CSIRO PUBLISHING
www.publish.csiro.au/journals/wr
Wildlife Research, 2009, 36, 721–736
Field-based evaluation of scat DNA methods to estimate population abundance of the spotted-tailed quoll (Dasyurus maculatus), a rare Australian marsupial Monica Ruibal A,E, Rod Peakall A, Andrew Claridge B,C and Karen Firestone D A
Evolution, Ecology and Genetics, Research School of Biology, The Australian National University, Canberra, ACT 0200, Australia. B Department of Environment and Climate Change, Parks and Wildlife Group, Planning and Performance Unit, Southern Branch, Queanbeyan, NSW 2620, Australia. C School of Physical, Environmental and Mathematical Sciences, University of New South Wales, Australian Defence Force Academy, Northcott Drive, Canberra, ACT 2620, Australia. D School of Biological, Earth and Environmental Sciences, University of New South Wales, Sydney, NSW 2052, Australia. E Corresponding author. Email:
[email protected]
Abstract Context. DNA extracted non-invasively from remotely collected scat samples has been used successfully to enumerate populations of a few endangered mammal species. However, scat DNA surveys relying on scent-marking behaviours need to identify if age- or sex-specific variations or seasonal changes in scat scent-marking patterns affect population estimates. Furthermore, owing to the low quantity and quality of scat DNA, a thorough assessment of the technique is needed when it is applied to different species to ensure that individual identification is reliable. Aims. In the current study, microsatellite genetic profiles derived from 208 remotely collected scats of the spottedtailed quoll (Dasyurus maculatus), a rare Australian marsupial carnivore, were compared with DNA profiles from tissue of 22 live-trapped individuals from the same study area to critically assess the reliability of the non-invasive method to estimate population abundance. Methods. Scat samples were collected at scent-marking sites over 4 consecutive months (April–July 2005), 7 weeks of which overlapped with the trapping program to allow direct comparisons of population estimates. Key results. Combining a multiple-tubes approach with error checking analyses provided reliable genetic tags and resulted in the detection of the majority of the live-trapped population (18 of 22 individuals). Ten additional individuals not known from trapping were also observed from scat DNA. A longer-term sampling regime was required for scats than for trapping to allow direct detection of a large proportion of the population and to provide a comparable population estimate. Critically, the 4-month scat collection period highlighted the importance of performing scat surveys during the mating season when scat scent marking is more frequent, and to avoid sex and age biases in scat marking patterns. Implications. Non-invasive scat DNA sampling methods that rely on scent-marking behaviours need to consider the duration of the sampling period and temporal differences in behaviours by the sexes and age groups to ensure that meaningful population estimates are achieved.
Introduction Conservation management initiatives for endangered species are often hampered by the difficulty in determining the size and ongoing status of remnant populations (Lindenmayer and Franklin 2002; Molina and Marcot 2007). Without knowledge of the changing state of such populations, management decisions and conservation efforts cannot necessarily be targeted optimally. The difficulty in enumerating populations of endangered mammal species is largely due to their rarity, exacerbated in most cases by the elusive or cryptic nature of such species (Creel et al. 2003). CSIRO 2009
This makes traditional methods of surveying populations often impractical and/or cost prohibitive (e.g. brown bears, Ursus arctos, Solberg et al. 2006). Over the past decade, several non-invasive DNA studies have concluded that scats or faecal pellets provide a viable DNA source to identify individuals and census populations. The advantages of using such materials are that samples can be collected remotely without the need to directly capture or observe individuals and that abundance estimates can be achieved over suitably short sampling periods of several consecutive days to tens of days (e.g. European badger, Meles meles (Frantz et al. 2003; 10.1071/WR09086
1035-3712/09/080721
722
Wildlife Research
Wilson et al. 2003); forest elephant, Loxodonta cyclotis (Eggert et al. 2003); brush-tailed rock-wallaby, Petrogale penicillata (Piggott et al. 2006); coyote, Canis latrans (Kohn et al. 1999); and wombats, Vombatus ursinus (Banks et al. 2002) and Lasiorhinus krefftii (Banks et al. 2003)). Despite the promise of non-invasive DNA sampling there are several pitfalls that need to be assessed in each new study species before the method can be used as a management tool. First, a critical concern is the risk of genotyping errors, principally allelic dropout and false alleles, which lead to mis-identification and biased population estimates (Taberlet and Luikart 1999; Waits and Leberg 2000; Piggott and Taylor 2003; Waits 2004; Dewoody et al. 2006). An effective approach employed to identify and minimise genotyping errors is to apply multiple independent PCR replicates of the samples at each locus (the multiple-tubes approach of Navidi et al. 1992; Taberlet et al. 1996; Frantz et al. 2003). However, this approach can be prohibitive when the number of samples is great or if the DNA source is prone to high genotyping errors because many replicates are needed to be confident of the resulting data (see Taberlet et al. 1996). Second, if recently deposited scat samples are needed to thwart high genotyping error rates in older samples, then the extended collection periods necessary to achieve accurate population estimates can make field collection labour-intensive and time-consuming. Thus, it is important to evaluate whether the non-invasive DNA method can promptly detect individuals within the population. When scat sampling is reliant on scent-marking behaviours it is also important to identify if there are age- or sex-specific variations or seasonal changes in scent-marking patterns (Dallas et al. 2003; Smith et al. 2006). For example, the rate at which individuals scent mark is predicted to increase when the function of the behaviour is associated with life history phases such as reproduction (Kruuk and Jarman 1995). Furthermore, under a reproductive function, scent-marking rates are predicted to increase among sexually mature individuals (Stoddart 1976). Depending on the species, this may be limited to reproductively mature or dominant individuals within one or both sexes, as is the case for male and female grey wolves (Canis lupus; Ryon and Brown 1990). Also, there may be differences in the type of scent marks that are deposited for each sex (Stoddart 1976). For instance, adult male honey badgers (Mellivora capensis) scent mark at latrines with faeces more often than do young males, and adult females rarely defecate at latrines (Begg et al. 2003). Consequently, a proportion of the population of this species would not be detected in a scat DNA study. Moreover, ageor sex-specific variations in the pattern and rate of deposition would lead to heterogeneous detection probabilities, which negatively bias population abundance estimates in mark–recapture studies (Otis et al. 1978). For these reasons it is imperative to ascertain if individuals of all ages and each sex can be detected from the scent-mark source and at different times of the year before non-invasive DNA methods are used to estimate population abundances. Application of non-invasive scat DNA sampling in the spotted-tailed quoll The spotted-tailed quoll (Dasyurus maculatus) is a medium-sized (up to 7 kg), forest-dependant marsupial carnivore endemic to
M. Ruibal et al.
eastern Australia (Edgar and Belcher 1995). This species was listed under Australian Commonwealth legislation in 2004 as an endangered species (Environment Protection and Biodiversity Conservation Act 1999) because of its significant range contraction since European settlement and more recent evidence that localised extinctions are continuing (DEH 2004). However, the extent and rate of the current decline is not known because of the difficulty in conducting broad-scale surveys that provide quantitative data on this cryptic animal (Long and Nelson 2008). Presently, the most effective method for surveying spotted-tailed quoll populations is live capture through cage trapping. However, the implementation of cage trapping is resource-intensive (Watt 1993; Jones and Rose 1996; SEFSTQWG 2003). Owing to the large size of the home ranges of this species (males, 363–2907 ha; females, 133–862 ha) (Belcher 2000; Belcher and Darrant 2004; Andrew 2005; Claridge et al. 2005; Glen and Dickman 2006), the survey effort needs to be broad in scale and prolonged to increase the likelihood of encountering most individuals (Long and Nelson 2008). Consequently, in the more rugged parts of this species range and where vehicle access is limited, live trapping is not feasible. Therefore, there is a need for alternative methods that achieve robust estimates efficiently over large geographic areas. Scat sampling potentially provides a practical method to survey spotted-tailed quoll populations owing to this species’ habit of defecating on prominent landscape features and the distinct shape of quoll scats compared with those of other predator species (Triggs 2004). Accumulations of quoll scats are commonly found on rock substrates such as large complex rock outcrops, slabs of exposed bedrock along riverine habitat, and at the tops and bottoms of rocky cliff lines (Alexander 1980; Kruuk and Jarman 1995; Belcher 2000; Dawson et al. 2002). These defecation sites, or latrines, are thought to be communal scent-marking sites (Kruuk and Jarman 1995; Belcher 2000; Dawson et al. 2002). However, it is not known if individuals of all ages and of each sex defecate at latrines. Observations of spotted-tailed quolls at latrines provided by two remote photographic camera studies (Belcher 1994; Claridge et al. 2004) demonstrated that not all visitations result in defecation, and observed that other scents in the form of urine and cloacal dragging are also used to scent mark. Thus it is imperative to investigate patterns of scat deposition by age and sex to ensure that all individuals can be encountered via the scent-marking habits of the species. The overall aim of this study was to evaluate the efficacy of scat DNA sampling to estimate population abundance for the spotted-tailed quoll. First, we evaluated the noninvasive approach for genotyping error rates (allelic dropout and false alleles), with subsequent analyses of the effectiveness of the multiple-tubes approach to eliminate any genotyping error. Second, following identification of individuals and matching of scat genotypes to trapped individuals, we assessed if age or sex biases in defecation rates impacts on population size estimates. Third, we assessed the potential of the method to provide meaningful estimates of population abundance in a survey period that is logistically feasible.
Remote census using scat DNA analysis in quolls
Wildlife Research
Materials and methods Study site and sampling methods The trapping and scat collection program covered ~7000 ha of the Byadbo Wilderness Area within the Kosciuszko National Park, south-eastern NSW, Australia (Fig. 1). A detailed description of the study site and the trapping program are provided in Claridge and Mills (2007). In brief, trapped individuals were permanently marked for unique identification with a Passive Integrated Transponder (PIT; Trovan Microchips Australia, Melbourne) and weighed to estimate age (sub-adult ~1 year old: , < 1500 g, < < 2000 g; adults 2+ years old: , > 1500 g, < > 2000 g; as per Dawson 2005). Ear tissue
N.S.W N.S.W
biopsies were taken for genetic analysis and used as reference profiles for the scat DNA genotypes. Ear tissue was preserved in 70% ethanol until DNA extraction. Quoll scats were collected from 20 trapping sites where active defecation sites (hereafter referred to as latrines) occurred. Sixteen other trapping sites were not included because four sites did not have rocky areas, ten latrine sites were inactive and two active latrine sites were not monitored. Scats were also collected from 10 latrine sites where traps had not been set. At each latrine site numerous boulders and/or bedrock areas were examined for scat deposition, with each single rock feature being considered a separate latrine rock within the latrine site. In total, 75 single latrine rock features were monitored for scats. Each latrine site was checked daily or every second day from April to July 2005. Only half of each scat was collected to minimise disturbance to the scent-marking habits of this species; the remaining half was skewered with a toothpick to prevent re-sampling and left in situ. Scats were placed individually in plastic zip-lock bags and then wrapped in aluminium foil before preserving the scats in the vapour phase of liquid nitrogen. Scats were preserved from the day of collection and during transportation (up to 2 weeks). In the laboratory, scat samples were stored at 20C until DNA extraction (up to 4 months later).
r ive
sR
In
ob
ge
Jac
bir
ah
Cr
ee
k
Th
atc
he
rs
M
ou
nta
in
Cr
ee
k
Study area
723
ns lto
Fe k
ee Cr
Sno wy
Riv er
N
0
0.5
1 Kilometers
Fig. 1. Map of locations where live-capture cage trapping and scat sampling were conducted between April and July 2005. Open squares represent sites where trapping and scat sampling were conducted; open squares with crosses represent sites where traps were set; and closed squares represent sites where only scat sampling was conducted. Grid references are based on Australian Geodetic Datum 66 (Zone 55).
Isolation and polymerase chain reaction amplification of DNA DNA was extracted from the ear tissue of live-captured quolls via the ammonium acetate extraction method of Bruford et al. (1992). Epithelial cells on the surface of scats were targeted for DNA extraction by swabbing the exterior surface of each scat using a sterile cotton bud soaked in the lysis buffer SLP (500 mM Tris-HCl pH 9.0, 50 mM EDTA, 10 mM NaCl; Deuter et al. 1995). The cotton bud was then vigorously rotated in a 1.5-mL tube containing 1 mL of SLP buffer. These steps were performed several times until the shiny exterior evident on the scat was removed. We then followed the extraction protocol outlined in Piggott and Taylor (2003) under ‘Method A Qiagen spin columns purification’. The DNA was eluted in 100 mL of AE buffer. Tissue and scat samples were genotyped at 10 microsatellite loci using redesigned primers based on the dasyurid primers of Firestone (1999) (Q1.3, Q3.3.1, Q3.3.2, Q3.1.2, Q4.4.2, Q4.4.10) and Spencer et al. (2007) (Dg1A1, Dg1H3, Dg6D5, Dg5G4) (see Appendix 1). The multiplex pre-amplification PCR method of Piggott et al. (2004) was applied to genotype scat samples. PCR reaction mixtures for each PCR step and DNA source are described in Appendices 2 and 3. Allele sizes were calibrated using Genescan-500 LIZ size standard (Applied Biosystems, Foster City, CA) with each sample. Fragment sizing was completed using GENEMAPPER software, ver. 3.7 (Applied Biosystems). Assigning the multilocus genotypes of scat samples The genetic profiles of DNA obtained from scats were validated in four stages, briefly described here (full details in Ruibal 2008). Stage one consisted of a sequential multiple-tubes approach based on multiple PCRs per sample and locus to verify genotype
724
Wildlife Research
profiles (see Appendix 4). In Stage Two the scat profiles were compared with each other and to the reference profiles obtained from ear tissue samples using a multilocus pairwise matches option in the program GENALEX 6 (Peakall and Smouse 2006). Samples that matched an ear tissue profile were assigned the microchip code of the individual. Scat genotypes unknown from the trapped population were referred to as unknowns. In Stage Three the software program DROPOUT (McKelvey and Schwartz 2005) was used to evaluate the prevalence of genotyping errors of unknown genotypes from the trapped population that may have mistakenly arose from single allelic mismatches using examining bimodality (EB) and difference in capture history (DCH) tests. During Stage Four the matching approach of Creel et al. (2003), which proposes that samples matching all but a single allele be assigned the same identity, was adopted post hoc to resolve the identity of nine samples that could not be resolved with PCR replication. This was considered a reasonable approach to eliminate overestimation biases of population abundance because the probability of individuals (among the trapped population) differing at two or more loci was high (see ‘Probability of identity’, below). Quantifying the overall genotyping error rate Once the scat genotypes had been assigned an identity, the frequency of allelic dropout and false alleles for the first two replicate amplifications was estimated for each microsatellite locus. GIMLET ver. 3.2 (Valiere 2002) was used to determine the total number of positive amplifications and genotyping errors observed per locus across all samples. The frequencies were then used to estimate locus-specific rates of allelic dropout and false alleles using the equations ADO1 and FA3 from Broquet and Petit (2004). Probability of identity Probability of identity equations for unrelated (PI) and related individuals (PIsibs) were used to estimate the average probability that two individuals would share the same genotype by chance alone for an increasing number of loci and over all 10 microsatellite loci (Taberlet and Luikart 1999; Waits 2001). PI and PIsibs estimates were calculated among individuals identified via trapping and scat sampling using GENALEX 6 following the formula provided in Peakall et al. (2006). The observed rate at which unique identity was discriminated among the trapped individuals and for the scat genotypes was compared with the theoretical predictions (PI and PIsibs). The observed PI rate was calculated as the proportion of all possible individual pairs with unique genotypes for each increase in the number of loci from 1 to 10. The individual pairwise genetic distance option available in GENALEX 6 was used to determine the rate of unique identity among all possible pairs for each dataset. Pairs of individuals with a genetic distance of zero were denoted as identical genotypes and any pairs with a genetic distance of 1 were denoted as unique genotypes. These values were then used to calculate the frequency of unique identity across all possible pairs of individuals for an
M. Ruibal et al.
increasing number of loci. Loci were arranged according to rates of PCR amplification success (high to low) as observed for the scat genotypes. Population abundance estimates To permit direct comparisons of abundance estimates between scat DNA data and trapping data, the scat dataset was restricted temporally to May and July to match the months when trapping was conducted. The resultant dataset is referred to here as the scat complement data. Population estimates were also derived for monthly scat collection data using statistical estimators. Model estimates were appraised under the assumption that the direct count of trapped individuals was the minimum number of individuals known to be alive (referred to here as MIN ktba). Closed population models were selected on an a priori basis because the fundamental assumptions of demographic (births and deaths) and geographic (migration in or out) closure were reasonable for this species. Despite the relatively long sampling period (maximum 8 weeks), births did not affect assumptions of closure at the time of year sampling was conducted because young are born at the end of July and remain in the pouch or den for some time beyond the completion of this study (Ruibal 2008); the chance of deaths occurring are likely to be minimal; and migration in or out would not be extensive because individuals are expected to have established home ranges. Although there were no barriers to prevent temporary movements outside the surveyed area, the risk of violating geographic closure was minimised; to ensure a reasonable chance of sampling individuals at multiple trap and latrine sites, sampling sites were spaced at distances (150–8500 m) encompassing the maximum straight-line distances known to be moved by male (2529–4430 m) and female (1865–3085 m) spotted-tailed quolls in the study area over 24 to 48 h periods (Claridge et al. 2005). Estimates of population abundance were derived for the trapping data and the scat complement data using two approaches that accounts for different features of non-invasive DNA sampling. First, as well as using conventional closed capture–mark–recapture (CMR) models (M(o), M(t), M(b), M(h): see Table 4) of the program MARK vers. 4.3 (White and Burnham 1999), we fitted the models with the addition of a misidentification parameter to mitigate the risk that individuals encountered from a single scat sample may not have been correctly identified (Lukacs 2005; Lukacs and Burnham 2005). The second approach, use of the program CAPWIRE (Miller et al. 2005), allowed the use of multiple detections of individuals within a single sampling session (i.e. per collection day) and account for heterogeneous scat deposition patterns among individuals. The fit of the models within MARK was evaluated using the Akaike Information Criterion (AICc, corrected for sample size). The model with the lowest AICc was regarded as the most parsimonious model to estimate population size. Models with DAICc (difference between AICc values) less than two and which had AICc scores close to that of the most parsimonious model were also considered parsimonious models (Burnham and Anderson 1998). When more than one model fitted a dataset
Remote census using scat DNA analysis in quolls
we weighted the estimates by model averaging over all parsimonious models. Data analyses of detection rates for sex and age classes To assess if there were age or sex biases in scat deposition rates, generalised linear models (McCullagh and Nelder 1989) assuming a Bernoulli distribution with a logit link function (observed = 1, not detected = 0 for each individual for each sampling day) were used to test for differences in the average deposition rate between males and females, and sub-adults and adults. Pseudo-replicate observations of individuals were not included in these analyses. Analyses were conducted for the trapping and scat complement datasets, with comparisons made to test for differences in the proportion of observations made for the sex and age classes. Scat genotypes that did not match to a trapped individual were excluded from the analyses as sex or age was not determined. The statistical analyses were conducted using GENSTAT for Windows, 9th Edition (VSN International, Hemel Hempstead, UK). The statistical significance of each variable was tested using deviance ratios (McCullagh and Nelder 1989).
Wildlife Research
725
A bimodal distribution in the minimum number of locus differences was produced for the EB test for eight samples with an allelic mismatch and the single false heterozygote genotype. The DCH test determined that five loci increased the number of new individuals faster than would be expected, with the allelic mismatch present in a total of 11 individuals. In all but two cases, this result corresponded to nine samples and loci with the one allelic mismatch, suggesting that genotyping errors were present in the data. Excluding the mismatched alleles in the nine profiles changed the distribution to that expected when the degree of error is low. Hence, the samples were assigned the identity to which they matched except for a single allele. Scat genotype identity assignment and probability of identity of individuals
From April to July 2005, 262 scat samples were collected. A total of 208 (79%) samples contained amplifiable DNA and were assigned a genotype and identity. Under the dichotomous rules of the multiple-tubes approach, DNA from 81 (39%) of the scat samples amplified across all 10 microsatellite loci, 68 samples (33%) amplified at 8 or 9 loci and 39 samples amplified at 7 loci. The remaining 20 samples only amplified once for 7 loci. These 20 samples unambiguously matched a reference profile (i.e. tissue samples or other scat samples) and so were included in the study.
In total, 28 unique genetic profiles were identified over the 4 month scat collection period. Eighteen profiles matched to genetic tags of trapped individuals: all six trapped females (1 sub-adult and 5 adults), 11 of the 15 trapped males (4 subadult and 7 adults) and the sub-adult intersexual individual. Four adult males were not detected through scat DNA analysis. Ten additional and unique genetic tags were observed and were identified as unknown individuals. Over the scat complement sampling period, 19 unique genotypes were detected from 56 scats. Sixteen of these matched to genotypes of trapped individuals (six females, nine males and the intersexual individual) and three belonged to unknown individuals. Two of the unknown genotypes matched profiles of a female and male individual trapped in the year following this study. Both individuals were determined to be sub-adults during this study based on weight and tooth wear (M. Ruibal, unpubl. data). All 28 scat genotypes were uniquely identified with the four most variable loci. The empirical finding that not all loci were required for unique identification was supported by the PI estimates across the full complement of loci of 3.12 106 for the scat genotype population and 4.6 106 for the trapped population, as well as the PIsibs estimate of 0.003 for both datasets (Fig. 2). The observed proportion of individuals with unique identity among the scat genotypes and the tissue samples was close to the predicted theoretical PI estimate (Fig. 2). When the loci were combined according to scat DNA amplification success rates (high to low), eight loci were required for unique identification (data not shown). This difference is a product of greater amplification success rates among the less variable loci which amplify smaller PCR products.
Genotyping error rate for scat samples
Detection rates for sex and age classes
Allelic dropout was observed in 5.4% of the first two amplifications performed across all loci and samples that were designated as heterozygotes. The rate of allelic dropout varied among loci (Table 1). False alleles were rare, with only 23 instances during the first two PCR amplifications across all loci and samples. All but one sample was resolved with further amplifications. This single sample mismatched a reference profile at a single allele (false heterozygote) across the 10 loci. This sample was combined with the genotype it matched against except for a single allele as per the matching approach of Creel et al. (2003).
A significant difference in the trap capture rate was observed among individuals (P < 0.001) and between the sexes (P < 0.001), with females (average s.e. = 7.3 1.5, range: 2–13, n = 6) trapped more often than males (average s.e. = 4.2 1.0, range: 1–16, n = 15). There was also a marginally significant difference between the age classes (P = 0.046), with sub-adults (average s.e. = 5.3 2.1, range: 1–16, n = 7) captured more often than adults (average s.e. = 4.8 0.9, range: 1–13, n = 15) (Table 2). For the scat complement data, a significant difference in the average number of scats detected for each sex was found (P = 0.017)
Results Trapping During May and July 2005, a total of 1563 trap nights resulted in the capture of 22 individual quolls on 109 occasions. Six females, fifteen males and one intersexual individual (with pouch and penis but lacking a scrotum, XXY karyotype; Ruibal 2008) including animals from the youngest possible capture age (approximately9 months, n = 7) tothe oldest(3+ years, n = 15) were captured. DNA from ear tissue samples from the 22 individuals was amplified for all 10 microsatellite loci. All individuals were uniquely identified with the four most variable loci. Microsatellite amplification success of scats
Allelic dropout (heterozygous genotypes only) False alleles
Locus 15 (n = 214) 2 (n = 358)
0 (n = 394)
Dg1H3R
17
(n = 324)
Dg1A1R
6 (n = 286)
9 (n = 191)
Dg5G4
2 (n = 366)
12 (n = 121)
Dg6D5R
0 (n = 395)
5 (n = 119)
Q13R
2 (n = 397)
2 (n = 114)
Q312R
2 (n = 392)
9 (n = 173)
Q331R
1 (n = 238)
16 (n = 116)
Q332R
6 (n = 386)
13 (n = 262)
Q4410R
2 (n = 393)
4 (n = 264)
Q442R
23 (0.6%) (n = 3605)
102 (5.4%) (n = 1898)
Total
Table 1. Summary of the frequency of allelic dropouts and false alleles observed for 208 scat DNA samples, with the total number of amplifications conducted per locus in each category shown in brackets
726 Wildlife Research M. Ruibal et al.
Remote census using scat DNA analysis in quolls
Wildlife Research
Theoretical PIsibs
Observed PI - tissue
Observed PI - scat
Probability of identity
Probability of identity
Theoretical PI
727
Number of combined loci
Number of combined loci Fig. 2. The theoretical predictions of the probability of identity for unrelated (PI) and related (PIsibs) individuals calculated across 10 microsatellite loci from tissue DNA of 22 live-trapped individuals, and the observed PI among the trapped individuals and the 28 scat genotypes. Loci were combined from the highest to lowest rates of amplification as observed for the scat samples. The inset figure shows the probability of identity at the range where 99% of the genotypes are expected to be unique.
(Table 2), with males (average s.e. = 3.2 0.8, range: 1–9, n = 10) encountered more often than females (average s.e. = 2.0 0.4, range: 1–4, n = 7). This difference was largely due to the non-detection of females from the first week of July (data not shown). This result suggests that females no longer deposit scats after this time at latrines, even though four females were caught in the trapping program during Table 2. Outcomes of generalised linear regression modelling of encounter rates of each age and sex recorded during trapping and scat DNA surveys Dot operator represents interaction between model terms. An individual identified as an intersexual was excluded from the trapping and scat complement datasets. Approx. F prob. = probability for variance ratios Change
d.f.
+ Week + Day + Sex + Age + Individual + Week.Individual Residual Total
6 21 1 1 18 120 420 587
+ Week + Day + Sex + Age + Individual + Week.Individual
6 21 1 1 14 96
Residual Total
336 475
Deviance
Mean deviance
Trapping 10.77 16.01 9.97 2.74 73.31 157.00 288.10 557.86
1.80 0.76 9.97 2.74 4.07 1.31 0.69 0.95
Scat complement 12.77 2.13 22.11 1.05 2.68 2.68 0.54 0.54 25.53 1.82 82.60 0.86 156.17 302.39
0.47 0.64
Deviance ratio
Approx.
F prob.
2.62 1.11 14.54 3.99 5.94 1.91
0.017 0.332
