Uncertainty associated with model predictions of surface and crown fire rates of spread

The degree of accuracy in model predictions of rate of spread in wildland fires is dependent on the model's applicability to a given situation, the validity of the model's relationships, and the reliability of the model input data. On the basis of a compilation of 49 fire spread model evaluation datasets involving 1278 observations in seven different fuel type groups, the limits on the predictability of current operational models are examined. Only 3% of the predictions (i.e. 35 out of 1278) were considered to be exact predictions according to the criteria used in this study. Mean percent error varied between 20 and 310% and was homogeneous across fuel type groups. Slightly more than half of the evaluation datasets had mean errors between 51 and 75%. Under-prediction bias was prevalent in 75% of the 49 datasets analysed. A case is made for suggesting that a ?35% error interval (i.e. approximately one standard deviation) would constitute a reasonable standard for model performance in predicting a wildland fire's forward or heading rate of spread. We also found that empirical-based fire behaviour models developed from a solid foundation of field observations and well accepted functional forms adequately predicted rates of fire spread far outside of the bounds of the original dataset used in their development. We examined error statistics of operational wildland fire spread models.We compiled 49 fire spread model evaluation datasets involving 1278 observations.Mean percent error varied between 20 and 310% and was homogeneous across fuel type groups.The analysis suggests that a ?35% error interval is a reasonable standard for model adequacy.