We are investigating how initial plant community pattern and composition influence restoration of riparian woodlands in Mediterranean-climate river systems. Since 2000, in an effort to recover endangered Chinook salmon populations, portions of the lower Merced River, California and its floodplain have been restored by re-filling and grading abandoned sand and gravel quarry pits and constructing a meandering river channel through the restored floodplain. Here, we analyze four years of plant community data to evaluate the relative strength of biotic vs. abiotic controls on early community development. In Year 1, some sites were planted with sterile barley (Hordeum vulgare). In Year 2, factorial management treatments included seeding with mixtures of native annual and perennial herbs, seeding with mycorhizal inoculum, and controls. Treatments spanned relatively dry to permanently saturated environments and all sites were flooded in one of four years. We analyzed species cover data in 218 plots along 23 transects using Detrended Correspondence Analysis and Non-metric Multidimensional Scaling. In Year 1, plant community composition was strongly related to moisture regime and unaffected by seeding with barley. In Year 2, community composition was dominated by weedy species that sorted out along the moisture gradient and secondarily related to seeding or mycorhizal treatments. These trends continued through Year 4, indicating that management efforts to set initial community composition were overshadowed by colonization and establishment of colonizing species that sorted along soil moisture gradients.