Background and aims : Plant roots provide mechanical cohesion (c r ) to soil on slopes which are prone to shallow landslides. c r varies in heterogeneous natural forests due to the spatial, inter- and intra-annual dynamics of root demography. Characterizing root initiation density and mortality, as well as how root growth is influenced by abiotic and biotic factors is essential for exploring a root system’s capacity to reinforce soil. Methods: In this study, root demography data were monitored using field rhizotrons during 1.5 years in two naturally regenerated mixed forests in the French Alps. These forests are composed of trees growing in groups (tree islands) with large gaps between the islands. Three categories of driving variables were measured: (i) spatial factors: altitude (1,400 m, 1,700 m), ecological patch (gap, tree island), soil depth (0.0–1.0 m divided into five layers of 0.2 m); (ii) temporal factors: month (12 months from March 2010 to February 2011), winter (winter of 2009–2010 and 2010–2011); (iii) biological factors: root diameter classes (]0, 1] mm, ]1, 2] mm, ]2, 5] mm (according to the international standard ISO 31–11, ]x, y] denotes a left half-open interval from x (excluded) to y (included)). Two types of two-part models, a Hurdle model (H) and a Zero-inflated model (ZI) were used to fit root data with a high zero population, i.e. if root initiation or mortality was zero during a given time period, or if roots were not present at all points throughout a soil profile. Results: Root initiation quantity decreased with increasing soil depth, as well as being lower in tree islands. Both soil depth and ecological patch interacted strongly with altitude. Root dynamics were significantly less active with a lower net production and c r increment in winter and spring than in summer and autumn. Roots which were ]1, 2] mm in diameter contributed the most to c r compared to other diameter classes, as they had a high production but a low mortality. With regard to model selection, both H and ZI demonstrated similar outcomes and underestimated extreme values of root demography data. Conclusion: All factors contributed towards explaining the variability of root demography and c r . We suggest taking into consideration the seasonality of root dynamics when studying root reinforcement