Assessing PM_2.5 Model Performance for the Conterminous U.S. with Comparison to Model Performance Statistics from...

Area of Science:

• Atmospheric Chemistry and Physics
• Air Quality Modeling
• Environmental Science

Background:

• Previous studies proposed using historical photochemical grid model (PGM) performance statistics for benchmarking new applications.
• A significant challenge is the lack of consistently calculated, spatially and temporally resolved performance statistics across the U.S.

Purpose of the Study:

• To calculate a consistent set of model performance statistics for PM2.5 and its components over the U.S. from 2007-2015.
• To use this multi-year dataset to quantitatively assess the performance of the 2015 Community Multiscale Air Quality (CMAQ) model simulation.
• To identify limitations in the benchmarking approach using historical model performance data.

Main Methods:

• Calculated annual, seasonal, regional, and network-specific performance statistics for PM2.5 and key components using CMAQ model versions 4.7.1-5.2.1.
• Compared 2015 simulation performance statistics against the derived multi-year (2007-2015) dataset.
• Conducted sensitivity simulations to assess the impact of specific parameterizations on PM2.5 concentrations.

Main Results:

• PM2.5 organic carbon performance in the 2015 simulation was favorable compared to the multi-year dataset, attributed to improved biogenic secondary organic aerosol and mixing parameterizations.
• Northwest region performance in 2015 was unfavorable for several species, suggesting a need for better wildfire emissions speciation and boundary layer modeling.
• Benchmarking is limited by wide regional/seasonal variations in statistics and temporal trends, indicating past performance may not be a reliable reference for recent years.

Conclusions:

• A consistent, multi-year dataset of PM2.5 model performance statistics is crucial for robust benchmarking.
• Regional and temporal variability necessitates a nuanced approach to model performance evaluation, moving beyond simple national benchmarks.
• Model parameterizations for ammonia, organic aerosol, and crustal cations significantly influence PM2.5 predictions, underscoring the need for accurate input data and process understanding.