Aims: Root architecture drives plant ecology and physiology, but current detection methods limit understanding of root placement within soil profiles. We developed a statistical model of root volume along depth gradients and used it to infer carbon storage potential of land-use changes from conventional agriculture to perennial bioenergy grasses. Methods: We estimated root volume of maize-soybean rotation and three perennial grass systems (Miscanthus × giganteus, Panicum virgatum, tallgrass prairie mix) by Bayesian modeling from minirhizotron images, correcting for small images and near-surface underdetection. We monitored seasonal and inter-annual changes in root volume distribution, then validated our estimates against root mass from core samples. Results: The model explained 29% of root volume variation and validated well against core mass. Seventh-year perennials had greater belowground biomass than maize-soybean both in total (11-16×) and throughout the profile (2-17× at every depth < 120 cm). Perennials’ relative depth allocations were stable over time, while total root volume increased through five years. In 2012 a historically hot, dry summer damaged maize while perennials appeared resilient, suggesting their large-deep root systems aid drought resistance. Conclusions: Perennial root systems are large, deep, and persistent. Converting row crops to perennial bioenergy grasses likely sequesters carbon in a large, potentially very stable, soil pool.