Reducing Data Uncertainty in Forest Fire Spread Prediction: A Matter of Error Function Assessment
Forest fires are a significant problem that every year causes important damages around the world. In order to efficiently tackle these hazards, one can rely on forest fire spread simulators. Any forest fire evolution model requires several input data parameters to describe the scenario where the fire spread is taking place, however, this data is usually subjected to high levels of uncertainty. To reduce the impact of the input-data uncertainty, different strategies have been developed during the last years. One of these strategies consists of adjusting the input parameters according to the observed evolution of the fire. This strategy emphasizes how critical is the fact of counting on reliable and solid metrics to assess the error of the computational forecasts. The aim of this work is to assess eight different error functions applied to forest fires spread simulation in order to understand their respective advantages and drawbacks, as well as to determine in which cases they are beneficial or not.
KeywordsError function Wild fire Prediction Data uncertainty
This research has been supported by MINECO-Spain under contract TIN2014-53234-C2-1-R and by the Catalan government under grant 2014-SGR-576.
- 5.Joint Research Centre: European forest fire information system, August 2011. http://forest.jrc.ec.europa.eu/effis/
- 6.Ebert, B.: Forecast verification: issues, methods and FAQ, June 2012. http://www.cawcr.gov.au/projects/verification
- 7.Finney, M.A.: FARSITE: fire area simulator-model development and evaluation. FResearch Paper RMRS-RP-4 Revised 236, Research Paper RMRS-RP-4 Revised (1998)Google Scholar
- 8.Gjertsen, U., Ødegaard, V.: The water phase of precipitation: a comparison between observed, estimated and predicted values, October 2005. https://doi.org/10.1016/j.atmosres.2004.10.030