A review of methods for introducing risk and uncertainty in forest

A review of methods for introducing risk and uncertainty in forest planning
Jordi Garcia-Gonzalo1, Maria Pasalodos-Tato2, José G. Borges1
1 Centro
de Estudos Florestais, ISA, Technical University of Lisbon, Tapada da Ajuda 1349-017 Lisbon (Portugal)
2 Centro de Investigación Forestal, CIFOR-INIA, Apdo. 8111, 28080 Madrid (Spain)
Email: [email protected]
Background
Forest planning is characterized by the long-term horizon of its
outcomes; uncertainty and risk are thus closely related to the
development of forest management plans due to:
Aims
Forest Management Decision
Support Systems (FORSYS)
The objective of this study is to present a review of different techniques
that have been used to address risk and uncertainty in forest planning
problems, and trying to give recommendations in regard to the
application of each method according to the problem case (Table 1).
Thus, the presentation of the review is problem oriented. This means
that based on the type of problems that forest manager may face we
provide a list of possible techniques that have been used or may be used.
Moreover, new approaches that may be useful for forest planning are
introduced.
•uncertainty related to the development of the trees
•the performance of timber markets
•the occurrence of catastrophic events, or even to the
•preferences of the forest owner.
Forest planning may be ineffective if it ignores these sources of
uncertainty and risk.
Methods
Table 1 Different problems that may be addressed when preparing
forest management plans.
The methods that address risk and uncertainty are grouped in
different categories attending to different dimensions of the
problem (i.e. temporal scale, spatial context, spatial scale,
decision making dimensions, objective dimensions and goods
and services dimensions), but also to different ways to
interpret risk and uncertainty (Table 1).
Dimensions
1
Temporal scale
Long term (strategic)
x
Medium term (tactical)
Short term (operational)
Spatial context
Non spatial
x
Spatial
Spatial Scale
Stand level
x
Forest level
Regional/national level
Decision making dimension
Unilateral
x
Collegial
Bargaining / participative
Objectives dimension
Single
x
Multiple
Goods and services dimension
Market non wood products
Market wood products
x
Market services
Non market services
Regarding temporal scale of management planning, three
levels are analyzed:
•Long term (strategic) refers to planning horizon
extending over more than 10 years.
•Medium term (tactical), refer to planning horizon
extending from 2 to 10 years.
Strategic
•Short term (operational ) refers to planning horizons
extending over one year or less.
Tactical
Spatial problems take into account the interactions of
decisions made in neighboring stands while non spatial means
that stands may be aggregated.
Operational
2
3
4
Cases
5 6
x
x
x
x
x
7
x
8
9
10
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
The different methods used to integrate risk and uncertainty in forest planning may be stochastic or deterministic.
Deterministic techniques may include the use of linear programming (LP) or dynamic programming combined with
sensitivity analysis or heuristics. Moreover, nonlinear programming may be used weighting the expected outcomes by its
probability of occurrence. Stochastic techniques may include mathematical programming, markov chains, dynamic
programming, stochastic simulation and the use of heuristics.
Results and Discussion
•Selection of harvesting machinery and its location which may be solved by using GIS and heuristics, LP
and IP.
•Tree bucking problem under uncertainty on transport costs and timber prices. This problem may be
solved by DP, LP and heuristics combined with scenario analysis technique.
STRATEGIC LEVEL
There are currently an impressive number of articles using different mathematical techniques to include
uncertainty and risk on forest planning. Tables 2 and 3 try to summarize some of them (not all). In
addition we have not found articles regarding some stand level problems, for example:
•Choosing harvest units to cut in certain period under climatic uncertainty. This problem may be solved
by using LP and scenario technique or stochastic simulation.
Table 3 Different methods used to integrate risk and uncertainty at forest planning level.
Simulation
(Deterministic)
Simulation (Stochastic)
Catastrophic events
Catastrophic events
Pasalodos-Tato et al. 2010, 2009, Thorsen and Helles
1998, Pasalodos-Tato and Pukkala 2008
González et al. 2005, Pukkala 1998, Moykkynen et al.
2000, González-Olabarría et al. 2008, Hyytiainen and
Haight 2009
Uncertainty in yearly
growth and catastrophic Valsta 1992, Pukkala and Miina 1997
events
Sensitivity analysis
Stochastic timber prices Pukkala and Miina 1997
(Scenario analysis )
Attitude towards risk
Pukkala and Kangas 1996
Preferences of the DM Pukkala and Miina 1997
Deterministic approach Catastrophic events
Martell 1980
Catastrophic events
Kao 1982
Stochastic dynamic
Kao 1984, Kooten et al 1992
programming (sensitivity Uncertainty in growth
analysis, stochastic
Preferences of the DM Couture and Reynaud 2008
simulation)
Stochastic prices
Lohmander 2007
Kaya and Buongiorno 1987, Buongiorno 2001,
Markov chains (scenario Uncertainty in growth
analysis technique,
Uncertainty in growth and Zhou et al. 2008, Insley and Rollins 2005, Rollin et al.
stochastic simulation)
prices
2005, Lohmander 2000, Lin and Buongiorno 1998
Scenario analysis or
simulation
LP
(including
lagrangian
relaxation
and fuzzy
set theory)
Probabilistic or chance–
constraint programming,
Fuzzy set theory
Robust optimization
Heuristic
SPATIAL
CONTEXT
Catastrophic events
Peter and Nelson 2005, Gassmann
1989, Boychuck and Martell 1996
Uncertainty timber growth
Hoganson and Rose 1987, Eriksson
2006
Post optimal or sensitivity Stochastic technological
analysis
coefficient
Scenarios of random
events
Scenarios of climate
change
TACTICAL LEVEL
Nonlinear programming
Dynamic programming
Optimize harvest scheduling (rotation length, thinning
regimes)
STRATEGIC/TACTICAL LEVEL
Table 2 Different methods used to integrate risk and uncertainty at stand level.
NO
SPATIA
CONTEXT
maximizatio
n of
sustained
harvest
volume
flows,
harvest
schedules,
the
maintenance
of
biodiversity
values
Risk as an objective
variable
Randomness in timber
growth
Pickens and Dress 1988
Pickens et al. 1991; Weintraub and
Vera 1991; Weintraub and
Abramovich 1995 ; Hof et al 1996
Uncertainty on production
Hof and Pickens 1991
requirements
Hof et al 1986; Pickens and Hof
Fuzzyness in timber yields
1991
Uncertainty in volume and
Palma and Nelson 2009
demand in harvests
Catastrophic event
Uncertainty in climatic
conditions
Risk of wind-throw
minimization or maxim.
Effect of minimizing
mean wind-throw risk
Wildfire simulators and the
Uncertainty in fuel
heuristic great deluge
management
algorithm
Preferences and the
HERO+ Priority function
attitude toward risk
Climate change
Reserve site selection
Catastrophic events e.g.
Markov chain processes.
risk of forest fire and
wind effects
Maintenance of wildlife
Meilby et al 2001
Garcia-Gonzalo et al. 2008
Zeng et al. 2007
Heinonen et al. 2009
Kim et al. 2009
Pukkala and Kangas 1996
Spring et al. 2005
Sabbadin et al. 2007
Spring and Kennedy 2005;Garcia
and Sabbadin 2001; Forsell et al.
2009
Spring et al. 2008
Innovative approaches: Three different groups of new approaches have been considered based on:
(i) fine tuning of current approaches already used: Using new methodologies of population based direct
search methods at stand level (Pukkala 2009). Using heuristics with continues variables. Using LP at
the stand-level by disaggregating the stand into raster cells.
(ii) addressing uncertainty related to preferences: Fuzzy set theory (Kangas et al. 2008) or stochastic
multicriteria acceptability analysis (SMAA). When several objectives are involved and are conflicting,
pareto-optimal or efficient technique may be used to help DM to set his preferences (Toth et al.
2002). Based on the portfolio technique, robust portfolio model has been developed in recent years.
(iii) other approaches: Information-gap, Uncertainty Features Optimization (UFO) and option analysis.
Conclusion
An assessment of each application in terms of its originality or its innovativeness is
beyond the scope of this paper, whose purpose, in fact, was to bring these applications to
the attention of forest management students, researchers and practitioners, as well as to
provide a general review and guide on which method to use according with the problem
faced by the decision maker.
Acknowledgments:
Part of this study was done in the context of a short term scientific mission (COST-STSM-FP0804-4839). In this context the authors would like to thank COST Action FP0804 Forest
Management Decision Support Systems (FORSYS). This research was also supported by Project PTDC/AGR-CFL/64146/2006 “Decision support tools for integrating fire and forest
management planning” funded by the Portuguese Science Foundation and by project MOTIVE - Models for Adaptive Forest Management, funded by 7th EU Framework Programme.
Lisbon. 21- 22 April 2010. Forest Management Decision Support Systems (FORSYS)