to hidden coordinates or permission to reconstruct a more detailed locality.
Suppose the public record reports only that an observation occurred in a broad protected-area region. A translator should not add a landmark from an internal volunteer message, even if it would make the narrative more engaging. A school may wish to celebrate a rare plant discovery, but publication has to balance education with risk to the organism and habitat. The source record’s disclosure rules should apply consistently across every language and promotional channel.
Test the same record through all language views, downloadable exports, previews and search snippets. A site that hides precise coordinates in one translated page but exposes them in another language’s alt text has not protected the information. Review maps, image metadata and download attachments as part of the release surface. Data protection is a system requirement; a translator should receive the approved public facts and an explicit boundary rather than being asked to decide conservation policy alone.
Verify event dates and temporal precision
An observation date communicates when evidence was collected, not when its translation was made. Keep eventDate stable unless a data curator has confirmed a correction. An uncertain historical date may be represented as a year or an interval rather than a fabricated day. Do not fill unknown month or day components with convenient zeros or assume an approximate date is exact. Publishing requirements and date formats must be checked against the dataset’s actual data profile.
A volunteer might record “late afternoon” in a field notebook while the structured event date contains only a calendar day. The translation should not invent a time zone or minute precision. If a source narrative mentions “after sunset,” that contextual phrase can be translated, but it remains a qualitative observation unless supported by specific measurement. Avoid allowing a polished multilingual dashboard to blur the difference between measured time, remembered approximate time and the database’s date granularity.
Create regression cases with a complete date, a month-only historical entry, a date interval, and a narrative referring to a season. Compare original and published structured values byte-for-byte where no transformation is intended. The user-facing translation may adjust date display order to match locale convention, but must retain the same underlying interval. If an event spans midnight, make sure the source meaning does not shrink to the first day merely because the target text is shorter.
Preserve detection and non-detection accurately
The term occurrenceStatus concerns detection or non-detection during an event. A record indicating absence is not the same as an observation where the organism was simply not photographed. Translators must know whether a note describes an observed organism, a targeted survey with no detection, or a missing data value. Words such as “not found,” “not present” and “not recorded” may imply different scientific claims depending on sampling design.
Suppose a standardized sampling event examined a defined area for an invasive plant and recorded no detection. The translation should describe that survey outcome, not assert that the species cannot exist anywhere nearby. Conversely, if an observer forgot to enter a species count, the language must not convert the blank field into zero. This distinction affects ecological inference: confirmed non-detection under a protocol is evidence with a different scope from an uncompleted or unknown observation.
Review quantity fields such as individualCount or the paired organismQuantity and organismQuantityType as data, not narrative decoration. A unit may represent individual organisms, cover percentage or another quantity type. A translator can explain it in natural language but should never translate a measurement into an unrelated unit or assume every numeric value counts animals. Check the pairing, controlled vocabulary and survey protocol before producing a summary that will be used in lessons or reports.
Review sampling effort before interpreting abundance
A large number of observations does not automatically indicate a larger population. A volunteer who spent three hours watching one site may record more birds than another who watched for five minutes. The event record may describe sampling protocol, size or effort; translation should preserve those distinctions. Changing “three observation hours” to “three observations” corrupts the relationship between effort and result and can mislead learners about how ecological data works.
When describing count comparisons, include the relevant denominator or methodological limitation if it is necessary for understanding. A public summary could say that five individuals were noted during a defined transect rather than announcing that the reserve contains only five individuals. An accurate translation may require more words than the English sentence to express this difference. Allocate space accordingly and ask a field scientist to verify the summary if it will support educational conclusions.
Keep scientific inference separate from field note translation. The raw observation “four flowers open” may support a later research analysis, but it is not itself a conclusion about pollination success or seasonal trends. A translator should not upgrade it into “the plant reproduced successfully.” A lesson author may add an explicitly labeled inference with supporting evidence, but that creates new explanatory content subject to review. The translation manuscript should preserve the original evidence rather than silently append research conclusions.
Deal with observer names and personal details responsibly
Names of collectors and observers can be part of provenance, yet public presentation may be subject to privacy or contributor agreements. Do not translate proper names into unrelated local names, reverse identity fields, or expose withheld contact details in a localized caption. If the platform offers a preferred display form, use it consistently according to the steward’s rules. The purpose of provenance is traceability, not the unnecessary disclosure of personal information.
A volunteer note might mention a colleague’s phone number or an exact home address. Such details are not automatically appropriate for public export just because they appear in the original. The data steward should decide what is redacted, retained privately, or summarized, and that decision should be implemented before translation. Translators should work on a controlled public subset and flag unexpected personal data. This reduces exposure while retaining enough context to explain the observation honestly.
When an observer’s testimony is quoted, distinguish direct quotation from an editorial summary. Preserve quotation boundaries, uncertainty and attribution, especially when the note involves a disputed identification. A translator’s clarification should not be presented as something the observer said. Keep a record linking the translated quotation to its source, including whether the source was handwritten, dictated or entered digitally. That provenance makes the educational material more trustworthy and easier to correct later.
Prepare a safe multilingual data model
A common mistake is to duplicate a whole Darwin Core dataset for each language and then let the copies drift. A safer educational system keeps the canonical occurrence data unchanged and maintains translations in a separate presentation layer keyed by occurrence identifier, field, locale and source revision. This is an editorial architecture, not an automatic Darwin Core extension standard. If the destination aggregator supports a specified multilingual mechanism, follow that contract rather than inventing field names.
For example, a display table can store an approved Malay occurrenceRemarks translation and its English source revision while preserving the original occurrence row. Each translated record should include the reviewer, date, locale and approval status. If a scientific correction changes the occurrence note, a flag can mark the translation as needing review. The reverse is also true: changing a target-language phrase must not overwrite the original evidence, date, ID or coordinates.
A text export for editors should explicitly label which columns are writable. Protect identifiers and taxonomic fields in the editing tool. Use validation to reject edited numeric coordinates, blank source references or unknown locale codes. Do not trust a spreadsheet’s column position after someone inserts a new field; map by declared header names. At reintegration, compare hashes or stable field values and generate a concise diff for a data curator before any publication decision.
Work through a small illustrative CSV record
This fictional occurrence row shows why field roles matter. The coordinates and date are invented for testing; they do not claim that a named species was actually observed at that place. The note is suitable for educational translation while the identifier and taxonomic string belong to the evidence layer. A CSV parser can check column structure, but biological accuracy and an approved Darwin Core Archive require further review.
occurrenceID,basisOfRecord,scientificName,eventDate,decimalLatitude,decimalLongitude,occurrenceRemarks
urn:school:survey:2026:017,HumanObservation,Ficus microcarpa,2026-07-11,1.3000,103.8000,"Leaves wet after rain; fruit not confirmed."
The translator may produce a Malay explanation of the final note such as “Daun basah selepas hujan; buah belum dapat disahkan.” The original record should remain the same, while a language-aware display layer points to it. The phrase must not become “fruit present” or “fruit absent,” because the observer explicitly did not confirm that fact. Keep the species name as recorded and avoid treating the synthetic coordinates as an invitation to publish a real collection location.
Test a second record where no observation was made, another with a damaged specimen and one whose locality is intentionally generalized. The same translation process should handle uncertainty and privacy consistently. If the display renders only the translated field note but omits its link to the occurrence ID, readers may confuse separate events. Provide a source reference in the interface and consider showing the original text for expert verification where licensing and privacy permit.
Diagnose a false certainty introduced by translation
A bilingual reviewer notices that “possible nest site near the path” became “confirmed nesting site at the path.” This is not a minor stylistic issue: it changes uncertainty, location and biological interpretation. Find the source occurrence note, read the relevant event context, and verify whether a later curator corrected the claim. If the source still says “possible,” the target must restore that uncertainty. The best repair uses natural target wording that does not sound more certain than the evidence.
Next check where the incorrect claim entered the workflow. It may have been a machine translation suggestion, an editor’s attempt to simplify for children, or an outdated display cache. Correct the translation record, regenerate affected previews and investigate whether public summaries or search snippets reproduced the false claim. Do not quietly alter the canonical source to match an erroneous translation. Record a correction note and a test fixture to keep the mistake from returning.
Finally ask whether the project should publish a clearer educational explanation of uncertainty. Such an explanation might teach the difference between an observed nest and a location where nesting was suspected. If added, it must be labeled as a learning note rather than substituted for the observer’s data. This turns a QA problem into an opportunity to teach good scientific reasoning without rewriting historical evidence.
Diagnose an unexpected change in geographical display
A translated interface suddenly places a record far from its intended region. Before blaming the map library, compare coordinate fields to the preserved source row. Look for a decimal comma substituted into CSV values, latitude and longitude reversed, or values rounded beyond their original precision. Inspect any display-only formatter and ensure that it has not overwritten machine coordinates. An attractive map layer may hide a serious data integrity defect.
If a private collection site is involved, verify the public generalization policy before attempting repair. The correct public result may intentionally show a broad area rather than the precise place. A translator should not restore a detailed pin from an internal source because it seems more accurate. Technical correctness, scientific uncertainty and conservation disclosure are different constraints that must all be satisfied. Repair the appropriate layer and have the data steward approve the final display.
Use a deterministic fixture with known synthetic coordinates and a documented expected display region. Test it in each interface language, export format and map view. Preserve coordinate precision as required by the dataset, and report whether the issue arose in data, serialization, mapping or translation. A localized note may be fixed in minutes, but a coordinate regression can invalidate comparisons and should trigger a broader data-quality check.
Diagnose a species-name collision
A target-language common name maps to more than one scientific taxon, and a school article suddenly describes two records as the same species. The correct response is not to choose the more familiar common name arbitrarily. Trace each occurrence through its scientific name and taxonomic reference, then examine the common-name sources for the relevant region. Keep the original classification unchanged while the display label is corrected or qualified.
Where a common name is ambiguous, show it alongside a scientific name or a carefully worded descriptor rather than implying certainty the source does not contain. The educational goal is for students to understand why common names are helpful but imperfect labels. A glossary can list multiple recognized names with context. It should not overwrite canonical identifiers or create a false claim that different species are taxonomically identical simply because their everyday names overlap.
Add regression cases for the disputed names and review other records sharing the same display term. Check search results and filters: an incorrect synonym assignment may route a learner to the wrong species page. If the issue is a genuine taxonomic revision, open the separate scientific-curation process. A translation correction should stay scoped to language, while classification changes require evidential authority and version tracking.
Validate a full round trip rather than only the file
Run the source record through export, translation management, reintegration, publication preview and download. At each stage compare the immutable identifiers, dates, coordinates and controlled values. Check that comments and quotations survive CSV quoting and character encoding. A field note containing a comma, quotation mark or non-Latin script can be truncated or shifted into the next column if the pipeline relies on manual string splitting instead of a CSV parser.
Confirm the released archive is mapped to the correct standard fields and that references resolve. A successful CSV parse establishes only syntax; it does not prove that basisOfRecord uses an accepted vocabulary or that an occurrence has a valid scientific identification. Use the relevant publisher’s validation tools and interpret their results with subject specialists. Document warnings rather than hiding them under a broad “translated successfully” flag.
For human review, compare a source view and target view side by side with the same occurrence identifier prominent. Ask a biologist to verify scientific substance and a language reviewer to verify clarity and register. For sensitive sites, use a restricted review workflow. This dual check makes translation faster because reviewers know which claims they are responsible for and can identify exactly whether a defect belongs to language, data mapping, biology or privacy.
Transfer exercise: historical specimen labels in a museum
A museum digitizes a century-old herbarium sheet. The handwritten label lists a former place name, an imprecise date and an uncertain plant identification. A translator is asked to prepare a modern Chinese learning caption while a software developer wants the original source fields rewritten so that the public archive looks tidy. Decide which elements must remain preserved, which can be explained separately and how the final description should communicate uncertainty.
The worked answer is to retain the original locality text and date precision in the appropriate evidence fields. Do not replace scientificName with an unreviewed modern taxon or silently remove an identification qualifier. Add an explicitly editorial caption that explains the historical locality if an expert has verified the relationship, and preserve uncertainty about the collection date. Link the translated display to the same occurrence or specimen identifier and keep any withheld collection details protected. A modern caption is a guide to evidence, not a new specimen record.
Test the solution by changing the audience to a research scientist rather than a primary-school class. The scientist must still be able to retrieve the original label, the curator’s interpretation and the translated explanation as distinguishable layers. If the public caption presents a precise date or location that the archival source never supplied, the transfer fails. This exercise shows why translation QA benefits from preserving evidence and interpretation separately, even when an elegant single sentence would be easier to publish.
Practical checklist for occurrence translation review
Confirm the record type and archive mapping, stable occurrence and event IDs, scientific names, controlled vocabulary values, date precision and any coordinates. Identify which notes are observational prose, which are original verbatim evidence, and which are editorial interpretation. Preserve uncertainty markers, quantities, negation and source attribution. Verify that no translation has added a sighting, expanded a habitat claim or inferred a species from a common name. Each check should be traceable to an individual record and source revision.
Next inspect public visibility: sensitive localities, generalizations, personal details, image metadata, downloadable rows and search snippets. Run CSV and archive validation as appropriate, then compare the original and translated display with a subject specialist. Test that every language view points to the same underlying occurrence and that an update to the source invalidates stale translations. Record unresolved questions and the responsible curator rather than letting them disappear inside a blanket editorial approval.
Limits and responsible publication
Darwin Core terms define a vocabulary and conceptual data model, not a universal translation mechanism for every field in every platform. A localized presentation table may be a reasonable editorial design, but a production publisher must follow its archive, aggregator and licensing contracts. This guide does not validate biological identifications, authorize disclosure of sensitive species, or replace curatorial expertise. Dataset quality expectations can also differ by dataset class, programme and publishing workflow.
The article illustrates how to protect occurrence evidence while making observations comprehensible across languages. It is not a guarantee that a particular CSV or translated label will be accepted by GBIF or another aggregator. For actual release, consult the latest official TDWG term definitions and the destination publisher’s current validation guidance, then test the final artifact. A reliable process keeps translation flexible while protecting the identity, time, place and uncertainty of the observation.
Summary: language should illuminate the sighting
A biodiversity translation is successful when a new reader understands the same field evidence without receiving stronger or different scientific claims. Protect identifiers, taxonomy, dates and measurements. Treat observational notes as careful testimony, keep uncertainty intact, and leave original verbatim fields available. Use an explicit language layer rather than duplicating biological records, and validate the entire publication route. Students and researchers can then learn from the translated note while remaining connected to the original observation.
Frequently asked questions
Should scientificName be translated into a local common name?
No. Keep the recorded scientific identification intact. Add a reviewed vernacular display label separately when it helps readers, and explain any ambiguity.
Can translated occurrence notes share the same occurrenceID?
Yes. They should refer to the same underlying occurrence. Separate translations do not constitute additional biological observations.
Why not overwrite verbatimLocality with a clearer modern place name?
The verbatim field preserves the original locality statement. A modern interpretation can be presented separately, with its evidence and uncertainty.
Is an approximate sighting the same as a confirmed identification?
No. Preserve hedges such as “possible,” “appeared” and “not confirmed” in the target language so the level of evidence does not change.
Can zero individualCount always mean the observer forgot to enter a count?
No. Zero, missing values and non-detection have different meanings. Interpret them with the survey protocol and occurrence status, not translation assumptions.
What if the public location is deliberately generalized?
Preserve the steward’s privacy or conservation policy in every language. Do not restore precise sites from internal notes or image metadata.
Does an ordinary CSV parse prove the dataset meets Darwin Core requirements?
No. It checks structure, not scientific correctness, accepted vocabulary usage, archive mappings or aggregator-specific publishing expectations.
How should common names that vary between regions be handled?
Use a reviewed terminology source, note regional variants when useful, and keep the scientific identification available for disambiguation.
Who should approve scientific uncertainty in a translated note?
A qualified data steward or subject specialist should approve the underlying claim; a language reviewer should check that the target wording faithfully expresses it.
When does a translation require a new biological record?
Translation alone does not. A genuinely new independent observation may require its own record, but that decision belongs to biological data curation, not language editing.
Authoritative sources and verification scope
- TDWG Darwin Core home and guides — conceptual model, vocabulary and implementation, inspected 28 October 2026.
- TDWG Darwin Core core terms — official term definitions and version metadata, latest list issued 13 October 2026.
- TDWG Darwin Core list of terms — current multi-namespace list dated 13 October 2026.
- GBIF technical data-quality recommendations — occurrence dataset requirements and recommended data.
- GBIF IPT Data Quality Checklist — expected links and value checks for occurrence/event publication.
- GBIF biological survey publishing guide — sampling-event and occurrence relationships, consulted for conceptual examples.