Get reliable science

Several studies have demonstrated that seed mass is related to different environmental factors. However, they have taken no account of the joint effects of spatial and phylogenetic information. We analysed the distribution pattern of seed mass along an elevational gradient (1040-2380 m a.s.l.) at the community level in grasslands of the southern Alps. First, we tested the influence of environmental filters (climate and soil properties) in determining community-weighted seed mass variation in mountain grasslands. Second, we verified the relative roles of environmental filters in determining seed mass variation after accounting for spatial and phylogenetic autocorrelation with an eigenvector filtering approach. Temperature, soil fertility, and soil pH were the most important predictors for explaining seed mass variation; specifically, warmer, low fertility, and alkaline grasslands showed a greater seed mass. Inclusion of spatio-phylogenetic filters in the model increased its fit and the variance explained and reduced autocorrelation significantly but had substantial effects on the parameter estimates, with temperature and soil pH becoming insignificant. This effect may be ascribable to spatially structured phylogenetic patterns and could likely result from the common evolutionary histories shared by many species at sites with similar environmental conditions. Therefore, the observed patterns between community-weighted seed mass and both temperature and soil pH are not independent of phylogeny, but they are explained by the shared history within genera and families. Nevertheless, soil fertility remained the most important predictor for explaining seed mass variation. The results of this work contribute to better understanding the combined effects of environment and evolutionary factors for determining seed mass distributions in the spatial context of mountain grasslands. The observed relationships with climate and soil properties are particularly interesting because they are potentially relevant when modelling plant trait composition under changes in land use and climate.