Design of forest supply chain under uncertainty: the impact of spruce budworm infestation on the wood supply

Abstract

The forest industry is very important from both environmental and economic perspectives for Canada. In 2017, production in the forest sector contributed around $25 billion to Canada’s real gross domestic product (GDP) through 210 thousand direct and 107 thousand indirect jobs. However, millions of dollars are the cost of damage of invasive species to forest owners such as government, industries, and private citizens. Revenue losses, prevention and control investments, and environmental mitigation efforts have cost Canada hundreds of millions of dollars during the last years. Spruce budworm (Choristoneura fumiferana (Clemens)) outbreaks is a well known major natural disturbance in eastern Canada. It is one of the most destructive insects in North America’s conifer stands. Reduction in the wood supply is one major direct impact of insect outbreaks. As an example, in 2017, more than 7 million hectares were defoliated by spruce budworm in Quebec. Repeated defoliation causes tree mortality, reduction of growth rates, and reduced lumber quality. There are different control methods to protect forest against insects and diseases. Silvicultural control methods such as salvage harvesting and pre-emptive harvesting, are used to satisfy the forestry companies’ demand and chemical methods like spraying biological insecticide Bacillus thuringiensis ssp. kurstaki (Btk) is taken into account to maintain trees alive during large-scale infestation for later harvest. In my article, titled “Salvage Harvest planning for Spruce Budworm Outbreak using Multistage Stochastic Programming”, we considered the effect of changes of outbreak intensity on wood values throughout the forest as the wood infestation can change the lumber quality. Salvage harvesting considered as an action to mitigate the economic and environmental damages. We propose a multistage stochastic mixed-integer programming model for harvest scheduling under various outbreak intensities. The objective is to maximize revenues of wood value minus logistic costs while satisfying demand for wood in the industry. Results show that when there is an outbreak throughout the forest, the first priority for salvage harvesting is to focus on forest areas with the lowest level of infestation. The other article, titled “The integration of spraying and harvesting to minimize the wood losses during an outbreak of Spruce Budworm” uses two control techniques, spraying and harvesting, against spruce budworm defoliation in the forest. In each period, the estimation of wood volume for each stand is updated based on its feature attributes, history of the defoliation, and whether it has been sprayed or not. This study provides a deterministic model addressing the questions of where and when should be harvested or sprayed to maximize revenues of harvested wood minus logistic costs and maximize the value of standing trees at the end of the planning horizon while satisfying demand for wood in the industry and minimizing the forest protection costs. The model has been applied to a case study located in the Bas-Saint Laurent region in Quebec. The results show that the benefits of harvesting outweigh the benefits of spraying and the models prefer to harvest rather than spraying; however, it does not mean that spraying is not effective. Spraying is helpful but it is not economical in comparison with harvesting. Furthermore, stands which have the highest wood loss, in other words, they have a high proportion of BF and WS and the cumulative defoliation score is around the turning point of the cumulative mortality curve are elected for harvesting. Finally, stands which have a highdensity ratio (volume to the area) are economical choices for spraying. While studying strategic forest management models, we observed two mistakes in the original formulation in one of the well-known models called Model II if the minimum number of periods between regeneration harvests is overlooked. The first is a mistake in the area constraints and the second in calculating one important parameter of the model representing discounted net revenue per hectare between periods. We provide a revised model together with comments on the computations of a parameter used in the model formulation. Then, in order to validate the problem identified, we solve the Model II with realistic data to address the modeling mistakes and explain how our revised formulation works with the same data. We also describe situations where the mistakes may have a larger impact and explain why they have not been identified earlier. This study is the first article called “How the minimum number of periods between regeneration harvests induces modeling mistakes in the well-known Model II forest management”.

Date	26 Feb 2021
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Mustapha Ouhimmou (Supervisor) & Mikael Rönnqvist (Co-supervisor)