Model evaluation should be a component of the model development process, leading to a better understanding of model behavior and an increase in its credibility. In this paper a model evaluation protocol is proposed that encompasses five aspects: 1) model conceptual validity, 2) data requirements for model validation, 3) sensitivity analysis, 4) predictive validation (incorporating statistical tests), and 5) model comparison. The proposed protocol was applied to evaluate fire behavior models that were developed to predict crown fire initiation and spread with potential application in fire management decision support systems. The evaluation protocol highlighted the limitations and the distinct behavior of the specific models and the implications of such limitations when applying those models to support fire management decision-making. The model limitations identified through these results helped the authors to characterize deficiencies in the state-of-knowledge of the determinant processes involved in crown fire behavior, thereby identifying pertinent research needs.
[This publication is referenced in the "Synthesis of knowledge of extreme fire behavior: volume I for fire managers" (Werth et al 2011).]