Accounting for climate change in a forest planning stochastic optimization model

Abstract

An approach is proposed for incorporating the variations in timber growth and yield due to climate change uncertainty into the forest harvesting decision process. A range of possible climate scenarios are transformed by a forest growth and yield model into tree growth scenarios, which in turn are integrated into a multistage stochastic model that determines the timber cut in each future period so as to maximize net present value over the planning horizon. For comparison purposes, a deterministic model using a single average climate scenario is also developed. The performance of the deterministic and stochastic formulations are tested in a case study of a medium-term forest planning problem for a Eucalyptus forest in Portugal where climate change is expected to severely impact production in the coming years. Experiments conducted using 32 climate scenarios demonstrate the stochastic model's superior results in terms of present value, particularly in cases of relatively high minimum timber demand. The model should therefore be useful in supporting forest planners' decisions under climate uncertainty

A Progressive Hedging Approach to Solve Harvest Scheduling Problem under Climate Change

“…The planning horizon of the problem covers 15 years, where each stand must be harvested by the end of this horizon, satisfying even flow constraints that aim to have smooth flow deviations within consecutive years. This is a similar problem than presented by [22]. The main difference is that in the present study spatial integrity of the planning units (i.e., stands) was considered, which made the problem much more difficult to solve.…”

“…Based on the scenario tree presented in [22] that considered 32 scenarios and in order to cover and represent a wide range of climate variability for the study case area, other 9 different scenario trees were generated. The scenario trees differed in the data structure (i.e., number of nodes or children per period) while maintaining the probability distribution of the original tree [28].…”

“…Stochastic programming takes all the scenarios into account without being subordinated to any one of them allowing developing policies that lead to solutions with a better expected objective value and that can be feasible for all scenarios. Few studies have focused on the use of stochastic programming to address climate change uncertainty [22,23]. Garcia-Gonzalo et al [22] developed a stochastic linear programming model in a harvest scheduling problem to demonstrate the advantage of using a stochastic versus a deterministic model based on expected values.…”

“…Few studies have focused on the use of stochastic programming to address climate change uncertainty [22,23]. Garcia-Gonzalo et al [22] developed a stochastic linear programming model in a harvest scheduling problem to demonstrate the advantage of using a stochastic versus a deterministic model based on expected values. Álvarez-Miranda et al [23] presented a stochastic goal programming model to deal with climate change uncertainty and included the use of conditional value at risk.…”

“…In this paper, a spatial large-scale multi-stage stochastic harvest scheduling problem considering uncertainty due to climate change was solved. In order to address climate change uncertainty when developing the harvest scheduling model studied in this paper, the 32 climate scenarios presented by Garcia-Gonzalo et al [22] were considered for tree growth. Even though Garcia-Gonzalo et al 2016 [4] used the same case study area and the same number of scenarios as in the present study; they simplified the problem (i.e., reduced the size of the mode by aggregating 1000 stands into 24 macro-stands (i.e., strata).…”