Estimating accurately and precisely changes in population size and density over time is essential to inform management decision-making and monitor species conservation status (Caughley & Sinclair 1994; Nichols & Williams 2006;). Failure to implement well-designed monitoring programs may lead to misguided management decisions and irreversible shifts in the ecosystem (e.g., Fairweather 1991; Gibbs et al. 1999; Field et al. 2005). It is also useful for understanding ecological, behavioral, or genetic processes (Ojaveer et al. 2004; Dochtermann & Peacock 2013; Valderrama et al. 2013).
When resources are unlimited, populations exhibit (a) exponential growth, shown in a J-shaped curve. When resources are limited, populations exhibit (b) logistic growth. In logistic growth, population expansion decreases as resources become scarce, and it levels off when the carrying capacity of the environment is reached. The logistic growth curve is S-shaped.
https://openoregon.pressbooks.pub/envirobiology/chapter/4-2-population-growth-and-regulation/
Methods based on direct animal surveys of marked animals, such as Capture-Mark-Recapture (CMR) and distance sampling, are generally considered to provide the most reliable estimates of population density of large mammals (Morellet et al. 2007). However, they are costly, not sustainable on a long-term and require animal capture and handling, which involve dangers of injury or death to the animal (Arnemo et al. 2006) and raise ethical issues for endangered species (Bekoff & Jamieson 1996).
Besides, in tropical rainforests and in mountains, monitoring populations of large terrestrial mammal species using direct counts can be particularly challenging (Thompson 2004, de Souza Martins et al. 2007), especially when animals are rare, elusive, solitary, largely nocturnal, highly mobile, far-ranging, and/or inhabiting remote or rugged habitats.
For these reasons, for such populations, monitoring often rely on indirect sampling methods, such as noninvasive field methods (Long et al. 2008), consisting in the collection of animal presence signs (such as hair, scats, tracks, visual observations, claw or teeth marks, photographs, videos, opened ant-hills, daybeds, livestock predation).
Classification of different methods to estimate density and /or relative abundance based on desirable characteristics for monitoring populations (A=relative abundance, D=density) in local management units, practicability, and applicability in epidemiological aspects. y=yes; ~=restricted; n=no; L=Low, H=High, M=Moderate, A=accuracy, P=precision; 1= (possible, but not usual); 2=yes with restrictions, 3= Maybe valid for monitoring and used as relative abundance for a given population, 4= Valid for monitoring along time and space as relative abundance when precision or accuracy is low; ; 5=possible
https://enetwild.com/2018/08/31/guidance-to-estimate-wild-boar-density/
Noninvasive field methods associated with genetic analyses (so called noninvasive genetic sampling, hereafter NGS) are increasingly popular and are usually considered as cost-efficient approaches (e.g., De Barba et al. 2010; Mowry et al. 2011; Sugimoto et al. 2012; Stansbury et al. 2014).
The relationship between non-disruptive, non-invasive and non-lethal DNA sampling methods. Non-invasive DNA sampling sensu stricto corresponds to the definition given by Taberlet et al. (6), Non-invasive DNA sampling sensu lato corresponds to the medical definition (71). Pictograms represent a non-exhaustive list of examples for which references are given below. From left to right and top to bottom: whole faeces sampling for species that use faecal territory marking (113), hairs collected in snow (51), hairs collected with unbaited barbed wire(44), DNA trap baited to attract animals (114), skin swabbing in the field without capture (95), capture of reptiles for buccal swabbing (115), gun darting of big mammals to collect tissue sample(116), biopsy on handled invertebrate (117).
Proportion of DNA sampling techniques (lethal, invasive, non-invasive, or both invasive and non-invasive) used in the identified relevant genetics studies, categorized by the research objective of the study (Type of study; left) and the genetic marker (Marker; right).
Proportion of identified relevant studies on genetics in amphibians, birds, carnivores, molluscs and rodents published in 2017–2018. Lethal and invasive methods of DNA collection are in red tones, non-invasive in blue tones, grey symbolizes that both invasive and non-invasive methods were used for DNA collection. Except for studies on carnivores, majority of studies used a lethal or an invasive genetic sampling technique.
Categories of Genetic Monitoring as defined by Schwartz et al. 2007
a: Number of articles in relation to the sampling method used between 2013 and 2018. The year 2018 is incomplete as the search was performed in July 2018.
b: Countries of origin of the samples analysed in the reviewed papers. Countries in grey were not represented in our review, countries coloured in various shades of green provided samples for 1 to 75 of the reviewed papers (see in-graph legend for colour scale).
c: Bipartite network of the main aim of the studies in blue, the type of sampling method used in orange (see Table 1 for definitions) and the nature of the samples collected in green. The horizontal width of the rectangles represents the number of articles in each category.
Non-invasive genetic sampling methods
Since 2009, three different methods of non-invasive genetic sampling have been used to estimate the annual Minimum Detected Size of the Pyrenean brown bear population:
1) a systematic collection of hair and scats on transect trails;
2) a systematic collection of hair using hair traps associated with automatically triggered camera traps;
3) an opportunistic collection of hair and scats throughout the bear range.
We assessed and compared the cost-effectiveness of these three different non-invasive genetic sampling techniques to estimate the annual size of the Pyrenean brown bear population from 2009 to 2018 where the size of the population, age and sex of individuals were known.
Copyright : ONCFS-Equipe Ours
Copyright : ONCFS-Equipe Ours
Copyright : ONCFS-Equipe Ours
Copyright : ONCFS-Equipe Ours
Comparison of the size (with its 95% Confidence Interval) of the centro-oriental subpopulation of brown bear estimated for each of the four sampling techniques (i.e. ST: Systematic by trails, SH: Systematic by baited hair traps, SC: Systematic by camera traps and OP: Opportunistic monitoring) using closed population capture-recapture models, with the Minimum Detected Size (MDS) estimated for the whole study area and the French side only, for each year of monitoring (between 2010 and 2012 for ST, SC and OP, between 2009 and 2011 for SH and between 2009 and 2012 for MDS).
Scat-detection dog method combined with non-invasive genetic sampling
To date, the Pyrenean brown bear population abundance has mainly been estimated from noninvasive genetic analysis of hair and scat samples, field-collected either opportunistically or using systematic monitoring approaches (Sentilles et al. 2020). Hair and scats are very important non-invasive clues to provide information on species and individual identity through genetic analyses (Foran et al. 1997; Waits et al. 2001) and to evaluate population dynamic parameters (e.g, Taberlet et al. 1995). Scats can also provide important insights into the ecology and natural history of the target species and inform research focusing on diet (e.g., De Barba et al. 2014), seed dispersal (e.g., Lalleroni et al. 2017), genetics (e.g., Kohn et al. 1995), physiological stress (e.g., von der Ohe et al. 2014), parasitism and health (e.g., Sheikh et al. 2017).
In the French Pyrenees, despite the participation to non-invasive Pyrenean brown bear population monitoring of more than 400 professionals and volunteers from the Brown Bear Network (ROB), bear scats are not easily detected by humans in the field, especially in such steep habitats with dense understory. For instance, between 2009 and 2013, scats represented only 4% of the total number of bear clues collected in the field, while hair represented 43%, tracks 17%, photos and videos from camera traps 16%, attacks on livestock and beehives 15%, and other cues (e.g. visual observations, scratches on trees) 5%. Besides, bear scats are not readily visually distinguishable in the field from scats of other mammal species with similar diet. In the Pyrenees, bear cub scats, in particular, can be easily confused with those of red foxes (Vulpes vulpes Linnaeus, 1758) and European badgers (Meles meles Linnaeus, 1758).
In North America, it has been fairly well established that large carnivore scat detection can be improved through the use of scat detection dogs (see references in Orkin et al. 2016, Grimm-Seyfarth et al. 2019). In 2013, we therefore decided to incorporate this technique in future efforts to monitor the Pyrenean brown bear population, by training a dog from an early age to detect bear scats.
Our results clearly show, using the Pyrenean brown bear as a case study, that even using a single scat-detection dog/handler team (see de Oliveira et al. 2012 for a similar study) can greatly improve the efficacy of collecting target scats, and increase the probability of acquiring genotypes and detecting individuals from scat samples (Sentilles et al. 2021 J Vertebr Biol).
Copyright : Cécile Vanpé
Copyright : Cécile Vanpé
Copyright : Cécile Vanpé
Copyright : OFB-Equipe Ours
Density estimation through morphological identification from camera trapping profiles
The ability to recognize and to follow individuals of a population over space and time is fundamental for wildlife managers and researchers in many ways, such as to evaluate individuals’ survival, reproduction, movements, habitat use and selection, home ranges, social behaviour and state of health (Morrison et al. 2011; Bolger et al. 2012). In addition, reliable individual identification is a key requirement in capture-recapture surveys (Stevick et al. 2001; Lukacs & Burnham 2005; Yoshizaki et al. 2009; Morrison et al. 2011; Goswami et al. 2012), the most robust approaches to estimate animal population size when complete counts are logistically impossible (Nichols 1992).
Traditionally, individual identification has been done by physically capturing and artificially marking animals using visible and unique marks (e.g., coloured ear tags, radio or GPS collars; Williams et al. 2002). However, these invasive techniques have many drawbacks (McMahon et al. 2011), including altering animal behaviours and physiology (Wilson & McMahon 2006), risking animal injury or death (McMahon et al. 2005; Arnemo et al. 2006), and being expensive and logistically challenging (Powell & Proulx 2003). In the case of endangered elusive large-bodied animal species, non-invasive methods, such as non-invasive genetic sampling (Waits & Paetkau 2005) and tracking natural markings from photographs (e.g., Hammond et al. 1990; Karanth 1995), are therefore clearly preferred to identify individuals.
With recent advances in digital technology and image-processing software and decreasing costs of camera traps, camera trapping has become increasingly popular over the last decade (Kays & Slauson 2008; O’Connell et al. 2011) to identify individuals based on unique phenotypic characteristics and estimate population size (Foster & Harmsen 2012). Recognition of individuals on photographs has been essentially done using unique obvious pelage patterns of individuals (e.g., stripes in tigers Panthera tigris: Karanth 1995; Karanth & Nichols 1998; rosettes in jaguars Panthera onca: Silver et al. 2004), more subtle unique natural markings (e.g., a set of natural features including black markings on legs and face, white tip of tail, tail shape, hair patterns on the flanks in maned wolfs Chrysocyon brachyurus: Trolle et al. 2007), or environment-induced unique markings on the body of the animals (e.g., scars and torn ears in tapirs Tapiris terrestris: Noss et al. 2003). Nonetheless, for the numerous species lacking such color patterns or natural markings, other individual characteristics, such as morphological traits, could be used (Goswami et al. 2012). Yet, studies using measurements of morphological traits extracted from camera trap photographs to recognize individuals are rare, and are restricted to measurements of ornaments (e.g., antler, horn, tusk) characteristics (Merkle & Fortin 2014 in free-ranging bisons Bison bison) or used a large combination and diversity of morphological features (e.g., shoulder height, tail and tusk length, tusk angle, presence/absence of ear fold, tear and hole, presence/absence of tumors and scars, ear lobe shape in Asian elephants Elephas maximus: Goswami et al. 2007, 2012). The reliability of such techniques is indeed constrained by the inter-individual variability of the traits at the population level, the stability of the traits over the duration of the study period, the visibility of the traits under different environmental conditions and different photograph quality and the precision of trait measurements (Stevick et al. 2001; Goswamy et al. 2011; Morrisson et al. 2011; Bolger et al. 2012; Foster & Harmsen 2012). Importantly, since misidentification of individuals can lead to serious biased in parameter estimates in the context of capture-recapture studies (Mills et al. 2000), the use of such techniques requires a validation using data from known individuals (e.g., captive or artificially marked animals, or animals identified by independent sources such as molecular tools; Stevick et al. 2001; Foster & Harmsen 2012).
Photographic capture-recapture methods have now been tested in a wide range of large carnivores including mainly Felidae (e.g., Karanth & Nichols 1998; Silver et al. 2004; Jackson et al. 2006; Kelly et al. 2008) and Canidae (e.g., Larrucea et al. 2007; Trolle et al. 2007). In Ursidae, individual recognition has been accomplished using pelage patterns (such as presence, shape and colour of facial, neck, tail and chest markings) in red pandas (Ailurus fulgens) (Shrestha et al. 2015), Andean bears (Tremarctos ornatus) (Rios-Uzeda et al. 2007; Russell et al. 2014; Reyes et al. 2017), and Asian black bears (Ursus thibetanus) (Higashide et al. 2013). But no study, to our knowledge, in Ursidae or even large carnivores, has yet tested the reliability of morphometric measurements to identify individuals from camera trap photograph whole body profiles and ultimately estimate population size (but see the BearID project in progress by Mélanie Clapham: http://bearresearch.org/research-applications/).
Camera traps have been used since 1993 to monitor the Pyrenean population of brown bear. In this study, we aim to assess the reliability of a set of morphological measurements to identify individual brown bears from the Pyrenean population on camera trap photograph profiles. More specifically:
1) we quantified the intra- and inter-observer measurement errors,
2) we evaluated the effects of the site, period, environmental conditions and photograph quality on measurements,
3) we determined the best clustering method to discriminate between individual bears identified independently thanks to molecular tools, artificial marks or natural marks,
4) we tested our method using a large dataset of bear photographs from the Pyrenean population.
Genetic non-invasive CMR method
Abundance of small populations of large carnivores can generally be reliably estimated using total counts of the different individuals detected over a time period (so called minimum population size). In the case of genetic identification methods, minimum population size is then defined as the number of unique genotypes identified among the genetic samples inside the study area (e.g., Creel et al., 2003; Solberg et al. 2006). However, exhaustive counts are often expensive, time consuming, and logistically demanding (Blanc et al. 2013). In addition, as population is growing larger and distribution is expanding wider, the risk of under-estimating population size using total counts is increasing sharply due to the rarely fulfilled assumption of complete detection of all individuals of the population (Solberg et al. 2006). The need to report the uncertainty (variance) of the population estimate becomes also crucial (e.g., Forney 2000; McGowan et al. 2011).
Capture-Mark-Recapture (CMR) surveys are then considered as the gold standard to estimating wildlife abundance while accounting for imperfect ability to detect all individuals in a population (Pollock 1976, 1980; Otis et al. 1978). CMR methods allow taking account for possible variations in sampling effort of the population, as well as temporal and/or spatial variations in the probability of detection of individuals or groups of individuals. It also allows estimating the demographic parameters of the population (e.g., survival, reproduction, emigration) and evaluating the uncertainty surrounding the estimates. With the rapid development of non-invasive genetic sampling (Taberlet & Luikart 1999; Taberlet et al. 1999), the CMR method, that was originally limited to live-trapping studies, has been adapted since the mid-90s (Woods et al. 1996, 1999) to the use of scats, hair or other non-invasive tissue samples to identify individuals (Bellemain et al. 2005; Lukacs & Burnham 2005; Miller et al. 2005; Petit & Valière 2006).
To date, the size of the Pyrenean brown bear population has been annually estimated using the Minimum Detected Size (MDS) index, which corresponds to the total number of different individuals detected in the population during the year (Sentilles et al. 2020). The aim of this study is therefore to develop a more robust method, based on capture-recapture modelling, to estimate accurately annual abundance of the Pyrenean brown bear population and its trends over time, while taking into account the probability of detection of individuals.
Capture rates from camera trapping
With the recent development of camera trapping and associated sophisticated analytical methods (e.g., Karanth 1995; MacKenzie et al. 2006), a new tool is now available to monitor cryptic terrestrial mammals, especially in tropical forests (Seydack 1984; Griffiths & van Schalk 1993; Cutler & Swann 1999). Camera traps have proved to be efficient to inventory species, including rare species (e.g., Trolle 2003; Srbek-Araujo & Chiarello 2005; Srbek-Araujo & Garcia 2005; Trolle & Kery 2005; Azlan 2006; Giman et al. 2007; Tobler et al. 2008). They can also provide various informations on the ecology and behaviour of elusive species such as home range size, daily and seasonal activity patterns, habitat use, reproductive cycle, group size... (e.g., Carthew & Slater 1991; van Schaik & Griffiths 1996; Gómez et al. 2005; Rivero et al. 2005; Azlan & Sharma 2006 ; Di Bitetti et al. 2006). Finally, for individually identifiable species (e.g., based on spotted fur patterns, scars, antler shape…), population estimates can be estimated based on capture-recapture models (Karanth 1995; Karanth & Nichols 1998; Trolle & Kery 2003; Silver et al. 2004; Jackson et al. 2006; Karanth et al. 2006; Soisalo & Cavalcanti 2006; Trolle et al. 2007), while for other species, occupancy rates can be calculated (MacKenzie et al. 2006; Linkie et al. 2007). While camera trapping has been extensively used to estimate abundance of feline populations (Karanth 1995; Karanth & Nichols 1998; Wallace et al. 2003; Kawanishi & Sunquist 2004; Maffei et al. 2004; Silver et al. 2004) or other large carnivore populations (Crooks et al., 1998; Kucera and Barrett, 1993; Mace et al., 1994; Linkie et al. 2007), the application of this method to large herbivores such as Artiodactyles and Perissodactyles is much more recent and focuses mainly on tapir spp. (Noss et al. 2003; Trolle et al. 2008).
The aim of this study was to test camera trapping to study the habitat use and activity patterns and to monitor the abundance of medium and large terrestrial mammals in the tropical rainforest of French Guiana, focusing specifically on 4 important game species of ungulates: the red brocket deer (Mazama americana), the grey brocket deer (Mazama gouazoubira), the collared peccary (Pecari tajacu) and the lowland tapir (Tapirus terrestris).
Copyright : Cécile Vanpé
Copyright : ONCFS-RNN Nouragues
Copyright : ONCFS-RNN Nouragues
Density estimation through individual identification from camera trapping based on natural marks
The development of camera trapping (Seydack 1984, Griffiths & van Schaik 1993) and associated analytical methods (Karanth 1995) has offered a promising tool to monitor population abundance of large and medium-sized Neotropical cryptic mammals (Srbek-Araujo & Chiarello 2005), in which individuals are potentially identifiable based on temporal or permanent natural markings such as scars, ear notches, antler shape, spots or other variation in coloration. Yet, while camera trapping has been extensively used to estimate densities of wild cats (e.g., Karanth & Nichols 1998 on tigers Panthera tigris, Silver et al. 2004 on jaguars Panthera onca) and have regularly been applied in other large carnivores (e.g., Mace et al. 1994 on grizzly bears Ursus arctos and Trolle et al. 2007 on maned wolfs Chrysocyon brachyurus), the application of this method to ungulates is much more recent and has focused on large species like the lowland tapir Tapirus terrestris (e.g., Noss et al. 2003, Trolle et al. 2008, Tobler et al., in press) or were based on artificial marks like in wild boars Sus scrofa (Heibensen et al. 2008). Only one deer species has been investigated so far based on natural marks, i.e. the white-tailed deer Odocoileus virginianus (Jacobson et al. 1997), but while males were individually identified based on antler configurations, population size and density were estimated using ratios of males and females as well as fawns and adults photographed. No study has as yet applied this method on deer species based on natural markings whatever the sex and the age of the individuals.
The aim of this study was to evaluate the feasibility, reliability and usefulness of the camera trapping approach for estimating red brocket deer densities using spatially explicit capture-recapture (SECR) models (Efford 2004, Royle & Young 2008, Efford et al. 2009), by taking advantage of three camera trapping surveys conducted from 2006 to 2009 in the Nouragues nature reserve in French Guiana (Tobler et al., in press). We also compared these densities obtained from camera trapping with those estimated from distance sampling analyses of diurnal line transects sampled simultaneously on the same study area.
A and B display two different females.
C and D show the same female on two different days.
E and F show two different fawns.
White circles indicate natural marks on the body of the animals used to identify the individuals.
Copyright : ONCFS
Density estimation through pellet-group count
We used pellet-group counts to estimate roe deer abundance in Sweden. More specifically, we carried out a faecal pellet-group count census (Neff 1968; Cederlund & Liberg 1995) over the entire 2,600 ha roe deer research area (Bogesund), both in 2004 and 2005, in the very first days of April, immediately after snow melt. The 604 sample plots were circular, their size were 10 m² and they were distributed regularly along North-South transects in order that they cross varying slope aspects and altitudinal zones. Transects were spaced 400 m apart, and plots 100 m apart on the transects. Each plot location was georeferenced and the plots were searched with a GPS. We considered that no pellet degradation occurred during winter due to snow cover and low temperatures. Because new pellets that have deposited since defoliation could be distinguished from old pellets in relation to dead leaves and needles that covered the latter but not the former (Kjellander 2000) when the survey is conducted in the spring (Robinette et al. 1958), we used a standing crop design with temporary plots. We considered that no pellet degradation occurred during winter due to snow cover and low temperatures. Each plot have been systematically read twice, clockwise and counterclockwise. The total number of pellet groups was recorded on each plot. A minimum of 10 pellets are needed to constitute a group. Moose was the only other deer species in the area, so the risk of counting faeces of other species than roe was excluded. The number of pellet groups on each plot was interpolated using a statistical prediction. The total number of pellet groups was estimated within each male’s territory (based on both 90% and 50% Kernels) for 2004 and 2005 as the sum of the point estimates falling within each territory using arcview3·2 software. This value was used as a relative index of local roe deer abundance. Although pellet group counts have been criticised as a method to estimate absolute population density (e.g., Robinette et al. 1958; Putman 1984), we considered that they provide an informative index of relative animal abundance which was adequate for our purposes (see Forsyth et al. 2007). We then assessed the effect of habitat quality on female local abundance, and in turn, the effect of female local abundance on male territory size and breeding success (Vanpé et al. 2009 J. Anim. Ecol.).
Copyright : Cécile Vanpé
Copyright : Cécile Vanpé
Copyright : Cécile Vanpé
Non-invasive molecular sexing
Many lemur species are arboreal, elusive, and/or nocturnal and are consequently difficult to approach, observe and catch. In addition, most of them are endangered. For these reasons, non-invasive sampling is especially useful in primates including lemurs. A key issue in conservation and ecological studies is to identify the sex of the sampled individuals to investigate sex-biased dispersal, parentage, social organization and population sex ratio.
Several molecular tests of sex are available in apes and monkeys, but only a handful of them work in the lemuriform clade. Among these tests, the coamplification of the SRY gene with the amelogenin X gene using strepsirhine-specific X primers seems particularly promising, but the reliability and validity of this sexing test have not been properly assessed yet.
In this study, we:
(i) show that this molecular sexing test works on three additional lemur species (Microcebus tavaratra, Propithecus coronatus and P. verreauxi) from two previously untested genera and one previously untested family, suggesting that these markers are likely to be universal among lemurs and other strepsirrhines;
(ii) provide the first evidence that this PCR-based sexing test works on degraded DNA obtained from noninvasive samples;
(iii) validate the approach using a large number of known-sex individuals and a multiple-tubes approach, and show that mismatches between the field sex and the final molecular consensus sex occur in less than 10% of all the samples and that most of these mismatches were likely linked to incorrect sex determinations in the field rather than genotyping errors.
I therefore showed that Di Fiore’s sexing test based on the co-amplification of the SRY gene with the amelogenin X gene using strepsirhine-specific X primers appears to be an interesting and reliable molecular sexing test for lemurs (Vanpé et al. 2013 Am J Phys Anthropol).