As a comparison, the Dijkstra (1959) algorithm employed in the Kourtz-0' Regan 

 (1971) fire spread model locates the paths of least resistance based on delay times 

 dependent only on the fuel descriptors of each individual cell, without dependence 

 on an adjacent cell. As each cell is reached--ignited--by the fire, the lapsed time 

 since fire starting time is recorded in the cell. Isochrones can then be drawn to 

 illustrate the progress of the fire. The two models are similar except for the 

 differences in delay time mentioned above. The Dijkstra model does not require up- 

 dating since it maps the path of least resistance to fire spread. Presently the 

 Kourtz-0 ' Regan model uses very large cells and average fuel parameters to describe 

 the average rate of spread and thus the delay time to consume the fuel cell. The 

 hexagonal model operating on cells that are subunits of the larger Kourtz-0' Regan 

 cell then can provide a distribution of rate of spread values needed to compute 

 the probable time a fire takes to consume a cell in the Kourtz-0' Regan model. Thus, 

 data obtained from the hexagonal model could be used as input to the Kourtz-0 ' Regan 

 mo de 1 . 



For a continuous fuel array the fireline intensity is the product of the reac- 

 tion intensity and the combustion zone depth (Albini 19 76) and assumes a constant 

 reaction intensity throughout the combustion zone. For the nonuniform array the 

 reaction intensity is assumed constant throughout the cell but may vary from cell 

 to cell. 



Following a suggestion by Frank Albini of the Northern Forest Fire Laboratory, 

 the fireline intensity is obtained by summing the products of the reaction intensity 

 and that portion of the cell contributing to the combustion process. Each column is 

 scanned perpendicular to the initial fire front (fig. 1) : 



1 



where I is the column fireline intensity, (Ij,)- is the reaction intensity of the 



th R 1 



i cell, F^ (appendix II) is that fraction of the i cell that is contributing to 



the combustion process, and D is the cell width. All column intensities then are 

 grouped to form a frequency distribution. 



The sum of the products, ^F.D, is the combustion zone depth if the fire is 



i 



burning perpendicular to the initial line of fire. Occasionally, portions of the 

 fire front may be burning to the side. Scanning down the columns would then give 

 some combustion zone depths that are unreasonably high. This results in spikes 

 in the distribution at high fireline intensities that can be easily located and 

 disregarded. 



6 



