RESUMEN
The decreasing abundance of mature forests and their fragmentation have been identified as major threats for the preservation of biodiversity in managed landscapes. In this study, we developed a multi-level framework to coordinate forest harvestings so as to optimize the retention or restoration of large mature forest tracts in managed forests. We used mixed-integer programming for this optimization, and integrated realistic management assumptions regarding stand yield and operational harvest constraints. The model was parameterized for eastern Canadian boreal forests, where clear-cutting is the main silvicultural system, and is used to examine two hypotheses. First, we tested if mature forest tract targets had more negative impacts on wood supplies when implemented in landscapes that are very different from targeted conditions. Second, we tested the hypothesis that using more partial cuts can be useful to attenuate the negative impacts of mature forest targets on wood supplies. The results indicate that without the integration of an explicit mature forest tract target, the optimization leads to relatively high fragmentation levels. Forcing the retention or restoration of large mature forest tracts on 40% of the landscapes had negative impacts on wood supplies in all types of landscapes, but these impacts were less important in landscapes that were initially fragmented. This counter-intuitive result is explained by the presence in the models of an operational constraint that forbids diffuse patterns of harvestings, which are more costly. Once this constraint is applied, the residual impact of the mature forest tract target is low. The results also indicate that partial cuts are of very limited use to attenuate the impacts of mature forest tract targets on wood supplies in highly fragmented landscapes. Partial cuts are somewhat more useful in landscapes that are less fragmented, but they have to be well coordinated with clearcut schedules in order to contribute efficiently to conservation objectives. This modeling framework could easily be adapted and parameterized to test hypotheses or to optimize restoration schedules in landscapes where issues such as forest fragmentation and the abundance of mature or old-growth forests are a concern.