A Biphasic, Multi-spectral Extension of Dozier’s Algorithm for Wildfire Characterization and Implication for Smoke Forecasting

Accurate parameterization of wildfires is crucial for estimating biomass burning emissions and their vertical injections, which significantly impact fire weather forecasting and post-impact assessments. Previous studies using Dozier’s single-phase assumption have indicated potential errors in fire temperature estimation of up to 100 K, with corresponding errors in fractional area reaching 50%. In our research, we developed a biphasic multi-wavelength extension of Dozier’s algorithm to more realistically represent burning fires in two phases: flaming and smoldering. Our method utilizes multiple fire-sensitive channels from visible to thermal infrared, available on advanced sensors like the Visible Infrared Imaging Radiometer Suite (VIIRS) and Geostationary Operational Environmental Satellite (GOES). We employ the Markov Chain Monte Carlo method, a Bayesian inference technique, to estimate the posterior distributions of flaming and smoldering temperatures and their corresponding fractional areas. Initial testing through an Observing System Simulation Experiment (OSSE) demonstrated the capability of our method, revealing that a biphasic model adequately represents wildfire heterogeneities. Our estimates significantly reduce Fire Radiative Power (FRP) uncertainty to 3%, compared to the classical Wooster method's 25%, and accurately reconstruct energy fraction indices (visible to far-infrared and mid-infrared to far-infrared). This approach provides reliable posterior distributions of fire parameters, facilitating ensemble model runs to quantify uncertainty for downstream emissions. Applying the algorithm to the FIREX-AQ Williams Flat events, based on multichannel fire radiance outputs from the Fire Light Detection Algorithm version 2, showed that our estimates could pinpoint high-temperature fire zones that evolved into pyrocumulonimbus (pyroCB) events, thereby offering valuable insights for parameterizing the vertical distribution of plume injections. We will also show some preliminary results of applying the estimated fire parameters for smoke injection characterization in the GEOS forecast system.