Determination of spatial and temporal patterns of rockfall events remains a serious challenge in most mountain areas and especially when it comes to quantitative hazard assessments, because of the scarcity and incompleteness of long-term records. This lack of reliable baseline data is particularly problematic in urbanized areas where rockfall risk tend to increase as urbanization is climbing up on slopes. On forested slopes, dendrogeomorphic methods have been applied repeatedly to fill this data gap, as they provide both spatial and temporal reconstructions of past rockfall events with high accuracy. In this study, the systematic mapping of 1,004 broadleaved trees at Saint-Paul-de-Varces (French Alps) was used to document a total of 1,516 rockfall scars visible on the surface of stems. We then coupled the so-called scar counting approach with the conditional impact probability concept, so as to estimate the likelihood of those rockfalls which did not collide with trees. This coupled method allows estimating and mapping of recurrence interval of rockfalls. Our results show a clear reduction of rockfall frequency in the down slope direction as well as a noticeable lateral change in it. This is consistent with the concave profile of the slope and the barrier effect of trees at the study site. These findings also demonstrate, on one hand, the usefulness of broadleaved tree species to reconstruct rockfall frequencies and, on the other hand, the efficiency of our approach to reveal spatio-temporal patterns of rockfall activity. The approach presented here could become a powerful tool for rockfall hazard assessments which, coupled with a 3D-modelling of block trajectories, will further allow computation of frequency - intensity maps on forested slopes.