Forest restoration in southern Amazonia: Soil preparation triggers natural regeneration

Forest regeneration in abandoned pastures in Amazonia has been well studied, but active restoration o f non- resilient pastures has not. In this work, we evaluated large-scale active restoration of intensively used pastures in southern Amazonia, where the highest deforestation rates are observed. With the construction o f the Jirau Dam in the Madeira river (state o f Rondonia, Brazil), a 3000-ha forest buffer zone has been established. This area was previously covered by African pasture grasses for cattle ranching. Eight mixed-species planting sites with variable grass management and ranging from 6 to 60 months post-planting were measured three times over 2.5 years. We also tested a gradient o f restoration intensity in an experimental planting: (i) no intervention (control), (ii) harrowing and herbicides to control grasses, and (iii) harrowing and herbicides plus tree planting. Our goals were to understand the initial trajectory o f actively restored sites, the role of harrowing and herbicide application in triggering natural regeneration, and the role of seedling planting on the initial vegetation structure. All tree species >= 30 cm in height were sampled in five circular 10-m-radius plots per site. Plant cover was also monitored using the step-point method. At 18 months, stem density ranged from 2500 to 14,490 ind ha(-1), demonstrating that density increased suddenly in most sites through colonization, although this was highly variable. Tree cover reached 81% in five years, virtually eliminating grass cover after 36 months. Recruits contributed more to basal area than planted seedlings. Although 17 species, on average, were planted in restoration sites, the number of species at the sites steadily increased over time, at a rate of 7 species yr(-1) per 1570 m(2). Cecropia spp. and Trema micrantha recruits had the highest stem densities and basal areas. Harrowing and grass control were enough to trigger succession in sites where natural regeneration was not taking place, shifting these sites to the highly resilient route o f early secondary forest succession in the Amazon. Future restoration efforts should use a combination o f methods, first evaluating the potential for natural regeneration and then gradually eliminating barriers.