Fig 1.
Overview of and photos of the study area.
(A)The big map shows Denmark with the study location marked with a red dot and the inserted map shows the location of Denmark in Europe. (B) Ortho photo of the bison enclosure (yellow line) showing the GPS points used for the vegetation analysis within the meadow (blue points) and within the release area (red points). (C) Photo of the bison herd feeding in the meadow. (D) Bison herd feeding at the hay rack in the release area. (E) Photo of the forest habitat in the eastern forest patch.
Table 1.
Description of the four different types of behaviour monitored while observing the bison herd.
Fig 2.
Response and predictor variables used for the spatial linear and logistic regression.
Response variables: (A) Response variable: Presence/absence and (B) frequency of occurrence of the bison herd summed across the whole observation period. Predictor variables: (C) forage quality, (D) tree cover, (E) SLA, and (F) elevation.
Table 2.
Description of the predictor variables used in the spatial linear and logistic regression.
Fig 3.
Daily and seasonal behavioural patterns.
Fraction of observations where the bison herd rest, feed, move or express other types of behaviour during a day throughout the observation period (spring, summer and autumn) (A). Time spent on a specific type of behaviour across the seasons (spring, summer, and autumn) (B). Different letters are attributed to seasons that vary significantly from other seasons within the same type of behaviour.
Fig 4.
Correlation between temperature and time spent resting or feeding.
Correlation between mean daily temperature and time spent on (A) resting or on (B) feeding across the observation period. Statistical results from Pearson´s correlation corrected for temporal correlation are shown.
Table 3.
Statistical results of chi-square test of habitat use of different habitats and management areas when accounting for their habitat availability.
Table 4.
Time spent by the bison herd in each type of habitat compared to the availability of the habitat types including Jacob´s food selection index.
Fig 5.
Correlations between frequency of occurrence and predictor variables.
Correlations between log frequency of occurrence and (A) elevation, (B) tree cover, (C) management area, (D) SLA, and (E) forage quality). Statistical results from Pearson´s correlation or Spearman´s rank correlation is shown for continuous variables and ordinal variables. An asterisk indicates a significant difference among management areas (C).
Fig 6.
Time spent feeding and resting across management area.
Number of feeding events or resting events of the bison herd observed in the meadow area or in the release area. n value indicates number of grid cell per management area. An asterisk indicates a significant difference in number of observations between management area.
Table 5.
Three best spatial linear models on log frequency of occurrence (ΔAICc <2) selected from eight tested models and ranked by AICc.
Table 6.
Summary of results after full model-averaging: Effects of each explanatory variable on frequency of occurrence (linear regression) and presence/absence (logistic regression) of bison.
Table 7.
The best spatial logistic model on presence/ absence occurrence (ΔAICc <2) selected from eight tested models and ranked by AICc.
Fig 7.
Relation between plant preference and SLA-values.
Correlation between plant-bison contacts, availability of plant species, and SLA-values of plant species. Un-preferred plant species (rectangle) is characterized by being abundant while having few contacts with a bison, Preferred plant species (triangle) is characterized by having a low abundance while having many contacts with a bison, and plant species with no preference (circle) is characterized by having increasing contacts with a bison with increasing abundance. Figure adapted from Borowski and Kossak, 1972.