A Data Assimilation Approach for Improved Operational Wildland Fire Forecasting using WRF-Fire

In this study, a computationally efficient data assimilation method suitable for operational wildland fire simulation is introduced and validated to ingest observed fire perimeters into a coupled wildland fire-atmosphere wildland fire simulation platform. Accurate wildfire simulation poses challenges due to multiple modelling uncertainties and errors. This is problematic because accurate fire-progression simulations are needed for pre-ignition mitigation and preparedness as well as post-ignition emergency response, especially considering the recent upward trend in wildfires frequency and severity in the US. Data assimilation is an effective approach to reduce modelling uncertainties and errors by integrating the models with real-world observations. In this study, an operational assimilation method is introduced in which the simulated fire is reignited from an observed fire perimeter to improve the forecasting accuracy. This method is applied to the 2018 Camp Fire simulation using the WRF-Fire coupled fire-atmosphere simulation platform in order to assess the effectiveness this method. First, a baseline case based on the current operational model setup used by Colorado Fire Prediction System is exploited to simulate the Camp Fire, showing non-negligible differences in fire propagation compared to observations. Next, the semi-continuous high-resolution Camp Fire perimeters derived from NEXRAD weather radar observations are used to assimilate the fire perimeter at different times. The forecast results after the assimilation are compared to the radar-driven fire perimeters in terms of the simulated fire perimeter, similarity with the observations, and time of arrival to several benchmark points. While maintaining the required computational demand for operational application of WRF-Fire, the proposed data assimilation resulted in more accurate fire propagation compared to the baseline cases in all evaluated benchmarks.