The Sensitivity of WRF-Fire Predictions of Area, Location, and Propagation Direction to Changes in Ignition Point Location and Time

In 2018, wildfires burned over 8.7 million acres in the United States, resulting in over $3.1 billion in suppression costs. Effective, efficient, and safe wildland fire management can be bolstered by decision support systems that provide accurate predictions of fire spread characteristics such as fire area, location, and direction of propagation. Operational wildfire behavior forecasts generated using the WRF-Fire decision support framework are initiated using ignition point locations and times estimated by fire response teams. If the ignition point goes uninvestigated, or if miscommunication occurs during the reporting of investigated points, their timing and location may remain inaccurate, resulting in unknown uncertainty in model forecasts.

To better understand how ignition point estimates translate to model accuracy, we performed a sensitivity study to bound the uncertainty of the area, location, and direction of fire propagation of WRF-Fire simulations based on adjustments to the ignition point location and time. Our results show that uncertainty in ignition point does translate to uncertainty in model output, and predicted area, location, and direction of propagation are related to the ignition point's association with other model inputs such as wildland fuel and wind. These variations in model output provide an indication of the limitations of the model due to uncertainty in the ignition points. This study provides motivation and context for further discussion between scientists and wildfire responders on data needs and collaboration opportunities in decision support system development.