Discussion
Knowledge of habitat selection is essential for evaluation of wildlife population density in a heterogeneous landscape. We find that inclusion of habitat heterogeneity is favored in the Spatial Capture-Recapture (SCR) model parameters of density and detection. The SCR model without landscape effects was 25 AIC units higher than the top model including landscape heterogeneity. The estimated densities across the habitats describe second-order habitat selection of the white-tailed deer population studied. Population densities were highest in agricultural areas and mixed forests and lowest in coniferous forests and transitional woodlands. We can integrate this preference over all landcover types in the study area to arrive at a landscape-specific estimate of density. We find that including heterogeneity increases the overall density estimate when compared with density estimated under the best supported homogeneous (constant density) SCR model. The increase in density is because the greatest proportion of the state space consists of agricultural areas (76%), which is also the preferred habitat (together with mixed forests which are, however, relatively uncommon).
In the SCR framework, we estimated habitat-specific densities, while accounting for third- order habitat selection through the effect of habitat on detection probability. The probability to detect an individual is highest in transitional woodland followed by mixed forest and is lowest in coniferous forest. Detection probability also decreases with increasing distance to land in agricultural use. Thus, inside a white-tailed deer’s home range, space is not used equally; the animals prefer to be close to fields and in transitional woodland compared to coniferous forest. In our study area, transitional woodlands are typically clear-cuts and these, as well as fields which are predominantly grain fields (wheat, oat, rye), likely present habitat with good food resources for the white-tailed deer.
About half (52%) of the white-tailed deer in the study area are males. This is an interesting finding, because the age and sex composition of the white-tailed deer is almost entirely shaped by the strong hunting pressure on this species. In general, harvested ungulate populations tend to have more females than males in Nordic countries, which may result in lower viability of the population (Langvatn and Loison 1999, Sæther et al. 2004). Hunting-bag statistics of white-tailed deer in Finland also imply a higher mortality of males and as a result, few males compared to females reach old ages (Kekkonen et al. 2016). However, DNA does not allow distinguishing fawns from adults, and fawns are probably a large fraction of the study population. Because home ranges of fawns are restricted to their mothers’ (Tierson et al. 1985), and because the spatial scale of the study is small and the number of individuals limited, our results may largely reflect fawns and their mothers. Based on the 17 individuals recorded in both study years, we can infer that at least 34% (14/41) of females and 12% (3/25) of males were adults in the second study year. For these reasons, the sex ratio of Finnish white-tailed deer would need more thorough investigation.
Our study area represents a typical Finnish rural area with low coverage of artificial surfaces (covering only 2% of the state space). Here, we did not find any preference or avoidance by white-tailed deer for artificial surfaces, which includes roads and human settlements. We expected that the distance to water, which in the study area were mainly two bigger streams with smaller ditches, would have been an important covariate for white-tailed deer density. Nevertheless, the models including that covariate were not supported by AIC. Presumably, there are other smaller stationary water sources in the area and thus larger streams and ditches are not the crucial source of water for white-tailed deer. It is notable that the habitat preferences later in the year likely differ, especially during winter when food sources are different and limited. In addition, snow cover may affect the availability of food and movement of individuals (Andersson and Koivisto 1980).
Collecting information on space use of individuals based on non-invasive genetic methods and analyzing this information with SCR provides a possibility to study habitat selection of animal populations, as done in this study. A possible advantage of this approach compared to telemetry approaches is that population-wide inferences can be reached at a potential lower cost and with clearly lower risk to the study organism. In addition, multilocus genotype data provides information not only for individual identification and density estimation, but also to study the genetic diversity and structure of populations, or relatedness and pedigree of individuals (Granroth-Wilding et al. 2017, Sun et al. 2017, Hagemann et al. 2018). However, DNA data alone does not provide information on age structure of the population, but integrating it with camera trapping data could provide this information (Furnas et al. 2018). Furthermore, non-invasive methods combined with telemetry of a few individuals can improve precision of SCR estimates by incorporating additional movement information (Royle et al. 2013a, Sollmann et al. 2016, Linden et al. 2018). Although the approach may vary, our findings add to the growing SCR-based evidence that it is important to consider the spatial heterogeneity of an area and habitat selection of animals. By acknowledging that landscapes are heterogeneous and incorporating that into our analyses, we quantify habitat selection, which captures crucial ecology in terms of how the organism uses its environment both in terms of population density and in terms of individual space use.