Genetic analyses and data quality
After microsatellite PCR screening for sample quality in 2003, we excluded 36% of hair trap samples, 85% of opportunistic hair and 73% of scats. In 2004, we culled 20% of hair trap samples, 76% of transect hair samples, 21% of opportunistic hair samples, 44% of transect scats and 51% of opportunistic scats.
The variability of the selected microsatellite loci was high with Ho = 0·80 and four to eight alleles per locus. PID(sib) at eight loci was low in both years (0·00065 in 2003, 0·0008 in 2004). The PID(sib) threshold for accepting a genotype was 0·033 in 2003 and 0·015 in 2004 (Appendix S1). As PID(sib) observed using the five loci with the lowest discriminatory ability was below 0·033 (2003) and 0·015 (2004), any sample successfully analysed with high confidence at five to eight loci was assigned a genotype.
Average per locus genotyping error rate was 3·7–7·7% for HT, 19·4% (2004 only) for opportunistic hair and 30·2–38·6% for scats. Fifteen per cent of hair samples that gave completed genotypes after one amplification were reamplified; all matched the original genotype. There were no one or two mismatch pairs in the eight-locus reference genotypes, indicating low probability of any undetected genotyping errors (Fig. S1). Multiple observations of a genotype were also taken as a measure of genotyping reliability. Individuals were detected from 1 to 59 samples in 2003 and 4 to 52 samples in 2004. The three genotypes sampled only once in 2003 were resampled in 2004.
All sex ID PCR replicates gave consistent results or matched the known sex based on field data. Forty of 181 samples that failed microsatellite genotyping in 2003 and 91 of 173 in 2004 were identified as bear samples after mtDNA fragment analysis.
Evaluation of NGS methods
In 2003, 363 samples were collected. HT provided the highest sample numbers (63%), and OP obtained similar numbers of hair and scat (Table 2a). In 2004, 801 samples were collected; HT samples were the most numerous (60%; Table 2b). TR provided equal quantities of hair and scat, whereas the number of OP faecal samples was double the number of hair samples (Table 2b).
Table 2. Results by sampling method and sample type for (a) 2003 and (b) 2004
| ||Samples collected||Samples analysed||Unique bears identified||Genotyping success|
|HT||227|| ||227||214|| ||214||8|| ||8||124|| ||124|
|HT||480|| ||480||229|| ||229||13|| ||13||181|| ||181|
Nine bears were identified in 2003 and 15 in 2004 (Table 2, Fig. 2). No single method detected all bears. Most bears (6 in 2003, 12 in 2004) were detected by both HT and OP, and in 2004 results from these methods combined identified all individuals that were sampled; in contrast, TR detected only two bears. With OP, more bears were identified from scat than from hair. HT and OP detected bears of all age and sex classes (Fig. 2). In 2003, one cub was sampled with HT; in 2004 a litter of two cubs was identified with opportunistic scats and a litter of three cubs with HT and opportunistic scats. The parents of all offspring were identified. All individuals sampled corresponded to translocated bears and their progeny, therefore remnant bears of the former population are assumed to be dead and no immigrants were detected. In 2004, 18 samples were collected at 10 damage sites. At eight sites, we documented the presence of a brown bear and, at six of these sites, the individual responsible for the damage was identified. Two of the bears identified were detected at >1 site.
Figure 2. Detection frequency of brown bears possibly present in the study area in (a) 2003 and (b) 2004 using hair traps (HT), opportunistic hair (OPh) and scat (OPs), transect hair (TRh) and scat (TRs). Age class and sex are indicated as follows: am, adult male; af, adult female; year 1-f, 1-year-old female; year 2-f, 2-year-old female; cub m, male cub; cub f, female cub.
Download figure to PowerPoint
In 2003, HT genotyping success was higher than OP considering hair and faeces combined as well as separately (among three tests the lowest χ2 = 27·15; d.f. = 1; P < 0·0001). In 2004, TR and opportunistic scats produced the lowest genotyping success rate (among eight tests the lowest χ2 = 13·71; d.f. = 1; P = 0·0002); HT yielded higher success rates than OP considering hair and faeces combined (χ2 = 25·82; d.f. = 1; P < 0·0001), but HT success was not significantly higher than opportunistic hair alone (Table 2b).
HT had lower error rates than faeces in 2003 (χ2 = 190·12; d.f. = 1; P < 0·0001) and OP in 2004 considering hair and faeces combined as well as separately (among three tests the lowest χ2 = 45·79; d.f. = 1; P < 0·0001). Among OP samples from 2004, hair produced a lower error rate than scat (χ2 = 13·75; d.f. = 1; P = 0·0002). In HT samples from 2004, one mixed sample was identified by more than two alleles at multiple loci.
Detection frequency varied greatly among individuals and with different techniques (Fig. 2). Using HT, individuals were detected during one to six sessions in 2003 (mean 3·1, SD 2·2) and one to seven in 2004 (mean 3·4, SD 1·9). Capture frequency was lowest (1–2 sessions) for TR in 2004. Combining opportunistic hair and faecal detections increased detection frequencies and bears were identified one to four different times (mean 2, SD 1·2) in 2003 and one to nine times (mean 3·6, SD 2·5) in 2004.
In 2003, bear samples collected at HT were found in 38·4% of grid cells, on the eastern side of the Brenta range (Fig. 1c), and approximate locations of bear samples collected opportunistically had a similar distribution. Individuals were sampled at one to seven trap sites, and at one to four locations with OP. In 2004, bears were sampled on both sides of the Brenta range (Fig. 1d,e,f). Overall, HT and OP provided the greatest spatial coverage and were the most effective at detecting bear movement. In 2004, bear samples were found in 53·6% of the grid cells with HT and along 8 of 17 transects. The number of locations where individual bears were detected ranged from 1 to 12 for HT, 1 to 8 for OP and 2 for TR. In general, HT and OP locations of bears detected with both methods were overlapping and complementary (Fig. 3).
Figure 3. Spatial distribution of multiple captures in 2004 for specific individuals: (a) adult male (am 2); (b) two adult females (af 3, af 5); (c) yearling female (1-year f) and male cub (cub m3). *, hair traps; , transects; •, opportunistic.
Download figure to PowerPoint
For HT, the number of samples collected, individuals and new bears identified varied greatly during the sampling period (Table S1a). Fewer bears and fewer new bears were detected in the later sessions. For example, no new individuals were sampled after session IV (July) in 2003 and session V (August) in 2004. Genotyping success varied with no clear trend but was generally lowest in the last sessions.
The temporal distribution of TR was not uniform over the surveyed area, but the number of samples collected and genotyping success were low in all sessions (Table S1b). No bears were identified after August.
OP was partitioned into three main seasons of bear activity (spring, summer and autumn; Table S1c,d). Hair sampling provided the highest sample numbers in the summer. Hair genotyping success was always highly variable, but generally was higher in the spring. More individuals were identified from hair during spring and no new bears were identified after August. Scat samples were more numerous in spring and autumn; scat genotyping success varied between years but was slightly higher in autumn than during other seasons. The number of individuals and new individuals identified was the lowest in the summer.
TR was not included in the population estimation analysis due to low recaptures. To maximize closure assumptions, we considered the first five HT sessions, and the sampling period May–August for OP (minimum count dropped to 9). The likelihood ratio test in capwire indicated individual heterogeneity (P = 0·002) in OP data, and the two innate rates model was used. capwire population size estimate for 2004 OP data was 13 (95% CI: 9–18). Supported models in mark (AICc ≤2) (Burnham & Anderson 2002) for HT did not indicate time variation in capture probability or behavioural response, and included the effect on capture probability of whether a bear was translocated or born in the area (Table 3). The estimate of capture probability from the top-ranked model was 0·54 (SE 0·069) and the model-averaged population size estimate was 14 (95% CI: 13–18).
Table 3. Model selection results from Huggins closed capture models in mark for one to five hair trapping sessions in 2004
|Model||AICc||ΔAICc||AICc weight||Model likelihood||No. parameters||Deviance|
In 2004, c. €71,000 was spent to implement hair and faecal sampling and to perform genetic analyses (Table 4). HT required the most resources in the field and in the laboratory. Field costs were higher than laboratory costs, except for OP. OP provided the lowest cost/genotyped sample (€125), cost/unique bear (€645) and cost/bear sample (€64), whereas TR costs were the highest. Within OP, hair had lower cost/genotype and cost/unique bear than scat, whereas cost/bear sample was slightly higher for hair than scat. The opposite trend of hair and scat costs was observed for TR.
Table 4. Costs (€) per sampling method estimated in 2004
| ||Field||Lab||Total||Cost/genotype||Cost/unique bear||Cost/bear sample|
|HT||31 417||7803||39 220||217|| ||217||3017|| ||3017||192|| ||192|
|TR||18 635||4180||22 815||6313||4365||2852||9470||7275||7605||997||1212||617|