Our results prove the number of independent lynx in the BBF has slightly increased over the past decade with a positive population growth rate. Survival rates were high and per capita recruitment was low indicating a low yearly population turnover. Reproductive parameters showed successful reproduction occurring every year, confirming the area represents a key source for the surrounding landscape.
Compared to results of closed population SCR studies on lynx conducted across Europe, density estimates of both open (0.68–1.31 individuals/100 km²) and closed population SCR models (1.02–2.39 individuals/100 km²) are among the highest. Gimenez et al. (2019) camera trapped lynx in the French Jura and Vosges Mountains between 2011 and 2016 and found densities ranging from 0.24 to 0.91 individuals/100 km2. Pesenti and Zimmermann (2013) conducted a camera trapping survey in winter 2007–2008 and 2009–2010 in the north-western Swiss Alps estimating densities of 1.47 and 1.38 individuals/100 km² respectively. In the southern Jura Mountains from December 2017 to February 2018, 42 found a density of 3.48 individuals/100 km² of suitable habitat as the highest across all reference areas in Switzerland.
The yearly baseline detection probability g0 was equal between sexes (Table 2). Although males are generally more active patrolling larger home ranges, females with kittens hunt at a higher rate resulting in augmented activity 43. We found a higher yearly detection function scale σ for males (Table 2), corresponding to larger home ranges, which in line with, e.g., Schmidt et al. (1997) and common in felids (e.g. Satter et al., 2019) with however no significant difference between sexes.
Open population SCR models in a Bayesian framework allowed us to separate survival from emigration and recruitment from immigration. Although this approach is still under development due to large camera trapping array requirements 37, it is reasonable to assume that true survival is at least as high as the value we estimated when the underestimation due to the inclusion of emigration is considered. Our combined yearly survival rates for subadults and adults (population mean = 85%) are among the highest observed for lynx in Central Europe. This is because our estimates come from strictly protected areas where animals have a higher chance to survive. However, the only values available for comparison are from telemetry studies, which are generally conducted with smaller sample sizes because of high-cost devices but allow to track animals’ fates. In Switzerland, Breitenmoser-Würsten et al. (2007) found an overall survival rate of 76% for adults and 53.3% for subadults. In Poland, the combined survival rate of subadults and adults was 63% 44. In three Scandinavian study sites, Andrén et al. (2002) found higher survival rates for subadults (70, 77 and 71%) and adults (87, 91 and 84%), likely due to lower human-related mortality throughout adulthood. Against our expectations, the survival rate was higher for males. This was due to a higher overall number of male apparent survival events, which consist of the number of consecutive detections over years including gaps during which the animal was alive but eluded detection. However, we did not find any strong difference in survival rates between sexes. A similar open population SCR study, conducted on a low-density ocelot population in Belize, also showed no significant differences in sex-specific survival, despite a survival rate of 0.86 for females and 0.78 for males (Satter et al. 2019). The authors suggested the statistical power was not enough to detect significant differences between sexes in these parameters, although they were able to determine sex for a large number of adult ocelots (n = 322).
Camera trapping does not allow assessment of the fates of all disappearing individuals. While our findings show increasing vehicle collisions in recent years (Table 1), we could not assess the actual impact of illegal killing since carcasses are seldom found and it only occurs outside the protected area, as confirmed by the high survival rates found inside it. Illegal killing still represents one of the main lynx mortality causes Europewide, responsible for 32% of known lynx death in Switzerland 5, 46% in Scandinavia 45 and 71% in Poland 44. In the vicinity of our study area, a high poaching rate has been suggested by a modelling approach 4.
We did not detect a significant difference in per capita recruitment between sexes (Table 2), similar again to Satter et al. (2019). Our results of generation time (2.64 years) and the average litter size of 1.97 (range 1–3) were close to those of other studies in Europe. It was proposed based on data of the Scandinavian population that lynx displays a common optimal litter size related to fitness rather than ecological conditions 18. In four areas of Scandinavia, Nilsen et al. (2012) investigated reproductive parameters and found a higher proportion of reproductive females ≥ 3 years of age than females of just 2 years of age, while average litter size was not significantly different (mean across sites = 2.10). Similar results were found in the south-central Scandinavian Peninsula where the yearly reproduction probability was 0.81 for ≥ 3 years old females and 0.77 for 2 years old females, while average litter size was 2.32 (range 1–4) 47. In Switzerland, the average litter size was 2.00 5. We were not able to assess the proportion of reproductive females because sex was not determined for all detected individuals.
The maximum age observed in the study area was 10 years, which was constrained by the study duration. In Switzerland, Breitenmoser-Würsten et al. (2007) found individuals 14–15 years of age through telemetry. Age cannot always be determined through camera trapping. As such, individuals already classified as independent in the first session reached higher values than those we observed, even if their exact age remained unknown.
According to the RAI results, roe deer abundance on the German side remained stable despite the ban on population control of this species in 2012. Contrastingly, the red deer population increased (Fig. 3). This is consistent with the top-down limitation of roe deer abundance by lynx in the study site (e.g. Heurich et al., 2012), with little influence on that of red deer. This finding might indicate that a higher consumption rate of roe deer was not possible in the area. However, especially male lynx might take advantage of increasing red deer numbers as they hunt calves and yearlings of that species 26. The observed abundance decrease in red fox could be due to intraguild predation as it coincides with increasing in lynx numbers, as reported in Scandinavia (e.g. Elmhagen and Rushton, 2007). However, considering that in the study area the red fox is only 1% of lynx kills 26, other factors such as food availability and diseases are likely driving red fox dynamics. Although RAI can be affected by biases attributed to changes in detection probability 50, we believe this method well represented the abundance trend of the species in question since the study design was uniform throughout the monitoring period.
For lynx, we found positive population growth rates, high survival rates and low per capita recruitment resulting in a low yearly population turnover. A stable number of reproducing females and stable prey base indicate carrying capacity for lynx in the study area has already been reached. Therefore, the increasing number of independent lynx most likely comprises dispersing individuals crossing the study area, a few of which settle replacing disappearing resident individuals. This indicates that lynx distribution has changed rather outside the study area.
Contemporary assessments of the BBA lynx population, based on coordinated transboundary camera trapping within an area of 13,000 km2, show a slightly positive population trend and confirm lynx presence and reproduction in the surroundings of the study area 24. In conclusion, the changes we observed could be related to locally decreasing poaching possibly as a result of both law enforcement and intensive long-term campaigns aiming to raise acceptance in the wake of high-profile poaching incidents.
Our study stresses open population SCR models can provide a wider range of demographic parameters useful for lynx monitoring than closed population SCR models. Data gaps affecting the earlier sessions possibly resulted in biased estimates, especially regarding those of closed population SCR models in a maximum likelihood form (Supplementary Table S2). Bayesian methods can better deal with the problem of incomplete detection because they produce posterior distributions of the demographic parameters incorporating the uncertainty resulting from data gaps by using the information deriving from other primary periods 11. This highlights the robustness of this method, which we recommend in future studies and monitoring, especially with incomplete detections. Nonetheless, closed population SCR models still represent a reliable tool for abundance and density assessments (e.g. Gimenez et al., 2019).