Effect of Meteorological Variability on Fine Particulate Matter Simulations Over the Contiguous United States

Abstract

This study quantifies the impact of meteorological variability on the Community Multiscale Air Quality (CMAQ) model‐simulated particulate matter of aerodynamic diameter 2.5 μm or smaller (particulate matter 2.5 [PM2.5]) over the contiguous United States (CONUS). The meteorological variability is represented using the Short‐Range Ensemble Forecast (SREF) produced operationally by the National Oceanic and Atmospheric Administration. A hierarchical cluster analysis technique is applied to down‐select a subset of the SREF members that objectively accounts for the overall meteorological forecast variability of SREF. Three SREF members are selected to drive off‐line CMAQ simulations during January, April, July, and October 2016. Changes in emissions, vertical diffusion, and aerosol processes due to meteorological variability dominate changes in aerosol mass concentrations over 55‐73% of the domain except in July when dry deposition dominates emissions and aerosol processes. Weather Research and Forecasting‐Advanced Research WRF (WRF‐ARW) simulations reproduced the variability of surface temperature very well but overestimated the 10‐m wind speed, precipitation, and at some sites the planetary boundary layer height. Averaged over CONUS, CMAQ simulations driven by all three meteorological configurations capture the observed daytime low and nighttime high PM2.5 mass concentrations but underestimated the observed concentrations likely due to faster advection and higher wet deposition in the model. PM2.5 levels across the three simulations agreed well during daytime but showed larger variability during nighttime due to dominance of aerosol, clouds, and advection processes in nighttime. The meteorology‐induced variability in PM2.5 is estimated to be 0.08–24 μg/m3 over the CONUS with larger variability over the eastern United States.

A Novel Ensemble Design for Probabilistic Predictions of Fine Particulate Matter Over the Contiguous United States (CONUS)

“…This indicates that the overall ensemble performance does not degrade from the first to the second day of forecast. This is likely because of the relatively stable performance of the three meteorological members as shown in Kumar, Lee, et al (2019) as well as nearly constant anthropogenic emissions between the two forecast days.…”

“…Replacement of NEI anthropogenic emissions with the ECLIPSE and EDGAR‐HTAP emissions increases surface PM 2.5 mass concentrations almost everywhere in the domain in all the months with the highest increase during January. Minimum enhancements due to anthropogenic and fire emissions perturbations are seen in July, which is likely due to stronger mixing of emissions within a deeper planetary boundary layer and higher deposition during July (Kumar, Lee, et al, 2019).…”

“…This study simulates 24 ensemble members using the CMAQ model Version 5.1 (United States Environmental Protection Agency [EPA], 2015). The meteorological fields are simulated using the Weather Research and Forecasting (WRF) model (Powers et al, 2017; Skamarock et al, 2008) Version 3.8.1 on a 481 × 369 × 43 domain that covers the contiguous United States (CONUS) at a horizontal grid spacing of 12 km × 12 km, a model top located at 50 hPa, and a time step of 20 s. The physical parameterizations for different WRF configurations are described in detail in Kumar, Lee, et al (2019) and briefly in section 2.2.1.…”

“…The SREF system is composed of 26 ensemble members, evenly divided between the Advanced Research WRF (WRF‐ARW) and the NOAA Environmental Modeling System (NEMS) NMMB models. For each set of 13 NMMB and 13 WRF‐ARW members, a control member is associated with 12 perturbed members (see Tables S1 and S2 of Kumar, Lee, et al, 2019). The SREF obtains the initial conditions from one of three sources: The North American Model (NAM) Data Assimilation System (NDAS) (Kleist et al, 2009; Rogers et al, 2009), the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) (Environmental Modeling Center, 2003), and the Rapid Refresh model (RAP) (Benjamin et al, 2016).…”

“…Uncertainties in the simulation of meteorological parameters arise from errors in initial and boundary conditions, and parametrization of atmospheric processes and land‐surface interactions in the models. Our ensemble design uses the meteorological variability simulated by the regional Short‐Range Ensemble Forecast System (SREF) produced operationally by NOAA (Du et al, 2015) as representative of the uncertainties in the meteorological parameters (see section 2.2 of Kumar, Lee, et al, 2019, for details). Uncertainties in emission inventories arise from both the structural inaccuracies (e.g., spatial and temporal aggregation, missing emission sources, and inadequate knowledge of relationship between human activities and emissions) and input data sets inaccuracies (e.g., activity data, emission factors, and combustion efficiency).…”