Soil carbon storage plays a key role in the global carbon cycle and is important for sustaining forest productivity. Removal of unpaved forest roads has the potential for increasing carbon storage in soils on forested terrain as treated sites revegetate and soil properties improve on the previously compacted road surfaces. We compared soil organic carbon (SOC) content at several depths on treated roads to SOC in adjacent second-growth forests and old-growth redwood forests in California, determined whether SOC in the upper 50 cm of soil varies with the type of road treatment, and assessed the relative importance of site-scale and landscape-scale variables in predicting SOC accumulation in treated road prisms and second-growth redwood forests. Soils were sampled at 5, 20, and 50 cm depths on roads treated by two methods (decommissioning and full recontouring), and in adjacent second-growth and old-growth forests in north coastal California. Road treatments spanned a period of 32 years, and covered a range of geomorphic and vegetative conditions. SOC decreased with depth at all sites. Treated roads on convex sites exhibited higher SOC than on concave sites, and north aspect sites had higher SOC than south aspect sites. SOC at 5, 20, and 50 cm depths did not differ significantly between decommissioned roads (treated 18–32 years previous) and fully recontoured roads (treated 2–12 years previous). Nevertheless, stepwise multiple regression models project higher SOC developing on fully recontoured roads in the next few decades. The best predictors for SOC on treated roads and in second-growth forest incorporated aspect, vegetation type, soil depth, lithology, distance from the ocean, years since road treatment (for the road model) and years since harvest (for the forest model). The road model explained 48% of the variation in SOC in the upper 50 cm of mineral soils and the forest model, 54%. Copyright © 2015 John Wiley & Sons, Ltd.