Sensitivity of simulated aerosol properties over eastern North America to WRF-Chem parameterizations

Aerosol properties and their climatic feedbacks are characterized by high uncertainty in both global and regional model simulations. We explore sources of uncertainty in the representation of aerosol properties using an ensemble of simulations performed at 24-km resolution with WRF-Chem over eastern North America. The sensitivity of aerosol optical depth (AOD) and near-surface fine particle concentrations (PM2.5) to planetary boundary layer (PBL) and aerosol schemes (modal with secondary organic aerosol versus sectional but excluding secondary organic aerosol), as well as different emission inventories (National Emission Inventory [NEI] 2005 versus 2011) is examined. We quantify the spread among ensemble members with respect to the model setup and compute statistical metrics to identify the run configuration that exhibits greatest fidelity relative to observations of aerosol and meteorological properties. Use of the Modal Aerosol Dynamics Model for Europe/Secondary Organic Aerosol Model scheme leads to highest agreement with Moderate Resolution Imaging Spectroradiometer clear-sky AOD observations particularly when the 2005 NEI is used (with either PBL scheme). These members exhibit small negative mean fractional bias over the simulation domain (<2%) and relatively high spatial correlation in summertime mean monthly AOD (>0.5). The aerosol scheme and NEI dominate the ensemble spread in AOD. Near-surface PM2.5 is also dependent on PBL scheme and is best reproduced in runs adopting a sectional approach and emissions for 2011. Thus, WRF-Chem configuration associated with highest agreement with AOD observations is not the same as for PM2.5, possibly reflecting the importance of columnar water vapor in dictating AOD or other unexplored uncertainties linking surface mass concentrations to column optical properties.