| California Spotted Owls and Fire and Fuels Management in the Sierra
Nevada: An Example Application of Bayesian Modeling to Forest Planning
Presented Danny Lee, Pacific
Southwest Research Station
See web version of PowerPoint presentation
Abstract
The recent USDA Forest Service decision to simultaneously amend the forest
plans of 11 national forests in the Sierra Nevada is emblematic of current
resource management conflicts on public lands. Much of the controversy
surrounding the Sierra Nevada decision concerns the choice between
conservation and maintenance of high-quality habitat for California
spotted owls and the desire to manage forest structure and composition in
order to reduce the incidence and extent of high intensity wildfires.
California spotted owls preferentially select forest stands with high
canopy closure, multistoried canopies, and higher-than-average amounts of
snags and downed wood-features than can contribute to the intensity and
spread of wildland fires. Using a combination of mechanical thinning and
prescribed fire, fire risks can be reduced, but at a potential cost to the
suitability of the stand as spotted owl habitat. The challenge is to find
a treatment level that achieves an acceptable balance between fire risk
reduction and maintenance of owl habitat. This tradeoff can be modeled
using Bayesian decision analysis that incorporates a combination of
empirical data, model forecasts, and professional judgment. At the heart
of this analysis is an influence diagram and associated conditional
probability matrices. The influence diagram has three types of nodes:
decision nodes, chance nodes, and consequences. Arrows or arcs show the
causal dependence among various nodes; the strength of the dependence is
captured in the conditional probability relation. For the spotted owl
example, decision nodes represent the how, where, and when of various
treatments. Chance nodes depict features such as stand condition, wildland
fire, and owl survival productivity for which there is uncertainty about
their future state. A single consequence node integrates across possible
outcomes to calculate an expected utility measure for a given decision. My
presentation will focus on 3 steps in building and using influence
diagrams: 1) building and parameterizing the influence diagram using
various information sources, 2) a comparison of alternative treatments
where the optimal decision depends heavily on current fire risk and stand
conditions, and 3) using the influence diagram to assess the value of
information that might be gained by experimental management.
^ top |
