Northern Prairie Wildlife Research Center
Although it would have been desirable to sample in each unit each year in a random fashion, for logistical reasons the south unit was surveyed in 1996 and the north unit in 1997. Of the alien species we found, yearly changes in population size are likely to be important in Melilotus officinalis, which is known to show strong year-to-year fluctuations (Turkington et al. 1978), and in annual species whose abundances may fluctuate in response to climatic variability. For most species, sampling frequency rather than cover minimizes differences associated with seasonal and yearly fluctuations (Elzinga et al. 1998). Vegetation was sampled between 15 May and 31 September. Taxonomy follows Barker et al. (1986).
We classified transects a posteriori as disturbed if any part fell within 100 m of roads, trails, campgrounds, picnic areas, buildings within the park boundaries, or fields seeded to non-native grasses. Thus, we defined as disturbed that native vegetation that occurred within 100 m of any of these features. Campgrounds and picnic areas themselves, for example, are not included in this analysis. Because the perimeter of the park either coincided with a road, which was already considered disturbed, or was adjacent to rangeland, we did not consider the perimeter, per se, disturbed. The 100-m buffers were generated using TNT Mips software (MicroImages, Lincoln, Nebraska, USA).
Although we chose to restrict this analysis to anthropogenic disturbance, we recognize that other forms of disturbance are found in the park, including seasonal flooding in the river bottoms, excavation and grazing in prairie dog towns, and ongoing erosional processes, among others. We also chose not to include an explicit spatial component within park units in our analysis. A preliminary analysis using general linear models revealed no systematic differences in number of alien species or frequency of individual species among strata (data not shown); vegetation type accounted for far more of the variability in these variables than did stratum.
Data analysis. Because we were interested in both main effects and interactions, only those vegetation types that included both disturbed and undisturbed transects and that occurred in both the north and south units of the park were used in the analysis, although we describe the excluded vegetation types in general terms as a point of comparison. We used Akaike's Information Criterion (AIC; Burnham and Anderson 1998), which incorporates the log-likelihood with a penalty for added parameters, to determine the best model out of all possible models containing park unit, disturbance, vegetation type, and their interactions. We also used AIC to compare models with only main effects for (1) number of alien species per transect, (2) frequency of alien species on transects that contained at least one alien species (hereafter termed "occupied transects"), and (3) presence/absence of each alien species that occurred on at least 10 transects. The difference between the AIC score for the best model and each model with only one of the explanatory variables is an indication of the relative importance of each variable. We chose to use the information theoretic approach, represented by AIC, rather than a statistical hypothesis testing approach for several reasons. First, our primary goal was to choose the best model out of several potential models, rather than from specific alternatives defined a priori. Second, we were more interested in effect size rather than significance of the various parameters; our sample size was large, so that virtually all factors were statistically significant, regardless of their biological significance. Finally, we wanted to compare effect sizes of the three explanatory variables, which we could not do with conventional hypothesis testing procedures, since vegetation type had many more levels and thus accounted for much more of the variation in these kinds of models than did either park unit or disturbance. All analyses were done using Proc Mixed (SAS Institute 1992).