FIREWX Tuning Wildfire Plume Models from Calibrated Imagery

Smoke plumes pose one of the largest threats in wildfire emergencies due to their capacity to carry embers and particulate matter over large distances, emphasizing the importance of having 3-dimensional models and predictions of smoke plumes. Previous studies have established methods to generate such 3-dimensional models through simulation of these smoke plumes, taking in inputs such as wind, fire perimeter, and vegetation. However, many of these methods have encountered real-world limited accuracy due to a lack of information on input vectors such as wind or temperature, which significantly impacts plume prediction accuracy.

This study addresses these limitations by matching forecasts of smoke plume with observed imagery through optimizing input vectors. First, smoke plumes are generated with the Navier-Stokes equations based on historical fire perimeters with known weather vectors, and Levenberg-Marquardt is then run to fit the simulated smoke plume to the calibrated imagery by fine-tuning the input vectors. Such an optimization of the smoke plume model promises not only an increased accuracy of smoke plume prediction, but also provides crucial information on weather vectors that allows for other forms of forecast for emergency response.