prilagođeno pretraživanje po punom tekstu

ŠUMARSKI LIST 3-4/2018 str. 50     <-- 50 -->        PDF

Shapiro-Wilk test and homogeneity of variances using Levene test. If the criterions of one-way ANOVA were not fulfilled, then we used nonparametric Kruskal-Wallis median test. When significant differences were detected in the Kruskal-Wallis test, we employed Dunn’s procedure for post hoc multiple comparisons. The seasonal (spring, autumn) difference of capture parameters was tested by Mann-Whitney U test (Zar, 2010). These statistical tests were computed in Statistica 8.0 software (StatSoft Inc. 2007).
The POPAN formulation (Schwarz & Arnason, 1996) of Jolly-Seber models (Jolly, 1965; Seber, 1965) was used to perform the comparative estimates of bank vole population size. The encounter history of POPAN models included the different forest habitats as three groups. To test the assumptions of open models we used Goodness-of-fit tests (RELEASE tests 2 & 3) which indicated that this model is suitable for estimating population parameters from data of three habitats. Using an information-theoretic approach (Burnham & Anderson, 2002; Mazerolle, 2006), we built 10 candidate models to determine the effects of three monitored forest stands. The global model included the following parameters: {φ(habitat+period), p(habitat+period), pent(habitat+period), N(habitat)}. It assumes that (i) φ (survival rate) differs between habitats and periods (the four 5-day trapping sessions), (ii) p (capture probability - given the animal is alive and available for capture) differs between forests and periods; and (iii) pent (probability of entry into the population per occasion) differs between forests and periods and it estimates N (super-population size) for habitats separately. Models were fitted using the logit link function for φ and p, the identity link function for N, and the multinomial logit link function to pent (White & Burnham, 1999). The model selection procedure was based on Akaike’s Information Criterion modified for small sample size (AICc). The model with the lowest value of the AICc (ΔAICc = 0) was the most parsimonious. To help evaluate the fit of the models, we also considered the difference in AICc (ΔAICc), as models which differ by less then 2 AICc units (ΔAICc < 2) receive substantial support from the data (Burnham & Anderson, 2002). All models were run in MARK 6.1 (White & Burnham, 1999).
During the four sampling months 740 individuals of bank vole were marked in the three different forests. The number of captures was 1401 while the total number of recaptures of our model species was 661. In case of three investigated forest habitats the capture data of trapping success for bank vole populations, and also their seasonal variations, showed differences. The total number of captures (ni) (Kruskal-Wallis test: H (2, N=60) = 16.14, P < 0.001) and the ratio of recapture (%) (H (2, N=60) = 14.74, P < 0.001) differed significantly between the forest habitats, while the number of marked animals (mi) did not show significant differences in the comparison of forest stands. Based on post hoc Dunn-test, both the number of captures (PFH-HU vs HUFM-CRO: z = 3.98, P < 0.001; RH-HU vs HUFM-CRO: z = 2.41, P .05) and recapture success (PFH-HU vs HUFM-CRO: z = 3.73, P < 0.001; RH-HU vs HUFM-CRO: z = 2.61, P < 0.05) were significantly higher in the two Hungarian forest habitats than in Croatia (Fig. 1).
Comparing the two investigated seasons, the total number of animals differed significantly in the Hungarian protected forest habitat (Mann-Whitney test: z = 2.08, P < 0.05) while the other two parameters did not show significant difference between summer and autumn (z = 0.34-1.62, n.s.) (Fig. 2). In case of the reforested habitat, the values of two capture parameters were significantly higher in the autumn than in the summer (ni: z = 2.79, P < 0.01; rr%: z = 2.23, P < 0.05), but in case of the number of known individuals, there were no significant results (mi: z = 0.37, n.s.) revealed by the Mann-Whitney test in comparing the two seasons. Based on data of the habitat under forest management, only the recapture rate was significantly higher in the autumn than in the summer (rr%: z = 2.38, P < 0.05), although the mean of recapture success was lower than in the other two forests in both two periods. The values of the other two parameters did not differ significantly between the two seasons (mi: z = 1.21, n.s.; ni: z = 0.19, n.s.) (Fig. 2).
Based on the analysis of capture-mark-recapture data, the Goodness-of-fit tests (Test 2 + Test 3: c2 = 5.03 - 7.19, n.s.) indicated that the candidate POPAN models are suitable to estimate the population size of bank vole in all three investigated forest habitats. According to model selection, the best reduced model (smallest AICc value and the largest Akaike (AICc) weight (wi)) was used to estimate the parameters (Table 2).