Predicting fire behavior in nonuniform fuel arrays is a problem requiring: 1. A method of assessing fuel nonuniformity, 2. A method of simulating fuel nonuniformity, and 3. An algorithm governing fire spread through a simulated array. Satisfying these requirements is the objective of this paper. The main concept is built around partitioning the fuel into a honeycomb array. Each cell is described independently according to its bulk fuel parameters (depth, load, average particle size, etc.). Field assessment is designed to meet the requirements of simulation. An algorithm simulates fire spread through the array by coupling predictions of heat flowing from a burning cell to predictions of the heat required for ignition of the adjacent cells. Ignition is allocated to the cell offering the least requirement for heat. Consequently, the fire moves nonuniformly through the array taking advantage of the path of least resistance. Methodology is emphasized. A simulated fire is initiated from a line source. Distortions in the propagating front result from fuel nonuniformities giving rise to a distribution of rates of spread rather than a single value. Analysis is appropriate for an assessment of a distribution of the fireline intensities. Examples are given for slash, residue after tree harvest, and a mixture of grass and sagebrush. Nominal windspeeds of 0 and 2 mi/h were chosen for the purpose of illustrating the technique for handling nonuniformity. Comparisons show that the previous alternative of combining all fuel to an average depth and load does not allow the land manager to assess the chance that patches of high risk fuel arrangements might result in unacceptable fire behavior.