Spring-fen habitat islands in a warming climate: Partitioning the effects of mesoclimate air and water temperature on aquatic and terrestrial biota.

Climate warming and associated environmental changes lead to compositional shifts and local extinctions in various ecosystems. Species closely associated with rare island-like habitats such as groundwater-dependent spring fens can be severely threatened by these changes due to a limited possibility to disperse. It is, however, largely unknown to what extent mesoclimate affects species composition in spring fens, where microclimate is buffered by groundwater supply. We assembled an original landscape-scale dataset on species composition of the most waterlogged parts of isolated temperate spring fens in the Western Carpathian Mountains along with continuously measured water temperature and hydrological, hydrochemical, and climatic conditions. We explored a set of hypotheses about the effects of mesoclimate air and local spring-water temperature on compositional variation of aquatic (macroinvertebrates), semi-terrestrial (plants) and terrestrial (land snails) components of spring-fen biota, categorized as habitat specialists and other species (i.e. matrix-derived). Water temperature did not show a high level of correlation with mesoclimate. For all components, fractions of compositional variation constrained to temperature were statistically significant and higher for habitat specialists than for other species. The importance of air temperature at the expense of water temperature and its fluctuation clearly increased with terrestriality, i.e. from aquatic macroinvertebrates via vegetation (bryophytes and vascular plants) to land snails, with January air temperature being the most important factor for land snails and plant specialists. Some calcareous-fen specialists with a clear distribution centre in temperate Europe showed a strong affinity to climatically cold sites in our study area and may hence be considered as threatened by climate warming. We conclude that prediction models solely based on air temperature may provide biased estimates of future changes in spring fen communities, because their aquatic and semiterrestrial components are largely affected by water temperature that is modified by local hydrological and landscape settings.