QUESTION : Can variation in the outcome of biotic interactions in relation to
environmental severity bemore accurately predictedwhen consideringmultiple
stress and/or disturbance variables?
LOCATION : Arctic-alpine tundra in Kilpisj€arvi, North Finland.
METHODS : To test the impact of including multiple environmental variables in
analyses of the outcomes of biotic interactions, we modelled reproductive effort
and cover of 17 arctic-alpine species as a function of Empetrum nigrum subsp.
hermaphroditum cover, geomorphological disturbance and soil moisture with
statistical interactions of the explanatory variables included.We implemented a
best-subset approach using generalized linear models (GLM) and selected the
bestmodel for each species based on Akaike’s information criterion (AIC).
RESULTS : For the majority of species, models including multiple environmental
variables were selected as best. Reproductive effort depended on one or both
environmental variables for all species, and 14 species were additionally influenced
by Empetrum,with the impact of Empetrum varyingwith abiotic conditions
in all but one of those species. Moreover, the three-way interaction of three
explanatory variables was included in the best-fit models for six species. The
impact of Empetrum on species cover showed a similar pattern, with 11 species
affected by Empetrum and its statistical interactions with one or both abiotic variables.
CONCLUTIONS : Biotic interactions have an important role in arctic-alpine vegetation,
but to fully understand variation in their effects multiple environmental
factors should be explicitly considered. In this study, the outcome of biotic interactions
was frequently dependent on two abiotic variables (and occasionally
additionally on their statistical interaction). Therefore, we demonstrate that
studies based on only one environmental factor may cause misleading interpretations
of the nature of biotic interactions in plant communities where there are
multiple independent variables underlying the habitat severity gradient.