Optimization of harvest planning of forest stands infested by Spruce Budworm using stochastic programming approach

Abstract

In the forest industry, harvesting process is one of the key critical processes as it supplies the primary raw material for different mills. However, due to several natural disturbances such as insect outbreaks, the impact and the effects on the tactical planning of forest supply chain can be irreversible. We consider the susceptibility, vulnerability, and increasing mortality by defoliation in trees over time caused by Spruce Budworm (SBW) infestation. The aim of this project is to use advanced optimization methods, in our case Stochastic Programming (SP), to maximize the market value of the harvested logs considering the occurrence of infestation over all the possible infestation scenarios. In our research method, we formulate a deterministic Mixed Integer Linear Programming (MIP) model which has then been extended into a Two-Stage SP model to deal with uncertainty related to the severity and propagation of the infestation; we also, as well, track the levels of infested volume inventory of the forest stands under the phases of SBW infestation according to their life cycle. The models are implemented in the modelling language of AMPL and solved using the commercial CPLEX solver. We tested the model for analyzing preliminary results to show the value of using SP in planning under uncertainty and the cost of the information. Then, we applied the model to a real case study in the North Shore region of the province of Québec (Côte-Nord) and compared deterministic and Stochastic Optimization (SO) methods with standard metrics for their evaluation. More precisely, we compute the Expected Value with Perfect Information (EVPI) and Value of Stochastic Solution (VSS) parameters, to analyze whether the method of Stochastic Programming is adequate for the project and the cost of the quality of the information and when we do not consider uncertainty. The optimization models offer better decision-making in forest management, reduce costs, increase the value in the entire chain and loss of trees as Spruce Budworm can lead to future outbreaks. Finally, we suggest some insights of the uncertain parameter that can affect the results of the optimization models and explain some suggestions that could improve the model if other attributes are included in the harvesting planning and the relevance of including other uncertainty parameters in forest planning.

Date	5 Oct 2017
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Mustapha Ouhimmou (Supervisor) & Mikael Rönnqvist (Co-supervisor)