RESULTS AND DISCUSSION 



Good control over the fuel variables (see tables 2 and 3) was maintained by care- 

 ful construction of the plots and by obtaining high precision in estimating them. 



Rate of Fire Spread 



The mathematical model predicted rates of spread that were higher than observed 

 rates on all but three of the plots (fig. 4 and table 4). Deviations (predicted values 

 minus observed values divided by observed values) ranged from -14 to 580 percent. The 

 average overprediction (percent deviation) was 82 for the ponderosa pine plots and 121 

 for the Douglas-fir plots. These are in reasonably close agreement considering the 

 complexity of the mathematical model and the slash fuels. 



Several reasons beyond experimental error probably account for the deviations. 

 The consistent overestimate of the spread rates indicates bias that could be explained 

 by certain aspects of the mathematical model: (1) Assumptions about the fuel; (2) the 

 moisture of extinction chosen for the slash; and (3) the method of weighting input 

 parameters by fuel surface area. 



Some bias may have been introduced on the plots having higher intensity fires 

 (some of the 30-ton-per-acre plots) because the 8-foot plot width limited flame develop- 

 ment. However, this bias should be small. An analysis of the effects of plot width 

 on flame size and related rate of spread indicates that an 8-foot plot width would re- 

 tard only slightly the rate of spread for fires having flames greater than 4 feet in 

 length (Anderson 1968) . No retardation of spread rate was indicated for fires having 

 flames less than 5 feet in length. 



11 



