Predictability of Change in Shoreline Vegetation in a Hydroelectric Reservoir, Northern Sweden

Abstract

We asked whether we could predict the responses of vegetation to specified perturbations acting on a known initial state using 10 yr of vegetation data and precise water level records from the Gardiken Reservoir on the Ume River in Sweden; this has a restricted flora and one obvious controlling factor: water level. Abundance and species richness were strongly correlated for both vascular and nonvascular plants (bryophytes and lichens). Duration of flooding was the most important variable controlling those independent variables chosen for study. The change from year t to t + 1 could be roughly predicted for community attributes such as richness and cover (R2 > 0.33) but not for species composition (R2 < 0.04). The initial (preceding-year) state of the system and the duration of flooding (also in the preceding year) were always the two most important independent variables. The major conclusions were that (1) even in this "simple" system with presumably one major control variable, our best regression equation accounted for only 41% of the variation, (2) community properties were more predictable than species composition, and (3) the similarity of the species pool, the lack of coupling from competitive interactions, and the high degree of immaturity may have reduced predictability of species composition.