Figure 2 - uploaded by Anthony Caravaggi
Content may be subject to copyright.
Photographs of (A) European hare Lepus europaeus (©Nigel Blake) and (B) Irish hare Lepus timidus hibernicus (©Shay Connolly) extracted with permission from Caravaggi et al. (2015), demonstrating clear intraspecific differences enabling species ID from both diurnal (C, European hare; D, Irish hare) and nocturnal (E, European hare; F, Irish hare) camera trap footage.
Source publication
Camera traps are used to estimate densities or abundances using capture-recap- ture and, more recently, random encounter models (REMs). We deploy REMs to describe an invasive-native species replacement process, and to demonstrate their wider application beyond abundance estimation. The Irish hare Lepus timidus hibernicus is a high priority endemic...
Similar publications
Effective wildlife management requires information on population status and distribution. Survey methods that provide estimates of these population parameters can vary greatly in effort required, area covered, precision of estimates, and cost. Trade-offs are required, because increasing precision and area coverage generally requires increasing fiel...
Citations
... capture time and date, location, activity, behaviour), this information is highly beneficial for studying various aspects of wild animal populations' biology and ecology. It can be used for tasks such as disease monitoring and control [31], understanding the animal's role in the ecosystem [36], monitoring invasive species [10], or measuring the involvement of humans in the animal's habitat and ecological restoration [8]. ...
We introduce a new wildlife re-identification dataset WildlifeReID-10k with more than 214k images of 10k individual animals. It is a collection of 30 existing wildlife re-identification datasets with additional processing steps. WildlifeReID-10k contains animals as diverse as marine turtles, primates, birds, African herbivores, marine mammals and domestic animals. Due to the ubiquity of similar images in datasets, we argue that the standard (random) splits into training and testing sets are inadequate for wildlife re-identification and propose a new similarity-aware split based on the similarity of extracted features. To promote fair method comparison, we include similarity-aware splits both for closed-set and open-set settings, use MegaDescriptor - a foundational model for wildlife re-identification - for baseline performance and host a leaderboard with the best results. We publicly publish the dataset and the codes used to create it in the wildlife-datasets library, making WildlifeReID-10k both highly curated and easy to use.
... We were unable to explicitly demonstrate if hares actively (Figure 1c). The home range sizes we recorded for the Irish hare population at Dublin Airport were similar to those reported elsewhere in Ireland (Caravaggi et al., 2016;Reid, 2023;Wolfe & Hayden, 1996). Hares do not inherently avoid tarmacked surfaces, however. ...
Understanding how animals move and use space within an environment is vital for the development and implementation of effective management actions. Within airfield environments, animal movement can present a substantial risk to aircraft, resulting in wildlife‐aircraft collisions (strikes) if animals enter into the manoeuvring areas of the airfield, namely the runways, taxiways and areas that connect the two (hereafter collectively referred to as ‘tarmacked areas’). However, reliable ecological data to inform management decisions can be difficult to obtain in such environments, due to access restrictions. Here, we present the first GPS data describing the movement ecology and spatial use of mammals on an airfield – Irish hare (Lepus timidus hibernicus), at Dublin International Airport – through the deployment of five GPS collars. A total of 4571 tarmacked area interactions were recorded between December 2021 and August 2022, with all five hares engaging with tarmacked areas. Between December and August, the highest number of interactions were recorded for the month of April (n = 1073), followed by March (n = 703). There was a mean of 4.3 (range: 0–65) interactions with tarmacked areas, per hare, per day throughout the study period. Hares most frequently engaged with tarmacked areas between 05:00 and 07:59, with some seasonal variation. The greatest cumulative distance moved per month was observed in May (505 km) and April (503 km). We identified that the average home range size of collared hares was 2.8 km² (±SD 0.1 km²), based on 95% Kernel Utilisation Distribution. Furthermore, we demonstrate that the hares incorporate tarmacked area habitat types into their home ranges with up to 13% of one individual's movements incorporating these areas. Our study demonstrates the suitability of GPS tracking devices for studying the movement ecology of high‐risk mammal species at airfields in order to inform airside management practices.
... Common to the more recent of these methods is the requirement to estimate the distances and angles from the camera trap to the locations of detected animals. Various methods have been developed for this, such as tracking animal movements on the images based on nearby landmarks (Rowcliffe et al., 2011) or calibration imagery (Henrich et al., 2023;Wearn et al., 2022), using a physical cane grid (Caravaggi et al., 2016), or using poles along the midline of the field of view (FOV; Hofmeester et al., 2017;Mason et al., 2022). ...
... Despite the obvious complexities of applying OCTDE in heterogeneous populations, this method has many advantages over those previously established. Firstly, apart from the camera trap, no extra equipment, such as cane grids (Caravaggi et al., 2016) or poles (Hofmeester et al., 2017;Mason et al., 2022), is needed when setting up a new camera site (although they can be used to support OCTDE by generating study-specific average size metrics). This is not only beneficial for making the standardisation and logistics of studies easier, but it also eliminates some weak points in study design, since cane grids and poles can be broken or knocked over by animals. ...
... such as REM (Rowcliffe et al., 2008). Further research could usefully assess how OCTDE can be complemented with methods relying on physical objects in the camera site (Caravaggi et al., 2016;Hofmeester et al., 2017;Mason et al., 2022;Rowcliffe et al., 2011) to achieve the optimal balance between practicality and accuracy. ...
Camera traps have great potential for generating wildlife insights by providing high resolution site‐specific data. Methods of data collection and analysis reliant on these tools for population density estimation can be relatively resource intensive, hindering their mainstream adoption.
Here, we explore the potential of population density estimates derived from a distance sampling method based on optics theory, which greatly simplifies the process of setting up camera sites and analysing data. Specifically, we (1) tested the method on human subjects in an artificial environment, (2) compared it to another method relying on virtual grids on images using wild populations of black‐backed jackal ( Canis mesomelas ) and African civet ( Civettictis civetta ) in South Africa and (3) deployed it to estimate wild boar ( Sus scrofa ) population densities in Hungary.
The initial human trials resulted in an estimate that was extremely close to true population density. When compared to the virtual grid method, results suggest that our distance sampling method can deliver accurate estimates with increased convenience and robustness against disruptions of the camera sites. The wild boar study resulted in a realistic density estimate, which can be used as a baseline when assessing future fluctuations in population density.
As this new approach does not have special requirements for setting up camera sites, it is efficient and widely applicable across other density estimation methods requiring an estimate for effective detection distance. Additionally, the method can be applied in the retrospective analysis of existing datasets.
... The use of reference images combined with image processing techniques provides means by which snow depth measurements can be remotely acquired, precluding issues with expense, installation logistics, and animal responses to objects left in the camera viewshed. While we have not found studies using reference images to measure snow depth, Henrich et al. (2022) placed transparent images containing distance markers on top of images and Caravaggi et al. (2016) superimposed gridlines onto images to assess distance to animals for distance sampling, eliminating the need for in situ distance markers. ...
... The edger package may also be useful for camera distance sampling techniques. The methods would be similar to those performed by Caravaggi et al. (2016). A reference image would need to be taken containing objects placed at regular distance intervals in the camera viewshed. ...
Remote cameras are used to study demographics, ecological processes, and behavior of wildlife populations. Cameras have also been used to measure snow depth with physical snow stakes. However, concerns that physical instruments at camera sites may influence animal behavior limit installation of instruments to facilitate collecting such data. Given that snow depth data are inherently contained within images, potential insights that could be made using these data are lost. To facilitate camera-based snow depth observations without additional equipment installation , we developed a method implemented in an R package called edger to superimpose virtual measurement devices onto images.
... For unmarked methods, it is well described that estimating movement parameters, and especially day range, limits their applicability because it is time-consuming to estimate this parameter (Nakashima et al. 2018). Initially, the estimation of day ranges and activities was based on tagging some animals in the population with telemetry devices (Caravaggi et al. 2016). Marking limits the applicability of unmarked methods because it increases the time and costs related to field application and commonly used analytical procedures applied to telemetry data underestimate the day range (Rowcliffe et al. 2012, Sennhenn-Reulen et al. 2017. ...
Motion‐sensitive cameras are popular as non‐invasive monitoring tools, and several methods have been developed to estimate population densities from camera data. These methods frequently rely on auxiliary movement data including the distance traveled by an individual in a day and the proportion of the day that an animal spends moving when individual recognition is not possible. The estimation of these movement parameters is time‐consuming, which could limit the applicability of cameras to estimate population density. To investigate the relevance of measuring movement parameters for the target population, we monitored 54 wildlife populations of red deer ( Cervus elaphus ), fallow deer ( Dama dama ), roe deer ( Capreolus capreolus ), wild boar ( Sus scrof a), and red fox ( Vulpes vulpes ) in different seasons through Europe with cameras. We estimated 91‐day ranges and activity levels. We fitted mixed models for day range and activity level as response variables to assess if the inter‐population variability in movement was explained by a set of a priori relevant geographical, environmental, biological, and management predictors. We then explored the bias in density estimates obtained in 25 independent populations when using predicted movement data. There was high intra‐species variation in day range and activity level among species and populations. Only species explained a small proportion of this variability; other predictor variables did not. We observed bias in densities when predicting the day range and activity for independent populations. Considering the intra‐species variability in movement parameters and the consequent unacceptable bias in density estimates, we recommend that monitoring and conservation programs estimate movement parameters for the target population and survey populations from camera data for more accurate density estimates. While this increases the handling time needed to estimate densities, it is worth the cost because of the reliability of camera‐based methodologies to estimate needed movement parameters.
... The disadvantages of this method are intensive labor and high costs in the field as well as the difficulty in estimating the daily travel distance of individual animals. Thus, only a few studies have been conducted using the REM model [52][53][54][55][56][57][58][59][60]. Therefore, a more efficient approach is highly needed. ...
Animal population abundance is a significant parameter for studies on invasive species that can threaten the ecosystem. Researchers have been developing population estimation methods since the 18th century, in order to evaluate species’ evolution and environmental effects. However, studies on the population density of the invasive species Callinectes sapidus are very limited. The present work, using a simulation model combined with field measurements, examines an innovative methodology for estimating the current population of the invasive species Callinectes sapidus in a shallow Mediterranean coastal lagoon. The methodology presented here builds the first stage of modeling and predicting the evolution of this species’ population in marine environments. The simulation model’s results are validated with an estimation of the total population based on juvenile abundance, and a curvature of the species population estimation based on cage catch is implemented. The simulation experiments presented here show the possibility of a robust prediction for blue crab population estimation.
... This survey method has provided a reliable, concealed, and intuitive method of wildlife investigation that greatly reduces the workload of investigation [23]. CT is becoming increasingly popular in ecology and can provide a great amount of information [24,25]. With continuous improvement in CT function and a reduction in equipment costs, CT has gradually become the main method for studying wildlife behavior [26]. ...
Simple Summary
Animals are subject to the daily and seasonal fluctuations of climatic factors that affect their thermoregulation. The purpose of the present study was to determine the activity patterns of wild plateau pikas (Ochotona curzoniae), with a special emphasis on the effects of thermal conditions. We set 14 camera traps in a coniferous forest, Yakushima, and monitored them for a total of 850 camera days between October 2017 and September 2018. During the cold season, pikas concentrated their activity around noon and exhibited unimodal patterns. During the warm season, pikas concentrated their activity in the morning and afternoon and exhibited bimodal patterns. Plateau pikas occupied a cool and windless microclimate during cold seasons and a cool and humid microclimate during warm seasons to avoid excessive cooling or overheating pressure. These results indicate that changes in the microclimatic conditions of their habitats (specifically, increases in temperature) might significantly restrict the daytime activity of plateau pikas. These results can provide baseline information for future research on pikas, especially that which is focused on their behavioral ecology and the effects of climate change in some Chinese alpine regions.
Abstract
Exploring the activity patterns of small mammals is important for understanding the survival strategies of these animals, such as foraging and mating. The purpose of the present study was to determine the activity of free-living plateau pikas (Ochotona curzoniae) in different months and seasons (cold and warm seasons), with a particular emphasis on the effects of weather condition. Based on a camera-trapping survey conducted from October 2017 to September 2018, we evaluated the activity patterns and activity levels of plateau pikas inhabiting the eastern Qinghai–Tibet Plateau in China. The effects of environmental factors on the activity of plateau pikas were examined using the generalized additive mixed model (GAMM). The results showed that: (1) The plateau pikas exhibited unimodal patterns of activity during the cold season (October–April). During the warm season (May–September), the activity patterns of the plateau pikas were bimodal. Their activity levels were highest in June. (2) During the cold season, their activity levels rose gradually over the course of the day to a peak near noon, and they were not significantly higher after sunrise than they were before sunset. During the warm season, their activity peaks were in the morning and afternoon, and their activity levels were substantially lower after sunrise than they were before sunset. (3) The plateau pikas were more active under conditions with lower ambient temperatures and precipitation during the cold and warm seasons. While relative air humidity was positively correlated with the activity of the plateau pikas during the warm season, wind speed was negatively correlated with the pikas’ activity during the cold season. Overall, these results collectively indicate that plateau pikas occupy habitats with cool and less windy microclimates during the cold season, and with cool and moist microclimates during the warm season. Information on the time allocation of pikas’ activity levels during different seasons should provide a baseline for understanding their potential for adaptation to climate change.
... In addition to image appearance discrepancies in datasets from different study sites, expanding surveys or monitoring programs can also change the composition of animal species recorded (Kays et al., 2020). For example, as data collections continue over time, previously undetected species may be encountered (Prach & Walker, 2011) [e.g., less frequent species (Pimm et al., 2014), recolonizing species (David Mech et al., 2019), reintroduced animals (Taylor et al., 2017), or invasive species that are harmful to the ecosystem (Caravaggi et al., 2016;Clavero & Garcia-Berthou, 2005)]. When novel species are introduced, our standard classification model is no longer effective because conventional AI methods require datasets to have fixed numbers of classes (Arjovsky et al., 2019). ...
Remote aerial sensing provides a non-invasive, large geographical-scale technology for avian monitoring, but the manual processing of images limits its development and applications. Artificial Intelligence (AI) methods can be used to mitigate this manual image processing requirement. The implementation of AI methods, however, has several challenges: (1) imbalanced (i.e., long-tailed) data distribution, (2) annotation uncertainty in categorization, and (3) dataset discrepancies across different study sites. Here we use aerial imagery data of waterbirds around Cape Cod and Lake Michigan in the United States to examine how these challenges limit avian recognition performance. We review existing solutions and demonstrate as use cases how methods like Label Distribution Aware Marginal Loss with Deferred Re-Weighting, hierarchical classification, and FixMatch address the three challenges. We also present a new approach to tackle the annotation uncertainty challenge using a Soft-fine Pseudo-Label methodology. Finally, we aim with this paper to increase awareness in the ecological remote sensing community of these challenges and bridge the gap between ecological applications and state-of-the-art computer science, thereby opening new doors to future research. © 2023 The Authors. Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
... Despite the number of options currently available to estimate density of unmarked species, many studies only calculate relative abundance metrics (Gilbert et al. 2021). This may in part be due to the recency of the methods for unmarked species, the fact that some of these have only been tested in simulations (but see Caravaggi et al. 2016;Loonam et al. 2021), and their technical requirements, such as the need for independent speed estimates (Rowcliffe et al. 2011). It is also not clear that methods for unmarked species can give reliable density estimates that are comparable to more established methods. ...
... It is also not clear that methods for unmarked species can give reliable density estimates that are comparable to more established methods. Previous studies have compared REM with visual census estimates (Caravaggi et al. 2016;Cusack et al. 2015), and REM, TTE and STE have been compared to SCR estimates from fecal samples identified with molecular methods (Loonam et al. 2021). No study, however, has performed similar tests using field data exclusively from camera traps. ...
Camera-traps have become one of the most common tools for studying wildlife abundance and population density. Traditionally, absolute density could be estimated only for species with individual markings, using capture–recapture frameworks. Newer methods allow to estimate density of unmarked species, but these have yet to be thoroughly tested and compared against capture–recapture methods. To make this comparison requires an identifiable species, for which both types of frameworks can be used. Here, we estimate the population density of ocelots (Leopardus pardalis) in the Osa peninsula, Costa Rica, comparing methods for marked and unmarked species. We deployed camera-trap grids between 2017 and 2019, identified individuals and determined spatially resolved individual detection histories, station-specific detection frequencies and times to first detection. Estimates obtained with methods for unmarked species (Time-to-Event and Random Encounter Model) varied widely among surveys, from 11 to 169 individuals/100 km², and were significantly different from spatial capture–recapture estimates (28.1 individuals/100 km²). Differences were largely driven by the non-random placement of cameras on human-made trails, which inflated the detection frequency. Maximizing the number of encounters benefits methods based on capture–recapture but is detrimental for methods based on random detections. Our results highlight the incompatibility between surveys designed for capture–recapture analyses, and those that assume random movement of animals. For recently developed unmarked species methods to be used for a larger and more diverse set of species, it is necessary to further test and define the requirements and factors that affect their calculations. This information will ultimately allow for a greater diversity of population and community studies.
... Barrett-Hamilton (1898) reported that several populations established successfully and proliferated, but most disappeared by the end of the 19th century. While there have been isolated reports of European brown hares (Fairley 2001;Sheppard 2004), the only confirmed extant population is 1,000-2,000 individuals in Mid-Ulster spanning south-east County Derry and east County Tyrone (Reid and Montgomery 2007;Caravaggi et al. 2015Caravaggi et al. , 2016. Irish hares have adapted to warmer conditions than elsewhere in the mountain hare range, exploiting habitats from mountains to the coast (Lysaght and Marnell 2016) in the absence of a lowland competitor. ...
... Thus, 56 roadkill hares were collected opportunistically from the vicinity of the nonnative L. europaeus population established in Mid-Ulster (Counties Derry and Tyrone) accumulated between 2003 and 2008. Whilst sightings of L. europaeus have been made near Baronscourt Estate, west County Tyrone (Reid and Montgomery 2007), the population in Mid-Ulster is the only known extant population (Reid 2011;Caravaggi et al. 2016). No attempt was made to attribute a species phenotype to each carcass, which was determined genetically (see below), as many roadkill carcasses were badly damaged. ...
... Bidirectional introgression was also recorded in the present study with hybrids backcrossing predominately with the native (69% of individuals with mixed nuclear sequences had L.t. hibernicus mtDNA) with L. europaeus backcrossing more limited: 31% of individuals with mixed nuclear sequences had L.t. hibernicus mtDNA). This > 2:1 ratio reflects the 3:1 non-native:native ratio in population density at the invading wavefront of L. europaeus (Caravaggi et al. 2016) suggesting the frequency of bidirectional introgression reflects the populations of parental and hybrid stock and availability of potential mates. ...
Introduced non-native species can threaten native species through interspecific hybridisation and genetic introgression. We assessed the prevalence of hybridisation and introgression between introduced European brown hare, Lepus europaeus , and the endemic Irish hare, L. timidus hibernicus . Roadkill hares ( n = 56) were sequenced for a 379bp section of the mitochondrial DNA D-loop and a 474bp segment of the nuclear transferrin ( Tf ) gene. A species-specific indel in the transferrin gene was present in L.t. hibernicus and absent in L. europaeus . Excluding three hares from which molecular data could not be recovered, 28 hares (53%) were native L.t. hibernicus , 7 (13%) were non-native L. europaeus and 18 (34%) were hybrids; of which 5 (28%) were first generation (F1) involving bidirectional crosses with mismatched nuclear and mtDNA (3 ♂ europaeus x ♀ hibernicus and 2 ♂ hibernicus x ♀ europaeus ). Mixed nuclear transferrin sequences suggested 13 (72%) of hybrids were at least 2nd generation (F2) with 9 (69%) possessing L.t. hibernicus and 4 (31%) L. europaeus mtDNA (the latter indicative of hybrid backcrossing with the non-native). The prevalence of hybridisation at similar mountain-brown hare contact zones throughout Europe is notably lower (4–16%) and typically unidirectional (♂ europaeus x ♀ timidus ). A high prevalence of bidirectional hybridisation and introgression (in association with projected climate change) may favour the introduced species over the native. Genetic surveillance and population monitoring are needed to further explore the potential conservation implications of European brown hare in Ireland.
