Search

Camera trap arrays can generate thousands to millions of images that require exorbitant time and effort to classify and annotate by trained observers. Computer vision has evolved as an automated alternative to manual classification. The most popular computer vision solution is the supervised Machine Learning technique, which uses labeled images to train automated classification algorithms.We propose a multi-step semi-automated workflow that consists of (1) identifying and separating bad-from good-quality images, (2) parsing good images into animals, humans, vehicles, and empty, and (3) cropping animals from images and classifying them into species for manual inspection. We trained, validated, and evaluated this approach using 548,627 images from 46 cameras in two regions of the Arctic (northeastern Norway, and Yamal Peninsula, Russia).We obtained an accuracy of 0.959 for all three steps combined with the complete year test data set at Varanger and 0.922 at Yamal, reducing the number of images that required manual inspection to 7.9% of the original set from Varanger and 3.2% from Yamal.Researchers can modify this multi-step process to meet their specific needs for monitoring and surveying wildlife, providing greater flexibility than current options available for image classification.

Camera traps have become popular for monitoring biodiversity and animal populations. Artificial intelligence is increasingly used to automatically classify large image data sets produced by camera traps and many tools that incorporate machine-learning models for automatic image classification have been developed over the last years. However, it is still challenging to combine tools for automatic classification with other tools for processing camera trap images and to adapt these tools to a specific study. Therefore, we propose a semi-automatic workflow for processing camera trap images in R. The workflow includes managing raw images, automatic image classification, a quality check of automatic image labels as well as the possibilities to retrain the model with new images and to manually review subsets of images to correct image labels. We illustrate the workflow with a case-study from the small mammal monitoring program of the Climate-ecological Observatory for Arctic Tundra. We first trained a classification model for small mammals and then transferred the model to new images, including images from newly established camera traps. We could show that retraining the original model with a small number of new images increased model performance and therefore highlight the importance of verifying automatic image labels when a model was transferred to new images. Furthermore, retraining the original model also decreased the time needed for manually reviewing images and correcting image labels substantially. Thus, the proposed workflow results in a data set with high accuracy and minimizes time needed for labeling images manually. This is especially useful for long-term monitoring where new images have to be processed continuously and methods have to be adapted over time. We provide all R scripts and the classification model for small mammals to make the workflow accessible to other ecologists.

Camera traps have become popular labor-efficient and non-invasive tools to study animal populations. The use of camera trap methods has largely focused on large animals and/or animals with identifiable features, with less attention being paid to small mammals, including rodents. Here we investigate the suitability of camera-trap-based abundance indices to monitor population dynamics in two species of voles with key functions in boreal and Arctic ecosystems, known for their high-amplitude population cycles. The targeted species—gray-sided vole (Myodes rufocanus) and tundra vole (Microtus oeconomus)—differ with respect to habitat use and spatial-social organization, which allow us to assess whether such species traits influence the accuracy of the abundance indices. For both species, multiple live-trapping grids yielding capture-mark-recapture (CMR) abundance estimates were matched with single tunnel-based camera traps (CT) continuously recording passing animals. The sampling encompassed 3 years with contrasting abundances and phases of the population cycles. We used linear regressions to calibrate CT indices, based on species-specific photo counts over different time windows, as a function of CMR-abundance estimates. We then performed inverse regression to predict CMR abundances from CT indices and assess prediction accuracy. We found that CT indices (for windows maximizing goodness-of-fit of the calibration models) predicted adequately the CMR-based estimates for the gray-sided vole, but performed poorly for the tundra vole. However, spatially aggregating CT indices over nearby camera traps enabled reliable abundance indices also for the tundra vole. Such species differences imply that the design of camera trap studies of rodent population dynamics should be adapted to the species in focus, and adequate spatial replication must be considered. Overall, tunnel-based camera traps yield much more temporally resolved abundance metrics than alternative methods, with a large potential for revealing new aspects of the multi-annual population cycles of voles and other small mammal species they interact with.

The extent to which persistent organic pollutants (POPs) with different physicochemical properties originated from the food (dietary input) was assessed in raptor nestlings. Lipophilic polychlorinated biphenyl (PCB) 153, 1-dichloro-2,2-bis(p-chlorophenyl)ethylene (p,p'-DDE), and hexachlorobenzene (HCB), and protein-bound perfluorooctane sulfonate (PFOS), were measured repeatedly in blood plasma of individual goshawk (Accipiter gentilis) and white-tailed eagle (Haliaeetus albicilla) nestlings, 1 to 3 wk after hatching and near fledging. Maternally derived POPs dilute as nestlings grow (growth dilution), and increasing plasma concentrations would indicate dietary input. First, plasma concentrations given no dietary input were estimated, and concentrations of p,p'-DDE, HCB, and notably PFOS were significantly higher than predicted from a growth-dilution scenario (approximately 1.5-fold to 2.5-fold higher; p 0.001). In contrast, PCB 153 declined in both species, although concentrations were still higher than predicted in white-tailed eagle nestlings (p 0.05). Second, the relationships between plasma POP concentrations and trophic position (δ(15) N) and dietary carbon source (δ(13) C) were analyzed, controlling for growth rate. Both δ(15) N and δ(13) C (measured in body feathers) were significantly associated to the accumulation of most POPs, except PFOS. In conclusion, pollutant data acquired in plasma of nestling raptors should be interpreted and further investigated in the light of individual feeding ecology, and the use of raptor nestlings as sentinels for POP monitoring could be optimized by correcting for different factors such as body condition, brood size, and age.

A mathematical model is developed and analyzed for studying the population dynamic consequences of the social system of Clethrionomys species. A patchy environment is assumed. The general version of the model incorporates density-dependent reproduction and survival; the specific version of the model for Clethrionomys assumes, in accordance with data, density independence in reproductive rate and survival of adults. Mature females are assumed to be territorial implying that maturation and dispersal of sexually immature females are regulated by the density of mature females.