The greatest threat to the protected Eurasian lynx (Lynx lynx) in Central Europe is human‐induced mortality. As the availability of lynx prey often peaks in human‐modified areas, lynx have to balance successful prey hunting with the risk of encounters with humans. We hypothesized that lynx minimize this risk by adjusting habitat choices to the phases of the day and over seasons. We predicted that (1) due to avoidance of human‐dominated areas during daytime, lynx range use is higher at nighttime, that (2) prey availability drives lynx habitat selection at night, whereas high cover, terrain inaccessibility, and distance to human infrastructure drive habitat selection during the day, and that (3) habitat selection also differs between seasons, with altitude being a dominant factor in winter. To test these hypotheses, we analyzed telemetry data (GPS, VHF) of 10 lynx in the Bohemian Forest Ecosystem (Germany, Czech Republic) between 2005 and 2013 using generalized additive mixed models and considering various predictor variables. Night ranges exceeded day ranges by more than 10%. At night, lynx selected open habitats, such as meadows, which are associated with high ungulate abundance. By contrast, during the day, lynx selected habitats offering dense understorey cover and rugged terrain away from human infrastructure. In summer, land‐cover type greatly shaped lynx habitats, whereas in winter, lynx selected lower altitudes. We concluded that open habitats need to be considered for more realistic habitat models and contribute to future management and conservation (habitat suitability, carrying capacity) of Eurasian lynx in Central Europe.

No abstract

1. Restricting movements to familiar areas should increase individual fitness as it provides animals with information about the spatial distribution of resources and predation risk. While the benefits of familiarity for locating resources have been reported previously, the potential value of familiarity for predation avoidance has been accorded less attention. It has been suggested that familiarity should be beneficial for anti-predator behaviour when direct cues of predation risk are unclear and do not allow prey to identify well-defined spatial refuges. However, to our knowledge, this hypothesis has yet to be tested.2. Here, we assessed how site familiarity, measured as the intensity of use of a given location, is associated with the probability of roe deer Capreolus capreolus being killed by two predators with contrasting hunting tactics, the Eurasian lynx Lynx lynx and human hunters. While risk of human hunting was confined to open habitats, risk of lynx predation was more diffuse, with no clear refuge areas.3. We estimated cause-specific mortality rates in a competing risk framework for 212 GPS-collared roe deer in two ecologically distinct areas of Central Europe to test the hypothesis that the daily risk of being killed by lynx or hunters should be lower in areas of high familiarity.4. We found strong evidence that site familiarity reduces the risk of being predated by lynx, whereas the evidence that the risk of being hunted is linked to site familiarity was weak.5. We suggest that local knowledge about small-scale differences in predation risk and information about efficient escape routes affect an individual's ability to avoid 1330 |

With the launch of the Sentinel-2 satellites, a European capacity has been created to ensure continuity of Landsat and SPOT observations. In contrast to previous sensors, Sentinel-2′s multispectral imager (MSI) incorporates three additional spectral bands in the red-edge (RE) region, which are expected to improve the mapping of vegetation traits. The objective of this study was to compare Sentinel-2 MSI and Landsat-8 OLI data for the estimation of leaf area index (LAI) in temperate, deciduous broadleaf forests. We used hemispherical photography to estimate effective LAI at 36 field plots. We then built and compared simple and multiple linear regression models between field-based LAI and spectral bands and vegetation indices derived from Landsat-8 and Sentinel-2, respectively. Our main findings are that Sentinel-2 predicts LAI with comparable accuracy to Landsat-8. The best Landsat-8 models predicted LAI with a root-mean-square error (RMSE) of 0.877, and the best Sentinel-2 model achieved an RMSE of 0.879. In addition, Sentinel-2′s RE bands and RE-based indices did not improve LAI prediction. Thirdly, LAI models showed a high sensitivity to understory vegetation when tree cover was sparse. According to our findings, Sentinel-2 is capable of delivering data continuity at high temporal resolution.