Sensitivity of the WRF-SFIRE Model to Changes in Fuel Moisture Content

The number and size of wildfires has been increasing over the past few decades due to climate changes resulting in longer fire weather seasons and increased burnable areas. These changes pose a significant threat to life, property, natural habitats, and result in increased economic costs from fire suppression efforts, fuel management, property loss, and other societal expenses. To address the negative impact of increased fire activity, advanced wildfire modeling can assist suppression crews, emergency planners, and public health officials in making more informed decisions.

However, wildfires are complex, multi-scale phenomena that are not fully understood, making modeling a challenging task. The scarcity of observational data about critical factors such as weather and fuel properties also adds to the difficulty, causing even the best wildfire models to produce inaccurate forecasts when initialized with incomplete and uncertain data. Understanding the effect of this noisy and incomplete data is crucial to evaluating the accuracy of forecasts generated by wildfire models.

In this presentation, we will investigate the impact of uncertain data on the WRF-SFIRE model. WRF-SFIRE is a coupled fire-atmosphere model that combines the Weather Research and Forecasting Model (WRF) with a spread model (SFIRE) implemented using the level-set method. Since atmospheric properties such as temperature, humidity, precipitation, and winds have an effect on fuel moisture content (FMC), WRF-SFIRE incorporates a fuel moisture component that can track changes to FMC during a fire simulation period. As a result, WRF-SFIRE is able to simulate behaviors such as diurnal variability in the fire rate of spread or fire extinction during periods of heavy precipitation.

This presentation will analyze the effect of variations in FMC when using WRF-SFIRE to simulate idealized and real-world wildfires. Simulations of real-world fires initialized with varying FMC will be examined to provide users of WRF-SFIRE with a better context to assess fire forecasts initialized using uncertain data. Under certain conditions, we will demonstrate that small changes to FMC can produce significant changes in the size of the modeled fire. Further analysis to help identify the source sensitivity in the model with respect to changes in FMC will be achieved by dividing the fuels in the fire domain into 1-hour, 10-hour, and 100-hour fuel classes and varying the respective FMC of each class independently.