Generation and evaluation of an ensemble of wildland fire simulations

Abstract

Numerical simulations of wildfire spread can provide support in deciding firefighting actions but their predictive performance is challenged by the uncertainty of model inputs stemming from weather forecasts, fuel parameterisation and other fire characteristics. In this study, we assign probability distributions to the inputs and propagate the uncertainty by running hundreds of Monte Carlo simulations. The ensemble of simulations is summarised via a burn probability map whose evaluation based on the corresponding observed burned surface is not obvious. We define several properties and introduce probabilistic scores that are common in meteorological applications. Based on these elements, we evaluate the predictive performance of our ensembles for seven fires that occurred in Corsica from mid-2017 to early 2018. We obtain fair performance in some of the cases but accuracy and reliability of the forecasts can be improved. The ensemble generation can be accomplished in a reasonable amount of time and could be used in an operational context provided that sufficient computational resources are available. The proposed probabilistic scores are also appropriate in a calibration process to improve the ensembles.

Novel method for a posteriori uncertainty quantification in wildland fire spread simulation

“…and the advance in time depends on ROS i , the rate of spread computed with the values of the environmental inputs at location x i and time t i , as follows: The description of the perturbation variables is presented in [13], where the distributions are mostly truncated normal. In the present study, we choose a wider support for some of the distributions and/or substitute them for uniform distributions.…”

“…The emulation and calibration procedure is applied to K = 7 fires that occurred in Corsica in 2017-2018 and that are presented in [13]. We also use the ensembles obtained in this previous study as reference for comparison and refer to them as "reference ensembles"…”

“…We evaluate the ensembles by following the methodology presented in [13]. We mainly focus on the accuracy of the ensembles by the means of the Brier skill score and a global version to summarize the performance of the ensembles on the seven fires.…”

“…[11,12]). In the case of probabilistic predictions, several evaluation criteria we proposed in [13] to assess the performance of the prediction system.…”

“…This methodology follows the setting presented in a previous study [13] where the un-certain inputs were identified. We try to obtain calibrated probabilistic predictions that are as relevant as possible based on the observation of 7 relatively big Corsican fires.…”