Camera-trapping: wild and domestic species occurrences in three Pyrenean pastures

Study Extent

The present dataset was collected during three sampling campaigns carried out between May and October 2021, 2022 and 2023 on three mountain pastures located in the Ariège Pyrenees. A total of 118 motion-triggered infrared cameras (Reconyx® Hyperfire 2) were set on the three mountain pastures. As a result of an excessive number of thefts during the three campaigns, one camera has been removed from the Barestet pasture and data from the three years has been removed from the dataset (dataset from 118 cameras instead of the 119 initially installed in 2021). Depending on the pastures and the years, according to the field constraints, the sampling started and ended at variable dates, but at least between 12 June and 6 October. Insofar as the project is likely to continue, the present dataset may evolve, following the same protocol and supplemented by additional years of data on one or more pastures.

Sampling

A former exploratory study, carried out in 2020 on one of the three study sites, tested the implementation of camera traps following a regular grid. However, the characteristics of the mountain environment and the topographical diversity of the pastures resulted in camera traps being set up in locations that were particularly unfavourable for the passage of wildlife, pastoral activity and humans (e.g. extremely steep slopes). As a result, the sampling was conducted following a regular grid of 400 m ✕ 400 m cells, in which one camera was installed at a chosen location of potential passages of the mammalian fauna (e.g. path, trail, mountain pass, thinning of vegetation etc.), thus providing a fine spatial anchor on the activity of the species and maximising their detectability. This fine-scale grid was relevant to the study in order to have the greatest possible insight into the spatilities of the various wild and domestic species, given the technical and financial constraints involved. For the purpose of consistency, the same mesh size was used on each of the sites, resulting in different numbers of cameras for the three pastures depending on their respective size (Arreau: 53 cameras, Barestet: 36 cameras, Ourdouas: 29 cameras). Depending on the vegetation and the configuration of the site, the cameras were installed on a tree or on a stake at a height varying from 50 cm to 1 m from the ground, at approximately 3 m and perpendicular to the potential path of the animal. Cameras were set to take three photos without delay between consecutive triggers in burst mode, as long as motion was detected. Cameras automatically record the date, time, moon phase, temperature of the captured events and the unique station’s label. Each camera was checked approximately every three weeks for data recovery, battery replacement and camera maintenance. Despite regular and frequent monitoring of the camera traps throughout the study period, some traps did not operate continuously over the study period due to various technical problems, thefts or damage by cattle. Although the sampling effort of each camera, in days, is indicated in the dataset (e.g. 134 days), the details of the functioning days of each camera for each pasture can be found on the following DOI: https://doi.org/10.48579/PRO/XUSK8V, in order to locate data gaps over time and distinguish them from periods without capturing events. During the entire study period, a total of 49,271 trap-days were sampled. The obtained pictures were manually identified on Timelapse Software (see https://saul.cpsc.ucalgary.ca/timelapse/pmwiki.php?n=Main.Licence). For wild species, two pictures separated by a time interval of at least two minutes were considered different capture events. When a capture event was recorded, only the first photo where the species was identifiable was annotated (i.e. species, temperature, date, time and number of individuals). When different individuals of the same species appeared later in the same capture event, the total number of individuals in the capture event was filled in using the first annotated photo. However, if an individual of another species appeared later in the same capture event, a new photo was annotated as the first identifiable photo of the species. Consequently, the date and time indicated for each capture event do not necessarily correspond to the exact number of individuals passing through at that precise time, but rather to the start of an event of variable duration containing that number of individuals. In some rare cases, two photos separated by an interval of more than two minutes were considered a single capture event. For example, an individual may have been sleeping in front of the camera, triggering it only occasionally. In such cases, a single event was recorded. Regarding livestock (i.e. cows, equids, goats and sheep), the photos were annotated on the basis of an hour's time. Thus, the minimum time interval between two consecutive capture events was one hour. In addition, as their behaviour can involve long periods of static in front of the cameras, in cases where a domestic species was continuously recorded for several hours, a new capture event was registered every hour after the start of the event. Moreover, for livestock data, the number of individuals was not recorded precisely since it mainly corresponded to the passage of herds. Thus, the value entered in the dataset is “many.” A precise number of individuals was only registered when an event showed only five sheep or less in order to identify isolated groups of sheep, which was relevant to answering the study goals. The dataset provided contains rows corresponding to each occurrence and, therefore, only shows information corresponding to the single annotated photo in each capture event. When two species appeared at the same time in a photo, the photo was annotated with both species, inducing two occurrences for one capture event. In the dataset, the row for this capture event was duplicated to obtain one row per occurrence.

Quality Control

The species identification of all tagged photos has been double-checked by the authors. At the first identification, the error rate was 0.32% and these identified errors were corrected at the second verification, bringing the final error rate to nearly 0%. Potential gaps (e.g. missing temperature information) have also been checked and photos of humans were systematically removed from the dataset.

Method steps

1. The study areas – three mountain pastures in Ariège – were selected on the basis of criteria relevant to the project (e.g. differences in pastoral activity, confirmed and dense bear presence, accessibility etc.). The choice of sectors was also dictated by the agreement of the pastoral groups present on the pastures to host the project for at least three years. Authorisations to use vehicles on the tracks approaching the pastures and authorisations to set photographic traps were then requested from the manager of these territories, in this case the Office national des forêts (ONF), which oversees the integrity of the forest estate, the conservation of structures and the protection of forest stands and natural environments, as well as the management of wildlife and hunting. A 400 m ✕ 400 m grid covering the entire surface of each mountain pasture was chosen according to the objectives of the study, as mentionned above and the location of the camera traps in each grid cell was chosen according to the most suitable place for fauna to pass (e.g. path, trail, mountain pass, thinning of vegetation etc.). When the camera traps were installed, the main habitat in which the cameras were installed was noted. As this information was not necessary for the project's objectives, habitat characterisation was done on sight and in an arbitrary manner. The cameras, set to take three photos without delay between consecutive triggers in burst mode as long as motion was detected, were installed approximately between May and October of the three years of the project, depending on field constraints (e.g. snow cover) and were checked approximately every three weeks. The data collected were processed using Timelapse software, according to the photo completion protocol explained above. Photos have been identified at the class, family or species level and photos of humans were systematically removed from the dataset. The dataset was then checked to control identification errors and potential gaps (e.g. missing temperature information) and to standardise the data. The dataset, deposited on the Global Biodiversity Information Facility platform (GBIF), was standardised in Darwin Core Archive (DwC-A) format to create an occurrence dataset corresponding to the standards recommended by Darwin Core.

Whenever possible, the taxonomic identification was made at the species level. Due to study objectives and identification difficulties, birds, equids and mustelids were not identified at the species level, but at the class level or at the family level, except for Western capercaillie and European badger. Thus, a total of one class, two families and nineteen species were recorded.

1. Aves
   common name: Birds rank: class
2. Equidae
   common name: Equids rank: family
3. Mustelidae
   common name: Mustelids rank: family
4. Bos taurus
   common name: Cow rank: species
5. Canis familiaris
   common name: Dog rank: species
6. Capra hircus
   common name: Feral goat rank: species
7. Capreolus capreolus
   common name: Roe deer rank: species
8. Cervus elaphus
   common name: Red deer rank: species
9. Felis silvestris
   common name: Wildcat rank: species
10. Lepus europaeus
    common name: European hare rank: species
11. Marmota marmota
    common name: Alpine marmot rank: species
12. Meles meles
    common name: European badger rank: species
13. Ovis aries
    common name: Sheep rank: species
14. Rupicapra rupicapra
    common name: Pyrenean chamois rank: species
15. Sciurus vulgaris
    common name: Eurasian red squirrel rank: species
16. Sus scrofa
    common name: Wild boar rank: species
17. Tetrao urogallus
    common name: Western capercaillie rank: species
18. Usrus arctos
    common name: Brown bear rank: species
19. Vulpes vulpes
    common name: Red fox rank: species
20. Dama dama
    common name: Fallow deer rank: species
21. Erinaceus europaeus
    common name: West european hedgehog rank: species
22. Genetta genetta
    common name: Common genet rank: species

The present study focuses on three Pyrenean summer pastures, Arreau (around 890 ha), Barestet (around 645 ha) and Ourdouas (around 527 ha), two of which are located in the Ariège Department and one of which is located in both the Departments of Ariège and Haute-Garonne, France. The three pastures are respectively part of one, two and one municipalities: Seix, Melles – Saint-Lary and Sentein.