Table 4. Model results for winter climate variability (mean number of above-zero days during January-March), July mean temperature, and bioclimatic zones (proportion of presumed optimal bioclimatic zone of the surrounding 25 km2 landscape) on characteristics of rodent population dynamics. The models for community contribution (abbreviated as “community cont.”) include all available sampling units, while the models for cycle amplitude include only the sampling units where a given genus was present, sample size given in parenthesis. For voles, the presumed optimal bioclimatic zone is the low alpine zone, for lemming the middle alpine zone. Results for best models are shown, dataset A refers to “all data” (n=22 locations), dataset N to “Norwegian data only” (n=17 locations). Values in parentheses show 95% confidence limits; predictor variable estimates for which the 95% confidence interval does not cross zero are denoted in bold. Random effects are given as standard deviation of variation; in parentheses is the proportion of variance assigned to location effect. Star denotes effects that were statistically significant when data from fall only was included in the models. Dash denotes variables that were included in the set of evaluated models, but did not appear in the best model, whereas empty cells denote variables that were not evaluated in the given model set. Reference level for the factorial variable sampling season is fall; the effect size estimate corresponds to a difference between fall and spring. See Appendix 1 for model selection table (Tables A6, A7) and results at 9 km2 and 1 km2 (Tables A8, A9).