Improving our understanding of individual wildfires by combining satellite data with fire spread modelling

Doutoramento em Engenharia Florestal e dos Recursos Naturais - Instituto Superior de AgronomiaWildfires pose real threats to life and property. In Portugal, the recent year of 2017 had the largest burnt area extent and number of casualties. A knowledge gap still exists in wildfire research related with better understanding individual wildfires, which has important implications for fire suppression, management, and policies. Wildfire spread models have been used to study individual wildfires, however, associated uncertainties and the lack of systematic evaluation methods hamper their capability for accurately predicting their spread. Understanding how fire spread predictions can be improved is a critical research task, as they will only be deemed useful if they can provide accurate and reliable information to fire managers. The present Thesis proposes to contribute to improve fire spread predictions by: i) Developing a methodology to systematically evaluate fire spread predictions ii) Thoroughly characterizing input data uncertainty and its impact on predictions; iii) Improving predictions using data-driven model calibration. The spread of large historical wildfires were studied by combining satellite data and models. The major findings of the present Thesis were: i) Satellite data accurately contributed to provide accurate fire dates and ignition information for large wildfires. ii) The evaluation metrics were very useful in identifying areas and periods of low/high spatio-temporal agreement, highlighting the strong underprediction bias and poor accuracy of the predictions. iii) Uncertainties in wind speed and direction, fuel model assignment and typology, location and timing of ignitions, had a major impact on prediction accuracy. iv) Predictions iii) Uncertainties in wind speed and direction, fuel model assignment and typology, location and timing of ignitions, had a major impact on prediction accuracy. iv) Predictions were improved by ‘learning’ from past wildfires, significantly reducing the impact of data uncertainty on the accuracy of fire spread predictions Overall, the work contributed to advance the body of knowledge regarding individual wildfires and identified future research steps towards a reliable operational fire system capable of supporting more effective and safer fire management decisions with the aim of reducing the dramatic impacts of wildfiresN/