Fire growth modelling at multiple scales

摘要

This paper presents a method to model fire growth over multiple time scales. These scales correspond to the confidence levels that can be placed in forecasted weather products, changing the model from deterministic to probabilistic as the range of the weather forecasts increases. Fire growth is modelled through the various stages in sequence, providing probability maps of potential fire extents. The shortrange model predicts fire growth on an hourly scale using a deterministic, eight-point fire growth model. The medium-range model predicts fire growth over several days. This model uses the same fire-growth engine but introduces forecast reliability through the use of ensemble techniques. The long-range model predicts fire growth on a scale of weeks. This model combines the probabilities of fire spread and of survival to produce a probable fire extent map. A case study of a large fire in Wood Buffalo National Park, Canada, is conducted to illustrate the use of the model.