Modeling PM2.5 Speciation Concentrations over California Using the MISR V23 Aerosol Product

Airborne fine particulate matter (PM) is the top environmental risk factor worldwide and has been linked to a variety of adverse health effects, including human respiratory disease, heart disease, and stroke. Ambient PM includes a complex mixture of particles with varying sizes, shapes, and chemical compositions. However, the relative toxicity of different PM types – mixtures of particles with different size distributions and compositions – is not well understood. To monitor PM, the US Environmental Protection Agency (EPA) has established PM ground monitoring networks, including monitors measuring PM₁₀ (particles with an aerodynamic diameter ≤10 μm), PM_2.5 (particles with an aerodynamic diameter ≤2.5 μm), and PM_2.5 species. However, the majority of US counties lack PM ground monitors, suggesting that the existing monitoring networks may not fully capture the complex spatiotemporal variability of PM concentrations and types.

Several recent studies have utilized remotely sensed aerosol products in combination with ground monitor data and other inputs in generalized additive models to improve the spatiotemporal resolution of ground-level particulate pollution predictions. In particular, NASA’s Multi-angle Imaging SpectroRadiometer (MISR) is an aerosol sensor aboard the Terra satellite that is capable of categorizing aerosols based on their size and shape.

This work builds upon previous studies that used the MISR aerosol product to model PM species (sulfate, nitrate, organic carbon, and elemental carbon) in southern California, with the aim of extending the results to the entire state. We use the newly reprocessed MISR V23 aerosol product, which provides 4.4 x 4.4 km² resolution aerosol optical depth (AOD) data partitioned into a prescribed set of aerosol particle models, in conjunction with PM_2.5 speciation measurements from the Chemical Speciation Network (CSN) and Interagency Monitoring of PROtected Visual Environments (IMPROVE) ground-monitoring networks as well as various land-use and meteorological inputs, to model speciated PM_2.5 over the state of California at yearly or finer temporal resolution. Preliminary results achieve an R² of 0.47 using AOD corresponding to a small, absorbing aerosol model, elevation, latitude, longitude, day of the year, and year inputs as statewide predictors of PM_2.5 elemental carbon. This is comparable to the performance of other studies over a small region of Southern California. This presentation will describe the model framework, the various input datasets that were tested, and the relative impact of each of the inputs on the model performance for several PM_2.5 species.