In Mediterranean and arid plant communities, water availability is considered to be the key limiting abiotic factor. Therefore, the main impact of recent climate change should result from changing rainfall patterns instead of rising temperatures or CO₂ levels.

Most predictions about climate-change impacts are based on space-for-time substitutions: They assume that both single species as well as entire plant communities will shift their distributions according to changing climatic conditions, and hence that their future distributions can be estimated from present-day spatial patterns.

However, despite substantial discussions about shortcomings of this approach, its validity has never been tested for plant communities. Here, we present the first such test.

We combined a steep natural rainfall gradient in Israel with additional experimental rainfall manipulations. During six consecutive years, detailed count data of annual plant communities was obtained from 400 permanent quadrats in four environments, ranging from wet-Mediterranean to arid conditions. Additionally, in the two intermediate environments rainfall was artificially increased and decreased by 30%, respectively. Hence, in manipulated plots the rainfall conditions resembled those of the adjacent drier respectively wetter ecosystem. According to the space-for-time approach, the plant communities in manipulated plots were expected to undergo a directed succession towards the corresponding wetter respectively drier plant community.
Results/Conclusions

There were strong year-to-year fluctuations in overall density, species richness and composition of the plant community in all environments. These community fluctuations were correlated to the annual rainfall amounts and patterns, and their magnitude increased along the gradient with aridity.

However, in both environments rainfall manipulations had no effect on either overall density or species composition after the six study years. Instead, the effect of rainfall manipulations was overruled by the year-to-year fluctuations, and therefore it did not lead to directional succession. Hence, our results contradict the predictions derived from a space-for-time approach.

We interpret that plant species in Mediterranean and arid environments were selected for adaptation to strong natural fluctuations in rainfall. Since the applied rainfall manipulations lie well in the range of natural fluctuations, the plant communities are highly resistant against these short-term climatic changes. The resistance may be additionally increased by the presence of long-lasting soil seed banks.

We conclude that long-term experiments are needed especially in highly variable ecosystems to study possible effects of climate change. Experimental approaches, which focus on mechanisms, should be favored over purely correlative space-for-time approaches.