Nationalparkverwaltung Bayerischer Wald
An integrated approach to model an ungulate population based on telemetry, genotyping , camera trapping and aerial survey data in the Bohemian Forest Ecosystem Henrich, M.
.5 R. ,
5 C. , Dupont,
5 P. ,
Bavarian Forest National Park, Freyunger Straße 2, 94481 Grafenau, Germany 2 University of Freiburg, Wildlife Ecology and Management, Tennenbacher Straße 4, 79106 Freiburg, Germany 3 Šumava National Park, 1. Máje 260/19, Vimperk II, 385 01 Vimperk, Czech Republic 4 Bavarian State Institute of Forestry, Hans-Carl-von-Carlowitz-Platz 1, 85354 Freising, Germany Norwegian University of Life Sciences, Department of Ecology and Natural Resource Management, NO-1430 Ås, Norway
Email: [email protected]
Estimates of red deer population densities in the Bohemian Forest Ecosystem have mainly been obtained by counts at feeding stations and in winter enclosures, in addition to nonstandardized indicators like browsing and hunting statistics. However, the more common occurrence of mild winters in the future due to climate change renders this method increasingly unreliable. There is considerable interest in establishing an appropriate methodology for long-term monitoring, which would enable the study of environmental variables’ effects on the population. The population in the study area is subject to a variety of natural and anthropogenic influences in space and time: • Strict forest reserves without human intervention vs. regular forestry and hunting • Natural disturbances by wind throw and bark beetle • Variation in the intensity of recreational activities • Reduction in the duration of snow cover and depth • Recovery of wolves The question is what the current population density, space use, sex ratio and reproduction rate are, and how these parameters might develop under different scenarios. Fig. 1: Effects of sampling coverage on parameter bias and standard deviation of abundance (N) and the scale parameter of the detection function (sigma). Red lines show the value of the parameters used during simulations. Violins and white dots indicate the shape and mean of posterior parameter distributions, respectively.
Since the area of interest encompasses 702 km², field work efforts had to be reduced to a manageable level while still delivering reliable results. Simulations of different study area coverages ranging from 20-100 % were run based on a simulated deer density of 1 per km² and detection probabilities derived from a previous study in Karwendel. Spatially explicit capture-recapture (SECR) models with a binomial observation component were fitted to each of 50 data sets that were created for each proportion. Based on the depicted (Fig. 1) decrease in precision and increase in bias, we chose a minimum coverage of 80 % of the area.
A grid with a resolution of 1 km² was superimposed on the study area, from which 155 cells were discarded partly randomly, partly due to inaccessibility. Faeces were collected in each of the remaining grid cells during systematic searches over a period of five weeks in summer 2018. In addition, camera traps were placed in the centre of randomly selected grid cells on the German side and a part of Šumava National Park. In subareas of the entire study area , thermal imaging by car and plane was used to detect red deer. Cells with high faeces densities were found more often in the Bavarian Forest National Park compared to the state forest Neureichenau and tended to be close to the state border. Search track
Fig. 2: Positions of grid cells, camera traps, subareas for the aerial survey and distance sampling tracks in the Bohemian Forest. The colour of the grid cells indicates the amount of faeces found there (data from Šumava National Park not yet available). As an example a search track is shown for G499 in the zoomed in panel.
Fig. 3: 55 female red deer were radio collared across the whole study area in 2018. Telemetry data is also available from 2002 to 2014. • Space use • Speed estimation
Fig. 4: 250 camera traps will collect data over a period of one year. • Population estimate via random encounter modelling (REM) • Sex ratio • Cow-calf ratio • Behaviour
Fig. 5: 3320 faeces samples were collected over an area of 547 km². • Identification of individuals • Population estimate via spatially explicit capture-recapture modelling (SECR) • Sex ratio
Fig. 6: Thermal imaging was used to detect deer during flights with an ultralight aircraft over 57.92 km². • Number and group size of animals • Validating detection probability with collared deer
Fig. 7: Distance sampling was done based on thermal imaging by car on transects with a total length of 105 km. • Population estimate via detection functions • Group size
Fig. 8: In spring a browsing survey was conducted in the study area. • Browsing pressure in comparison to population density estimates