Prediction Time Assessment in a DDDAS for Natural Hazard Management: Forest Fire Study CaseI

This work faces the problem of quality and prediction time assessment in a Dynamic Data Driven Application System (DDDAS) for predicting natural hazard evolution. In particular, we used forest fire spread prediction as acase study to show the applicability of the methodology. The improvement on the prediction quality when using a two-stage DDDAS prediction framework has been widely proved. The two-stages DDDAS has a first phase where an adjustment of the input data is performed in order to be applied in the second phase, the prediction. This paper is focused on defining a new methodology for prediction time assessment under this kind of prediction environments by evaluating, in advance, how a certain combination of simulator, computational resources, adjustment strategy, and frequency of data acquisition will perform, in terms of prediction time. Since the time incurred in the hazard simulation is a crucial part of the whole prediction time, we have defined a methodology to classify the simulator's execution time using Artificial Intelligence techniques allowing us to determine upper bounds for the DDDAS prediction time depending on the particular input parameter setting. This methodology can be extrapolated to any DDDAS for predicting natural hazards evolution, which uses the two-stage prediction scheme as a working framework