Understanding the distribution of native versus exotic plant diversity in California's grassland landscapes

The invasion of California's grasslands by exotic plants has been so extensive that the questions of where remnant native-dominated grasslands are located and why have been termed an ecological murder mystery. We examined the effects of isolation from roads and other indicators of human disturbances, and their interactions with soil type (non-serpentine and serpentine), aspect (cool, warm, and neutral slopes), and land use (presence and absence of livestock grazing), on native and exotic plant diversity in an inland California foothill grassland landscape. We found that patterns of native persistence and exotic invasion differed depending on plant life form and interactions among predictors. Non-serpentine grassland sites >1000-m from roads (distant) had more native grass species and fewer exotic forb species than sites 10-m from roads (near). In serpentine grasslands, there was no effect of distance from roads. Similarly, remote grassland islands (grassland patches ≤2 ha and isolated within a chaparral or oak woodland matrix) contained significantly greater native grass cover than grasslands >10 ha and contiguous with roads and other grasslands.; 64 pairs of plots uncleared and cleared of above-ground plant material at all near and distant sites from the first study. On non-serpentine soils, Centaurea's seedling survival was greater in uncleared (but not cleared) 10-m than >1000-m plots. These results suggest that beyond the seedling stage, biotic resistance inhibits Centaurea invasion of non-serpentine grasslands at sites distant from roads more than at sites near roads.; Finally, I used classification-tree analysis to ask whether distance from roads, soil, aspect, and land use can predict locations of hot spots of native plant diversity, as well as the presence of 3 exotic species. Classification-tree models correctly predicted characteristics of sites containing high native perennial grass and native annual richness in 83% and 63% of sites that contained those characteristics. In addition, they correctly predicted characteristics of sites containing Centaurea solstitialis, Aegilops triuncialis, and Taeniatherum caput-medusae in 79%, 62%, and 83% of cases, respectively. By illustrating how combinations of natural and anthropogenic factors may interactively influence vegetation patterns, I propose that my approach can be used as a conceptual framework for understanding how to (1) locate and protect remnant native plant communities in invaded landscapes and (2) explain confounding exceptions to generalizations.