Review of Methods for Developing Probabilistic Risk Assessments. Part 1: Modeling Fire

D.A. Weinstein and P. B. Woodbury

Cornell University

The USDA Forest Service has recognized a need to develop integrated approaches to assess the probable effects of multiple stresses.  As part of this effort we conducted a state-of-the-science review of probabilistic regional risk assessment methodologies.  The goals of this review were to: (1) Describe methodologies currently in use, identifying the methods that are capable of evaluating the threats to ecosystems from fire and fuels, invasive species, loss of open space, unmanaged outdoor recreation, and other key stresses;  (2) Evaluate the usefulness of these methodologies for the Forest Service, including the advantages and disadvantages of each of these methods; and (3) Provide preliminary evaluation of the available databases as sources for these methodologies.  This paper presents the conclusions of this analysis, highlighting methods useful for evaluating the risk to fire as an example.  A companion paper presents the results of our survey of methods available for evaluating the risk of invasive species.

Much effort has gone into creating a capability of predicting fires throughout the region, both in their likely location and frequency. To create this capability, fire modeling systems have been established using a fine-scale grid of data on the landscape, such as fuel loads, vegetation, and climate trends. For example, LANDFIRE is a system that has been adopted by the Forest Service for assessing the risk of fire throughout the U.S. LANDFIRE depends heavily for this assessment on well-tested models such as FARSITE. 

The great proliferation of fire modeling systems in different portions of the U.S. suggests that each has specific strengths in simulating fires in the area for which the model was originally designed. Systems that can be applied in many different areas have obvious advantages.  However, it is also useful to have the predictions made by a system include information on the distribution probabilities of fires of different sizes, intensities, and the heterogeneity of fire types at any given location.  Not all systems are capable of providing this information.  Further, there is an increasing need for flexible classification of forest types in order to be able to assess risks across a number of stresses at a given location.  Perhaps most importantly, systems must be designed to track changes in fire susceptibility as climate changes.  Without this capability, it is unclear whether the relationship between vegetation, fuel loadings, and fire that will be shaped by future climates will be accurately predicted.  A modeling system such as MAPSS has a much higher likelihood of being able to track such changes in relationships.  We identified the large number of models capable of being used to track changes in vegetation and their resulting effect on changes in fire frequency.

Maintaining so many different types of models might be unwieldy and confusing to potential users.  However, we strongly advocate that fire risk be estimated by a number of fire models run in parallel.  If different models, especially those using different approaches and different data, predict similar patterns of risk, it will increase confidence in these predictions and make them more useful for management decisions.

Wednesday Morning Plenary

corresponding author:

David Weinstein
Cornell University
Department of Natural Resources
8 Fernow Hall
Ithaca NY, 14853
607-351-4214
daw5@cornell.edu