Improvements of NAQFC PM2.5 predictions using two different bias correction methods

Particulate matters with diameter less than 2.5 µm (PM2.5) are of great concern to air quality in the U.S. However, PM2.5 predictions provided by the NOAA National Air Quality Forecasting Capability (NAQFC) are still in the developmental stage. Reduction in seasonal and diurnal biases must be achieved before NAQFC PM2.5 forecasts become operational. The NAQFC consists of the NOAA NCEP regional operational weather forecasting model, the Non-hydrostatic Multi-scale Model on the Arakawa staggered B-grid (NMMB) and the EPA Community Multiscale Air Quality (CMAQ) model. Recent major upgrades of NAQFC include (1) projected emission based on the National Emission Inventories (NEI) in year 2011, (2) refinement of CMAQ model vertical grid resolution, (3) dynamic update of lateral boundary conditions for CMAQ dust aerosol species by using the real-time NEMS GFS Aerosol Component (NGAC) dust forecasts, and (4) implementation of bias corrections for PM2.5 predictions. In this presentation, we examine two different bias correction approaches and their applications in the NAQFC PM2.5 predictions. These two approaches are the Analog Ensemble (ANE) and the Kalman-filter combined with historical forecast analogs (KFAN). Both ANE and KFAN demonstrate substantial enhancement of PM2.5 predictions. We evaluate the performance of these two methods on the diurnal pattern and day-to-day variability of PM2.5 for different seasons and regions. We discuss the impacts of the number of analog ensemble and the optimal training period on both the ANE and KFAN capabilities for capturing daily variability of PM2.5 forecasts. Furthermore, we present detailed analyses of episodes with large dust and wildfire smoke concentrations when the methods showed deficiencies.