Mobilising Legacy Georeferencing Efforts
Bibliographic record
Abstract
As we progress towards a globally accessible natural history collection, the ways in which we digitally curate, share and use our data will inevitably change. Digital specimen records enable access by in-country experts, increase the opportunity for specimen enhancement by the scientific community, and improve the breadth of research to which the specimens contribute. Digitisation workflows capture an image often with minimal transcription; they do not include the further enhancement of specimen records, such as georeferencing - the addition of coordinates to text based locality information. Typically used to georeference herbarium specimens, the point radius method assigns a coordinate and a measurement of maximum uncertainty (Wieczorek and Chapman 2020). Some herbarium specimens that lack coordinates still cannot be georeferenced using this method due to poor label data or the unavailability of contextual information. Data derived from herbarium specimens form the basis of species distribution modelling, taxonomic research and IUCN extinction risk assessments. If a specimen record does not have coordinates, then it is often discarded early in the data mining process, highlighting the importance of data enhancement through georeferencing, which confirms the species occurrence in space and time. The Kunming-Montreal Global Biodiversity Framework targets help to guide the most important applications of collection data for biodiversity conservation; Target 4, for example, aims to halt species extinction and protect genetic diversity (Convention on Biological Diversity 2023). Extinction risk assessments of plant species are underpinned by distribution maps derived from preserved specimens and observation records with coordinate information either recorded at the time of collection or through subsequent georeferencing of the specimen. However, over half of all herbarium specimens on Global Biodiversity Information Facility (GBIF) do not have coordinates and so robust extinction risk assessments often require a georeferencing step before applying the IUCN’s criteria. Other data types contributing to extinction risk assessment, such as population size and trend are often lacking and, where available, subjective and restricted to an expert’s firsthand knowledge of the species, making the data untraceable for the wider community (Nic Lughadha et al. 2019, Willis et al. 2003). As highlighted in Bloom et al. (2018), the estimated distribution of a species differs depending on the origin of the coordinate data. For example, non-georeferenced records tend to inflate the estimated distribution of a species. Working with collaborators with regional geographical expertise, and using datasets following standardised protocols, will more likely result in accurate and precise georeferenced records. As herbarium specimen labels can contain qualitative context on collection localities, the process of georeferencing is subjective and so an indicator of confidence and the georeferencer's method is important for end user trust and usability of the record. As more herbarium specimens are digitised, there is a growing need to produce georeferenced locality data at scale.. Although there are robust tools that can be used to supplement manual georeferencing, georeferencing of all specimens in natural history collections currently lacking coordinates is not feasible to resource, as georeferencing is time and resource heavy. The Royal Botanic Gardens, Kew now contributes 5.8 million herbarium specimen records to GBIF, many of which will have between three and six duplicate specimens located in herbaria around the world. Enhancing these records with existing georeferencing efforts accumulated through the completion of several thousand extinction risk assessments will reduce duplicated effort and uncover georeferenced localities that would otherwise go undocumented. This dataset will also help in establishing protocols to apply when georeferencing plant collections in the future, particularly in data-poor tropical regions.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.002 |
| Science and technology studies | 0.003 | 0.002 |
| Scholarly communication | 0.002 | 0.010 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.013 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".