Background and aims Identifying root types helps understand the diversity of the morphogenetic capacities of root axes. Unfortunately, root types are generally defined arbitrarily. We used an unsupervised clustering method to define root types and examine the ontogeny of the Pinus pinaster root system. Methods K-clustering was first used on four root traits, including three geometric architectural traits: basal cross-sectional area, root tropism, parent root tropism, and branching angle. Based on these groups, we assigned all the root axes to a root type. Results Clustering yielded the same five groups of lateral roots and explained the same percentage of variance (70%) whatever the tree age. This way, 11,004 root axes from 69 excavated root systems ranging from 3 to 50 years were assigned in 5 types. P. pinaster root axes showed large differences based on root tropism and parent root tropism intensity. The larger horizontal shallow roots represented most of the woody volume throughout the life cycle of the trees. The frame of the central part of the root system was almost completed in 4-year-old trees. Conclusions In addition to the taproot, five types of lateral roots were identified in P. pinaster based on root geometry and size, and no consistent age effect was reported. This method could improve the connexion between root trait data and root modelling.