only you (and your optimization model) can prevent forest fires

This month’s OR blog theme is the environment. I am going to write about forestry, a topic that was reviewed by Andres Weintraub and Carlos Romero in their Interfaces paper “Operations research models and the management of agricultural and forestry resources: A review and comparison.”

The US Forest Service has been using linear programming models to plan long-range harvesting and tree rotations (FORPLAN, SPECTRUM, Timber RAM). A search on their web site yielded a few OR papers that go back to the 1960s. Forestry models can have sustainability constraints, where sustainability ensures that the harvest rate does not exceed the rate of natural regeneration and that the harvesting costs that do not exceed the rate of financial growth.  Mixed integer programs have been used to model a wide range of shorter-range problems, such as balancing harvesting with road building, how to perform weekly harvests, how to match the timber supply with demand, locating harvesting machinery, and truck routing.

These forestry models balance efficiency with ecological issues such as biodiversity, wildlife, preservation, soil protection, water quality and scenic beauty.  (I’ve never added a “scenic beauty” constraint to a model. It sounds like fun!) Forest managers have to deal with ever more stringent constraints, such as protecting more habitat/protected areas, planning for habitat patches with no grown trees using adjacency problems (areas that are cut cannot be next to one another, leading to checkerboard-like cutting patterns), and requiring corridors of forest to remain to allow wildlife species cover, feeding, and foraging opportunities.

OR forestry models have succeeded for many reasons, including these:

• They are suitable for the forestry problems at hand
• Forestry needs efficiency due to global competition
• Managers are familiar with OR tools
• Forestry firms are large enough to benefit from the economies of scale
• Users have been involved in building models and implementing the results
• Collaboration between OR professionals, academics, and forestry experts

Forestry problems have different features than agriculture problems. The forestry problems are often longer-term and the data are often harder to obtain than in agriculture problems. In agriculture, problems are often solved for a current season, where the costs, prices, and yields can be reasonably estimated from past experience. This doesn’t hold for forestry problems, where the sources of uncertainty often stem from future prices, tree growth rates, and disasters.

Some agriculture problems are variations to the classical Diet Problem that determines how to cheaply feed farm animals blends of various food stuffs. Of course, this makes sense. I am always surprised when I find out that an idealized OR model that I’ve used in the classroom really makes a difference (see Mike Trick’s post on using the Secretary Problem to find love).

I have written indirectly about OR and the environment before:

• carpooling using social networking
• bicycle sharing and operations research
• OR and the food system
• local produce consumes more energy
• commuting to work using public transportation
• climate change needs operations research
• depleting fossil fuels