Response of Mediterranean grassland species to changing rainfall

(1) Long-term changes (over 14 years) were investigated in the composition of a species-rich Mediterranean grassland (`dehesa') in south-western Spain. The frequency of every species of herbaceous vascular plant was recorded annually at the end of the growing season, from 1977 to 1986 and again in 1990, in permanent plots along a transect. Monthly rainfall data over the same period were typical of the Mediterranean climate: there was virtually no rainfall in summer (June-September), when potential evapotranspiration was high; rainfall in the cooler part of the year was unpredictable in amount and distribution. (2) The vegetation was dominated by winter annuals, which germinated after the autumn rains and fruited in late spring or early summer; they comprised 86 of the 99 species recorded. Most of these species displayed considerable variations in abundance over the period of study and the abundance of some fluctuated violently from one year to another. (3) Correlations were examined between the logit of total annual frequency of each species (n = 11 years) and rainfall totals over the growing season (October-May) and its three components: autumn (October-November), winter (December-February) and spring (March-May). Evidence for trends in logit frequency that might represent successional changes was sought from similar correlations with time. (4) The abundance of many species was highly correlated with rainfall, which accounted for up to 83% of year-to-year variance (in Juncus capitatus). Rainfall over the whole growing season was a good predictor of abundance for more species than any of its component periods (25 species with significant associations out of 99). Cluster analysis resolved five groups of species with similar year-to-year variations in abundance: two groups contained most of the species showing significant positive correlations with rainfall, and another group contained a small number of generally infrequent species that showed significant negative correlations. The species in the remaining two groups had few significant correlations with rainfall. Only three moderately abundant species were significantly correlated with time. Successional change was apparently much less important than rainfall as a source of variation in plant frequency. (5) The demographic mechanisms underlying these relationships, their implications for the regulation of community structure and their possible value in detecting the consequences of climatic change in the Mediterranean region are discussed.