Forest ecosystem management, based partly on a greater understanding of natural disturbance regimes, has many variations but is generally considered the most promising approach to accommodating biodiversity concerns in managed forested regions. Using the Lake Duparquet Forest in the southeastern Canadian boreal forest as an example, we demonstrate an approach that attempts to integrate forest and stand-level scales in biodiversity maintenance. The concept of cohorts is used to integrate stand age, composition and structure into broad successional or stand development phases. Mean forest age (MFA) because it partly incorporates historic variability of the regional fire cycle, is used as a target fire cycle. At the landscape level forest composition and cohort objectives are derived from regional natural disturbance history, ecosystem classification stand dynamics and a negative exponential age distribution based on a 140 year fire cycle. The resulting multi-cohort structure provides a framework for maintaining the landscape in a semi-natural age structure and composition. At the stand level the approach relies on diversifying interventions, using both even-aged and uneven-aged silviculture to reflect natural stand dynamics, control the passage ('fluxes') between forest types of different cohorts and maintain forest-level objectives Partial and selective harvesting is intended to create the structural and compositional characteristics of mid- to late-successional forest types and, as such, offers an alternative to increasing rotation lengths to maintain ecosystem diversity associated with over mature and old-growth forests. The approach does not however supplant the necessity for complementary strategies for maintaining biodiversity such as the creation of reserves to protect rare, old or simply natural ecosystems. The emphasis on maintaining the cohort structure and forest type diversity contrasts significantly with current even-aged management in the Canadian boreal forest and has implications for stand-level interventions, notably in necessitating a greater diversification of silvicultural practices including more uneven-aged harvesting regimes. The approach also presents a number of operational challenges and potentially higher risks associated with mu1tipl-stand entries, partial cutting and longer intervals between final harvests. There is a need for translating the conceptual model into a more quantitative silvicultural framework. Silvicultural. © 2002 Elsevier Science B.V. All rights reserved.