Integrated optimization of strategic and tactical planning decisions in forestry

The traditional approach to plan the forest products value chain using a combination of sequential and hierarchical planning phases leads to suboptimal solutions. We present an integrated planning model to support forest planning on the long term with anticipation of the impacts on the economic and logistic activities in the forest value chain on a shorter term, and we propose a novel optimization approach that includes acceleration strategies to efficiently solve large-scale practical instances of this integrated planning problem. Our model extends and binds the models implemented in two solver engines that have developed in previous work. The first system, called Logilab, allows for defining and solving value chain optimization problems. The second system, called Silvilab, allows for generating and solving strategic problems. We revisit the tactical model in Logilab and we extend the strategic model in Silvilab so that the integrated planning problem can be solved using column generation decomposition with the subproblems formulated as hypergraphs and solved using a dynamic programing algorithm. Also, a new set of spatial sustainability constraints is considered in this model. Based on numerical experiments on large-scale industrial cases, the integrated approach resulted in up to 13% profit increase in comparison with the non-integrated approach. In addition, the proposed approach compares advantageously with a standard LP column generation approach to the integrated forest planning problem, both in CPU time (with an average 2.4 factor speed-up) and in memory requirement (with an average reduction by a factor of 20).