Northern Canadian boreal forest has a considerable ecological and economic importance, with the black spruce forest type occupying a large extent of this ecosystem. Organic layer thickness and its relationship to topography are two key factors affecting tree growth and productivity of black spruce boreal forests of the Canadian Clay Belt region. This study linked multi-scale models of organic layer thickness and topography to improve our understanding of how these variables influence forest productivity and its distribution at different spatial scales within the Clay Belt region, northwestern Quebec. Field data were used to calculate site indices, which were used as estimators of forest productivity. Organic layer thickness was determined from field measurements obtained by manual probing, whereas topographic variables were extracted from multi-scale LiDAR-derived digital terrain models (DTM) at four resolutions, i.e., 5-, 10-, 15- and 20-m. Correlations between individual predictors and site index were found to be weak; however, few were significant, viz., organic layer thickness. Regression tree-based models were fitted using two different sets of explanatory variables at the four scales: organic layer thickness and topography (model 1); and topographic variables only (model 2). Organic layer thickness, aspect, and slope were the most important variables explaining forest productivity (63% and 31% total variance explained for models 1 and 2, respectively). Model 1 was found to be scale-independent, since the total explained variance was similar under the four resolutions, whereas with model 2, effects of topography on productivity were greater for coarser scales (highest R2 at 20-m resolution). Both models indicated higher forest productivity on southwest-facing slopes (i.e., >2.2%) with shallow organic layers (<35 cm), so then where organic horizons are the deepest the tree productivity is low. In contrast, lowest site indices (expressing low productivity) were found in areas with very deep organic layers (>85 cm). The resulting models could be applied at operational scales to predict site index at locations for which organic layer thickness information and DTM exist. Such information could be used to help forest managers in predicting how forest growth will respond to various harvesting activities.