Evaluation of a Lagrangian Air Quality Model against Various Observations and 3D Air Quality Models

Air quality regulations have reduced emissions of pollutants in the US, but many prognostic studies suggest that the future air quality in the US still can be degraded by global change factors, such as climate and land use. Three-dimensional Eulerian models are computationally expensive for the long-term simulations of the future air quality. To study how future air quality at a local scale will be influenced by these factors in an efficient way, we have developed a Lagrangian air quality modeling framework, called HYSPLIT-MOSAIC. It consists of the HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory), an air trajectory model developed by NOAA, and the MOSAIC (Model for Simulating Aerosol Interactions and Chemistry) model, a gas and aerosol chemistry and dynamics model developed at PNNL. To evaluate our framework, we compared our simulations against various observations and 3D Eulerian model outputs. Three-dimensional Eulerian models contain more comprehensive processes to provide more detailed air quality simulations than our Lagrangian model, so in this study we used them as another benchmark. Using 3D model simulations, it allows us to examine the error in input data, including the meteorology, emission data, initial and boundary conditions.

We have evaluated HYSPLIT-MOSAIC against CARES (Carbonaceous Aerosol and Radiative Effects Study) field campaign data from central California, EPA’s AQS measurements from 28 sites over California as well as WRF-Chem-MOSAIC simulations for the CARES period (June 2010). Additionally, we have been evaluating against 55 AIRNOW sites over the Pacific Northwest and the CMAQ-based AIRPACT5 air quality forecasting system, since November 2017. Overall HYSPLIT-MOSAIC results are comparable to the observations and 3D model results. It captures the diurnal patterns of ozone, although it overestimates by 10% to 50% respect to observations. Our model tends to underestimate PM2.5 concentrations compared to observations. For PM2.5 compositions, sulfate, ammonia and black carbon concentrations are predicted reasonably, but nitrate concentrations are overpredicted by up to 50% than observations. Because our framework does not include all the organic aerosol species, it underestimates their concentrations. The observed seasonal change of ozone and PM2.5 concentrations from winter to spring in 2018 are captured in HYSPLIT-MOSAIC.

Our evaluation results demonstrate that the HYSPLIT-MOSAIC framework is capable of predicting air quality at local-scale. Since it is computationally efficient compared to 3D Eulerian models, HYSPLIT-MOSAIC is appropriate for performing numerous ensemble long-term runs with multiple climate scenarios. For the future air quality assessment, we will utilize high resolution statistically downscaled climate data to provide the required meteorological data for our Lagrangian model.