Tuesday, 30 January 2024: 4:30 PM
339 (The Baltimore Convention Center)
Wildfires are a major natural emission source of aerosols and trace gases in the atmosphere, leading to air quality degradation and adverse effects on human health. The complexities of wildfire emissions' quantification, coupled with the interplay of chemical transformations and the atmospheric transport of emitted compounds, make the precise prediction of wildfire air quality impacts a formidable challenge. Anticipating future wildfire emissions on a sub-seasonal to seasonal (S2S) time scale is significant for effective wildfire air quality forecasting and decision-making. In this study, we combine fire records and weather and climate data to estimate Week 3-4 wildfire emissions using statistical and machine learning (ML) models. For the statistical approach, we use the cumulative distribution function (CDF) of the normalized fire weather index (FWI) and vapor pressure deficit (VPD) along with the day 0 Fire Radiative Power (FRP) to predict FRP for the existing fires over the span of 7-30 days. We compare the accuracy of these FRP predictions against static methods like persistent FRP and consistently decreasing FRP, further evaluating their precision. Notably, our FRP predictions using the fire index CDF exhibit higher accuracy compared to that from the static FRP prediction methods, reducing model error by 19.23% during the 2020 Gigafire period. For the ML approach, we utilize the fire index, FRP, terrain data, meteorology data, and climate indices to predict the FRP S2S changes for each fire. The data from 2020-2021 are used to train the ML model. The predicted FRP was used to estimate the change in PM2.5 emission, which was then used as an input of the CMAQ model to study the impact of the different emission prediction methods on surface PM2.5 variations at the S2S time scales. The predicted PM2.5 concentration was compared to the AirNow observations. By implementing this advanced wildfire emission forecast model, we enhance wildfire air quality predictions, offering a practical avenue to bolster decision-making support and safeguard public health.

