Fire-growth modelling using meteorological data with random and systematic perturbations

Abstract

The focus of this investigation is to quantify the effects of perturbations in the meteorological data used in a fire-growth model. Observed variations of temperature, humidity, wind speed, and wind direction are applied as perturbations to hourly values within a simulated weather forecast to produce several forecasts. In turn, these are used by a deterministic eight-point fire-growth model to produce an ensemble of possible final fire perimeters. Two studies were conducted to assess the value of applying perturbations. In the first study, fire growth using detailed, one-minute data was compared to growth based on the more commonly used hourly data. Results showed that the detailed weather produced fire growth larger and wider than the hourly based data. By applying perturbations, variations in the flank and back-fire spread were captured by the random-perturbation model while the forward spread fell within the 20 to 30% probability prediction. A sensitivity analysis based on the observed variations showed that wind speed accounted for a 44% difference in area burned, while temperature accounted for only a 16% difference. In the second study, case studies were conducted on four observed forest fires in Wood Buffalo National Park. Results showed that daily fire-growth predictions using simulated weather forecasts over-predicted fire growth using actual hourly weather observations by 27%. Systematic-perturbation models best compensated for this with most fire growth falling within the predicted range of the models (52 out of 63 days).