1. Soil resource partitioning and dispersal limitation have been shown to shape the tree community structure of mature tropical forests, but are poorly studied in the context of forest succession. We examined the relative contributions of both ecological processes to the variation in the species composition of young tropical secondary forests at different spatial scales, and if the relative importance of these two ecological processes changed during succession. At the species level, we examined if the association between species abundances and soil fertility differed between early and late successional species and/or changed over the course of succession. 2. We used vegetation and soil data from 47 secondary forest sites with two plots each in a tropical agricultural landscape. A distance-based redundancy analysis and variation partition were employed to examine the relative importance of spatial distance (proxy for dispersal limitation) and heterogeneity in soil nutrients (proxy for soil nutrient partitioning) at the landscape scale, and a linear regression to test their effects at the local scale. We examined interspecific variation in species’ responses to successional age and soil nutrients with a joint species distribution model. 3. Dispersal limitation and soil niche partitioning drove considerable variation in the composition of plant communities at local and landscape scales. The relative contribution of these two ecological processes changed with scale (local versus landscape), topography (lower slope versus upper slope plots). At the species level, significant abundance-soil fertility associations were mostly positive. Most species became less responsive to soil nutrients over the first few decades of tropical forest succession, probably because light became the main limiting resource in older forests. 4. Synthesis. Our key finding is that spatial heterogeneity in soil resources and spatial distance jointly drive compositional variation within and across early successional forests. Our results highlight that a network of forest fragments enhances the resilience of ecological processes and the potential of secondary forests to restore and preserve biodiversity in human-modified landscapes. To advance our understanding of ecological succession, we need to move beyond single-factor and local-scale studies and examine the effects of multiple variables on succession at multiple spatial scales.