• California;
  • Climatic variation;
  • Disturbance;
  • Gopher;
  • Long-term study;
  • Serpentine;
  • Vegetation dynamics

Abstract. We present data from the first 11 years of a longterm study of the dynamics of an annual grassland on serpentine soil in Jasper Ridge Biological Preserve, Northern California. Annual rainfall amounts and distributions varied greatly over the period 1982-1993, as did the amount and distribution of gopher disturbance. Temporal variation in gopher disturbance showed no relationship with rainfall, but spatial variation in disturbance frequency was related to soil depth. The disturbance regime experienced by the grassland is complex, both spatially and temporally, and most of the area is disturbed at least once every 3-5 years. Plant species abundances showed a variety of responses to climate variation and disturbance. Abundances of individual species in any given year could not be linked directly to rainfall amount due to hysteresis effects and other interactions. The grassland composition changed markedly over the study. Exclusion of gophers suggested that changing abundances of several species were linked to gopher disturbance. In particular, perennial species' abundances increased greatly in the years following exclosure, but then subsequently declined. Data on plant densities on gopher mounds disturbed at different times of year and in different years indicate that the local species composition remains distinct for a number of years following disturbance. Disturbance history is hence a major factor controlling local community variation. Changing species importances, a complex disturbance regime and the importance of disturbance history make prediction and modelling of this system difficult. It is suggested that the same is probably true for many plant communities, and that long-term studies must be an essential part of ecological research programs. This study illustrates the practical problems inherent in maintaining long-term field experiments and in analyzing complex time series data which suffer from inadvertent deviations from the original experimental design.