Data that represent average worst fire weather for a particular area are used to index daily fire danger; however, they do not account for different locations or diurnal weather changes that significantly affect fire behavior potential. To study the effects that selected changes in weather databases have on computed fire behavior parameters, weather data for the northern Rocky Mountains were treated as probability distributions, then used in computer simulation to estimate distributions of rate-of-spread (ROS) and fireline intensity (FLI). Sensitivity of ROS and FLl to weather input changes was analyzed by varying the source and amount of weather data, and diurnally adjusting temperature and relative humidity. In eight representative cases, a minimum amount of data produced the lowest cumulative probabilities of ROS and FLl, and data from a higher elevation produced the highest values. For long-term planning, within the region studied, a small subset of weather data distributions was adequate for estimating probabilistic distributions of ROS and FLI. Joint probabilities of ROS and FLI differed substantially among test cases. Fire behavior values obtained with observed data were higher than those obtained with diurnally adjusted data. The simulation techniques used are appropriate for use in long-term fire management planning models.