A model for rapid PM_2.5 exposure estimates in wildfire conditions using routinely available data: rapidfire v0.1.3

Area of Science:

• Environmental Science
• Public Health
• Atmospheric Science

Background:

• Wildfire smoke exposure events are increasingly frequent in the western US, impacting public health.
• Current methods for estimating smoke aerosol exposure are time-consuming and difficult to reproduce.
• The growing frequency of these events necessitates rapid exposure assessment tools.

Purpose of the Study:

• To introduce the rapidfire R package for rapid and reproducible wildfire smoke exposure assessments.
• To develop a tool that utilizes routinely generated, publicly available data for near real-time analysis.
• To provide high-quality, spatially resolved estimates of aerosol concentrations during smoke events.

Main Methods:

• The rapidfire R package integrates multiple data sources: air quality monitoring, satellite observations, meteorological modeling, smoke modeling, and low-cost sensors.
• A machine learning approach, random forest regression, is employed to fuse these diverse datasets.
• The package is designed for efficient data harvesting and processing within a month of an event.

Main Results:

• Estimates of ground-level 24-hour average particulate matter were generated for California wildfires between 2017-2021.
• The rapidfire package successfully produced spatially resolved aerosol concentration estimates.
• The generated estimates demonstrated excellent agreement with independent filter-based monitoring data.

Conclusions:

• The rapidfire R package offers a valuable tool for rapid and accurate wildfire smoke exposure assessment.
• This tool addresses the critical need for timely data to study public health impacts of smoke events.
• The integration of diverse data sources and machine learning provides a robust method for exposure estimation.