In this paper, we introduce an ecologically relevant measure of drought stress and its relationship with vegetation characteristics. We used process-based simulations of reference drought stress on inclined surfaces with different soil type, slope, and aspect to get an insight in the drought statistics that explain small-scale differences in vegetation characteristics. Mean intensity, duration, and frequency of drought events are commonly integrated into the so-called “dynamic drought stress” TSdyn. We introduce a simple, physiology-based alternative measure of drought stress TSupp, based on the finding that plants respond to extremes rather than to mean intensities of stress events. Using extremes instead of mean intensities makes information on duration and frequency of stress events superfluous. We show that compared with TSdyn, TSupp (i) is a better predictor of the fraction of xerophytes within a vegetation plot, (ii) is simple, transparent, and easy to interpret as only the uppermost intensities are involved, i.e. minimum parameters are needed, and (iii) does not involve empirical parameters, as are used in TSdyn. Focusing on extremes rather than on means is especially important in predicting climate change effects, as especially these extremes are predicted to increase due to climate change. Copyright © 2011 John Wiley & Sons, Ltd.