A Novel Ensemble Design for Probabilistic Predictions of PM2.5 for the NAQFC

Poor air quality (AQ) in the U.S. is estimated to cause about 60,000 premature deaths with costs of $100–$150 billion every year. To reduce such losses, the National AQ Forecasting Capability (NAQFC) at the National Oceanic and Atmospheric Administration (NOAA) produces forecasts of ozone, particulate matter less than 2.5 mm in diameter (PM_2.5), and other pollutants so that advance notice and warning can be issued to help individuals and communities limit the exposure and reduce air pollution-caused health problems. The current NAQFC, based on the U.S. Environmental Protection Agency Community Multi-scale AQ (CMAQ) modeling system, provides only deterministic AQ forecasts and does not quantify the uncertainty associated with the predictions, which could be large due to the chaotic nature of atmosphere and nonlinearity in atmospheric chemistry.

This project aims to take NAQFC a step further in the direction of probabilistic AQ prediction by exploring and quantifying the potential value of ensemble predictions of PM_2.5, and perturbing three key aspects of PM_2.5 modeling: the meteorology, emissions, and CMAQ secondary organic aerosol formulation. This presentation focuses on the impact of meteorological variability, which is represented by three members of NOAA’s Short-Range Ensemble Forecast (SREF) system that were down-selected by hierarchical cluster analysis. These three SREF members provide the physics configurations and initial/boundary conditions for the Weather Research and Forecasting (WRF) model runs that generate required output variables for driving CMAQ that are missing in operational SREF output. We conducted WRF runs for January, April, July, and October 2016 to capture seasonal changes in meteorology. Estimated emissions of trace gases and aerosols via the Sparse Matrix Operator Kernel (SMOKE) system were developed using the WRF output. WRF and SMOKE output drive a 3-member CMAQ mini-ensemble of once-daily, 48-h PM_2.5 forecasts for the same four months. The CMAQ mini-ensemble is evaluated against both observations and the current operational deterministic NAQFC products, and analyzed to assess the impact of meteorological biases on PM_2.5 variability. Quantification of the PM_2.5 prediction uncertainty will prove a key factor to support cost-effective decision-making while protecting public health.