Fire size distribution parameters (mu, sigma) for Dynamic Fire Extension

[Brian Miranda]

Dynamic Fire Extension users,

I have recently started using a new approach to estimating the fire size distribution parameters (mu and sigma) and thought I would share the method with the community in case others would find it useful.  These parameters can be difficult to interpret without thorough understanding of the lognormal distribution.  For information on the lognormal distribution, and what these parameters are, I find the Wikipedia entry quite helpful (http://en.wikipedia.org/wiki/Lognormal_distribution).

The new method I am using is to fit a lognormal distribution to the historical fire size records within R, using the packages 'fitdistrplus' and 'truncdist'.  The attached text file contains code for R for performing this fitting.  The code I use fits a truncated log-normal distribution because our fire data almost always has a minimum fire size (tiny fires don't get reported) and sometimes has a maximum size.  The summary of the fitted model provides parameter estimates for mu (meanlog) and sigma (sdlog), along with standard errors of those estimates.

I find this method preferable to simply taking the mean and standard deviation of the log-transformed fire sizes because it just seems to work better, plus it provides the standard errors of the parameter estimates which can be incorporated as variability among simulation replicates if desired.  I am also attaching a sample data set that can be used with this script.

Feel free to contact me with any questions.

-Brian