Reconstructing the Historical Frequency of Fire: A Modeling Approach to Developing and Testing Methods
© Joseph Gordon Fall 1998
SFU

Research project submitted in partial fulfillment of the requirements for the degree of master of natural resources management in the School of Resource and Environmental Management. Report No. 225.

Abstract

Fire is a prevalent natural disturbance in most of British Columbia's forest ecosystems. Recently, scientists and forest managers have recognized the important role fire plays in regulating forest ecosystems and maintaining biodiversity. In response, B.C. Government initiatives propose to use an ecosystem's historical disturbance dynamics for guiding forest management. Gaining an understanding of the methods used to estimate historical fire frequency, along with the limitations of these methods, and the sources of uncertainty and magnitude of bias in such estimates, will be critical for developing such ecosystem-based management objectives.

In Chapter 2, I review the published fire history literature, focusing particularly on the methods, underlying models, and calculations used to estimate historical fire frequency. This review is presented as an interactive tutorial, to aid a novice reader gain an understanding of some of the more difficult aspects of fire frequency reconstruction and interpretation. Some sample pages and a description of the tutorial are provided along with instructions on how to obtain the complete package.

All fire history studies rely on a series of inferences based on a set of physical evidence left by fire. This physical evidence contains inherent errors, most often of unknown magnitude. In addition, other errors are introduced when a researcher samples this evidence to create a data set, and estimates the history of fire occurrence from this data set. In Chapter 3, I present a methodology for quantifying the level of confidence that should be placed in an estimate of historical fire frequency made from tree-ring based fire interval data. In this approach, I use a spatial simulation model of the fire regime to generate synthetic fire histories. I propose and use new techniques to model the formation and survivorship of fire evidence in the tree-ring record. These models introduce errors into the synthetic fire histories based on the types of errors thought to be present in the physical data. Finally, a spatial model of fire history sampling is used to simulate errors introduced by the researcher.

I use Monte Carlo simulation to derive a confidence interval for the empirical estimate of fire frequency made in a recent fire history study. The results indicate that it is possible to reliably estimate historical fire frequency from fire interval data. However, the greatest source of uncertainty in this estimate is the probability with which fire evidence is formed in the tree ring record. This source of error has received little attention in the literature, and so I conclude by recommending that this problem be given serious study. In the mean time, I recommend that researchers minimize this source of uncertainty by collecting samples from several trees at each sampling point in the landscape.