A long ecological series can be numerically precise and still compare different biological meanings under the same species name.
This guide shows how reference vouchers, versioned names and targeted re-identification restore comparability without pretending every record is equally certain.
Choose a reading route
Understand identity and taxonomy
Build and run a voucher audit
Transfer and define limits
Audit and source the dataset
Chapter contents
- 1. Separate identity, name and concept
- 2. Why names alone are unstable
- 3. What counts as a reference voucher
- 4. Design a risk-based voucher plan
- 5. Run a reference-voucher audit
- 6. Worked example: the disappearing bird
- 7. Build a taxonomic crosswalk
- 8. Changed-context transfer check
- 9. Limits and misconceptions
- 10. Frequently asked questions
- 11. Practical dataset audit
- 12. Sources and reporting template
1. Separate identity, name and concept
Misidentification assigns the organism to the wrong taxon. A synonym change alters the accepted label while the taxon may stay substantially the same. A split, lump or revised circumscription changes which organisms the name includes. Each problem needs a different repair.
| Problem | What changed | Repair |
|---|---|---|
| Misidentification | Record-to-organism link | Re-examine evidence |
| Synonym | Accepted label | Map old and current names |
| Concept drift | Taxon boundary | Version concepts and reassess comparability |
2. Why names alone are unstable
A scientific name may not specify the taxonomic concept used by an observer in a particular year. Long-term monitoring should retain the verbatim name, checklist version, identifier, identification confidence and current mapped name.
Never overwrite an old determination. Append later determinations with dates and responsible people so the history remains auditable.
3. What counts as a reference voucher
A voucher is preserved evidence tied to a record and curated so identity can be checked later. It may be a specimen, photograph, recording, tissue aliquot or DNA sequence. The evidence type must capture the diagnostic characters and comply with permits and conservation ethics.
| Evidence | Useful for | Limit |
|---|---|---|
| Specimen | Morphology and later re-identification | Collection may be harmful or prohibited |
| Photograph | Visible field marks and context | Hidden characters may be absent |
| Audio | Vocal taxa | Silent individuals and overlapping calls |
| Tissue or sequence | Genetic comparison | Reference-library and contamination quality |
4. Design a risk-based voucher plan
Concentrate effort where errors matter: difficult species pairs, new range records, rare taxa, observer changes and records driving management. State what is retained, where it is deposited, the identifier used and how it links to the observation.
- Obtain permits and follow ethics.
- Use persistent catalogue identifiers.
- Capture date, place, method and identifier.
- Preserve verbatim and later determinations.
- Sample routine records as well as unusual ones.
5. Run a reference-voucher audit
Draw a stratified sample across years, sites, observers and confidence levels. Re-identify without showing the old label when feasible. Record agreement, corrected identity and unresolved status. Then test whether error is random or concentrated enough to change a trend.
Never force uncertain material into a species. “Genus only” or “unresolved pair” may be the most accurate result.
6. Worked example: the disappearing bird
A wetland series has 100 records labelled “reed marsh warbler.” Vouchers exist for 60. Re-examination finds 40 correctly identified, 12 belonging to a similar species and 8 unresolved. A later split also divides the historical concept into northern and southern lineages.
| Class | Count | Treatment |
|---|---|---|
| Confirmed target | 40 | Retain with concept version |
| Similar species | 12 | Correct and document |
| Unresolved | 8 | Exclude from species-level trend or model separately |
| No voucher | 40 | Retain as legacy uncertainty, not confirmed |
The apparent decline may partly reflect later observers separating the similar species more accurately. Reanalyse with an uncertainty band and test whether the trend survives removal of unsupported records.
7. Build a taxonomic crosswalk
Link each verbatim name and checklist version to the current accepted name, concept status and decision rule. One-to-one synonyms can often be harmonised. A split may require place, date, morphology or genetics; otherwise a record may need to remain at the broader historical concept.
Version the crosswalk. Taxonomy will change again, and future analysts need to know which mapping produced the result.
8. Changed-context transfer check
Move from preserved insects to a bird survey. Physical collection is usually inappropriate, so evidence becomes photographs and audio clips attached to a sample of detections. The logic stays: target risky identifications, preserve originals, blind a recheck and keep unresolved records visible.
A camera-trap study adds image quality, angle and classifier version to the audit trail.
9. Limits and misconceptions
Vouchers do not guarantee a correct name, and DNA does not replace taxonomy. A reference collection can contain mistakes. Some organisms cannot be collected responsibly; many legacy records lack reviewable evidence; and some concept changes prevent a single harmonised series.
| Misconception | Correction |
|---|---|
| A database accepted the name | Name matching does not verify the organism |
| A synonym is a new species | Names can change without a new concept |
| One voucher validates every record | Vouchers must represent variation and risk |
| Unresolved records are useless | They may support broader statements |
10. Frequently asked questions
How many vouchers are enough?
There is no fixed percentage. Sample across the sources of risk and increase effort where disagreement would change the conclusion.
Should names update automatically?
Only unambiguous mappings should be automated, and the original must remain. Splits and lumps need review.
Can citizen-science photographs be vouchers?
They can serve as reviewable evidence when diagnostic features, metadata and provenance are adequate.
11. Practical dataset audit
Make identity uncertainty as visible as sampling uncertainty before publishing a trend.
- Is the original name retained?
- Is the checklist or backbone version recorded?
- Can vouchers be found by persistent identifiers?
- Were rechecks stratified by observer, year and difficult taxon?
- Are corrected and unresolved records propagated into analysis?
12. Sources and reporting template
Report: “Records used [name source/version]. A stratified audit re-examined [n] of [N] with [voucher types]. Agreement was [value]; [n] were corrected and [n] unresolved. Crosswalk version [x] mapped names. The trend [did/did not] persist after unsupported records were excluded.”