Rapidfire: PM2.5 Estimates for California Wildfire Smoke Events

A machine learning approach fuses air quality monitoring, satellite data, and meteorological models to produce spatially-resolved PM2.5 estimates. The rapidfire R package generated these estimates for several large wildfire smoke events in California from 2017 to 2021. This dataset and its accompanying scripts were created by Sean Raffuse at UC Davis to support the rapidfire manuscript.

Use Cases

• Train models to predict ground-level PM2.5 based on satellite observations and meteorological data.
• Analyze spatial patterns of smoke exposure during specific California wildfire events.
• Validate air quality estimates against independent filter-based network measurements.
• Study the correlation between wildfire smoke events and public health outcomes using high-resolution exposure data.

Strengths

• Estimates show excellent agreement with independent filter-based PM2.5 network data.
• Covers several large wildfire smoke events over a five-year period from 2017 to 2021.
• Integrates multiple routinely generated public data sources, including official monitoring, satellite observations, and low-cost sensor networks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to its focus on California and the western United States.

Provenance

Source

University of California, Davis

Collection Method

Data fusion using a random forests regression model on harvested air quality, satellite, meteorological, and smoke model data.

Time Range

2017-2021

Geography

California, western United States