results of the previous study
The work of Flores-Garnica (2001) focused on the need for a comprehensive approach forest fuel mapping and estimation. The predominant fuel type occurring within the field area were fallen woody debris of varying size, placing the field area within the Timber Litter Group of fuel complexes following Anderson (1982).
Four fuel maps were generated based upon the distribution of live woody fuels, and upon the distribution of surface fuels according to their fuel lag time (1-Hr, 10-Hr, 100-Hr; Table I). The data were presented graphically with statistical data summaries. Five statistical interpolation methods: spline, inverse distance weighted (powers 1 and 2), polygonal mapping,and theissen polygons; and seven geostatistical interpolation methods: Ordinary Kriging, Universal Kriging (1st and 2nd degree), Cokriging, Point Kriging, and Block Kriging, were applied (Flores-Garnica 2001). Validation for the model results was based primarily upon the mean square of the errors (MSE) and the correlation coefficient (CC), with a secondary evaluation based upon Akaike’s Information Criterion (AIC) to insure modeling parsimony (Flores-Garnica & Omi 2003).
The three models based upon fuel lag time (1-Hr, 10-Hr, 100-Hr) were best estimated by Cokriging because of a strong cross-correlation between fuel distribution and elevation. Inverse Distance Weighted (IDW) predictions for the 10-Hr fuels provided the next best estimate, and IDW was found to be the most consistent statistical interpolation predictor of fuel loads (Flores-Garnica 2001). However, the follow-up study of Flores-Garnica & Omi (2003) concluded that Ordinary Kriging was the best linear unbiased estimator of fuel loads.
Finally, Flores-Garnica (2001) used the fuel distribution estimates to generate a spatial simulation model that was compared to the USDA Forest Service fire modeling algorithm FARSITE (Finney 1998).
