Large‐scale and long‐term wildlife research and monitoring using camera traps: a continental synthesis
Bibliographic record
Abstract
Camera traps are widely used in wildlife research and monitoring, so it is imperative to understand their strengths, limitations, and potential for increasing impact. We investigated a decade of use of wildlife cameras (2012-2022) with a case study on Australian terrestrial vertebrates using a multifaceted approach. We (i) synthesised information from a literature review; (ii) conducted an online questionnaire of 132 professionals; (iii) hosted an in-person workshop of 28 leading experts representing academia, non-governmental organisations (NGOs), and government; and (iv) mapped camera trap usage based on all sources. We predicted that the last decade would have shown: (i) exponentially increasing sampling effort, a continuation of camera usage trends up to 2012; (ii) analytics to have shifted from naive presence/absence and capture rates towards hierarchical modelling that accounts for imperfect detection, thereby improving the quality of outputs and inferences on occupancy, abundance, and density; and (iii) broader research scales in terms of multi-species, multi-site and multi-year studies. However, the results showed that the sampling effort has reached a plateau, with publication rates increasing only modestly. Users reported reaching a saturation point in terms of images that could be processed by humans and time for complex analyses and academic writing. There were strong taxonomic and geographic biases towards medium-large mammals (>500 g) in forests along Australia's southeastern coastlines, reflecting proximity to major cities. Regarding analytical choices, bias-prone indices still accounted for ~50% of outputs and this was consistent across user groups. Multi-species, multi-site and multiple-year studies were rare, largely driven by hesitancy around collaboration and data sharing. There is no widely used repository for wildlife camera images and the Atlas of Living Australia (ALA) is the dominant repository for sharing tabular occurrence records. However, the ALA is presence-only and thus is unsuitable for creating detection histories with absences, inhibiting hierarchical modelling. Workshop discussions identified a pressing need for collaboration to enhance the efficiency, quality and scale of research and management outcomes, leading to the proposal of a Wildlife Observatory of Australia (WildObs). To encourage data standards and sharing, WildObs should (i) promote a metadata collection app; (ii) create a tagged image repository to facilitate artificial intelligence/machine learning (AI/ML) computer vision research in this space; (iii) address the image identification bottleneck via the use of AI/ML-powered image-processing platforms; (iv) create data commons for detection histories that are suitable for hierarchical modelling; and (v) provide capacity building and tools for hierarchical modelling. Our review highlights that while Australia's investments in monitoring biodiversity with cameras position it to be a global leader in this context, realising that potential requires a paradigm shift towards best practices for collecting, curating, sharing and analysing 'Big Data'. Our findings and framework have broad applicability outside Australia to enhance camera usage to meet conservation and management objectives ranging from local to global scales. This review articulates a country/continental observatory approach that is also suitable for international collaborative wildlife research networks.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.019 | 0.049 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.003 |
| Bibliometrics | 0.017 | 0.020 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.006 | 0.004 |
| Open science | 0.001 | 0.004 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.007 | 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 source (direct Gemma or distilled Codex), 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".