Large-scale and long-term wildlife research and monitoring using camera traps: a continental synthesis

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.

Keywords: Australia; big data; biodiversity conservation; data sharing; occupancy modelling; sampling methods; terrestrial vertebrates.

Fig. 1

Temporal trends in the use of wildlife cameras in Australia from peer‐reviewed literature. (A) Year of publication (dark grey bars) and the last year of sampling (light green bars). The black line and numbers indicate the median time lag between the year sampling ended and the year the item was published. (B–D) Median effort, which remained highly variable but may have started to plateau at 75–100 cameras deployed per survey that were left for 50–65 days, thus totalling ~5000 total trap nights. All values on the y‐axis in (B–D) are log₁₀ transformed; black dots in the box plot indicate the median value for the year sampling ended; red points are outliers. Trap nights are calculated as the number of cameras multiplied by duration.

Fig. 2

Preferences and motivations for different methods used to sample wildlife. Results show responses from 132 Australian camera users who completed an online questionnaire. Sampling approaches demonstrated a strong preference for cameras regarding (A) data quality and (B) cost‐effectiveness. (C) Results of a binary logistic regression showing odds ratios where values >1 (vertical black line) indicate a positive association (more likely) and <1 a negative association (less likely). An asterisk next to the variable name indicates the overall model was significant using a chi‐squared test. A solid filled point is used where the 95% confidence intervals do not overlap 1.

Fig. 3

Only half of Australian wildlife camera analyses correct for detection probabilities. (A) Summary of published studies. (B) Responses from 132 Australian camera users who completed an online questionnaire. Methods were grouped by whether they ignored bias associated with the imperfect detection, such as in presence/absence, naïve occupancy (percentage of cameras detecting the species) or capture rates [or their modified forms, such as relative abundance index (RAI) or Allen's index]. Animal behaviour studies were primarily concerned with behaviours like carcass and nest discovery and likely used indices to understand issues such as bait uptake. (C) Results of a binary logistic regression showing odds ratios where values >1 (vertical black line) indicate a positive association (more likely) and <1 a negative association (less likely). An asterisk next to the variable name indicates the overall model was significant using a chi‐squared test. A solid filled point is used where the 95% confidence intervals do not overlap 1.

Fig. 4

Questionnaire participants indicated which taxa they sampled with cameras, summarized here with larger areas reflecting more users. Animal symbols courtesy of the NESP Resilient Landscapes Hub, nesplandscapes.edu.au.

Fig. 5

Distribution of camera surveys in Australia. (A, B) Number of camera surveys per 2500 km² hexagonal cell from published literature and questionnaire data (A, with the inset showing values per state), and potential (unconfirmed) sites from the Atlas of Living Australia (B). (C) Published (grey) and questionnaire survey locations (green), with symbol size scaled to the number of cameras reported to be deployed. (D) Published and questionnaire surveys per habitat, ordered from largest area (left) to smallest (right). Above the bars, drawings indicate the biome: trees represent largely forested biomes; grass indicates grassland or savannah; spinifex grass indicates shrub and desert biomes. The inset map shows the spatial distribution of each biome in Australia. Specific biome names from the ecoregion layer, from left to right, are: (i) deserts and xeric shrublands; (ii) tropical and subtropical grasslands; (iii) mediterranean forests; woodlands and scrub; (iv) temperate grasslands and savannahs; (v) temperate broadleaf and mixed forest; (vi) tropical and subtropical moist broadleaf; and (vii) montane grasslands and shrublands.

Fig. 6

Framework for a distributed network of wildlife camera data providers and its outputs. Data are ingested as raw images or spreadsheets, then standardised, collated, and stored permanently as a data commons, located in existing trusted data providers such as the Atlas of Living Australia (ALA) and the Terrestrial Ecosystem Research Network (TERN). ALA and TERN must add new functionality to store and share detection histories as opposed to presence‐only data. The data are then analysed using hierarchical modelling on high‐performance computing clusters (HPCC) with open‐access code hosted on GitHub. This transparent workflow scales rapidly and is designed for timely outputs for government or NGO reports or peer‐reviewed publications. Data contributors are updated with biannual newsletters and collaboration requests generated between users requesting specific access to data sets. AI, artificial intelligence.

Fig. 7

Case studies using cameras for large‐scale long‐term monitoring in Australia. (A) Locations of the 10 major ‘Eyes on Recovery’ study sites denoted by circles with the extent of areas that burnt according to the Google Earth Engine Burnt Area Map (GEEBAM) data set. Low to Moderate burning is represented in orange and High to Very High is in red. (B) Locations of the 204 WildCount monitoring sites denoted by hollow white circles distributed across 146 national parks and reserves in New South Wales; Fire Extent and Severity Mapping (FESM) where burns that were Low to Moderate are represented in orange and High to Extreme are in red (redrawn from Fig. 1 of Lavery et al. 2024).