A comparative study of nucleation parameterizations: 2. Three‐dimensional model application and evaluation

Abstract

Listen

Translate article

Following the examination and evaluation of 12 nucleation parameterizations presented in part 1, 11 of them representing binary, ternary, kinetic, and cluster‐activated nucleation theories are evaluated in the U.S. Environmental Protection Agency Community Multiscale Air Quality (CMAQ) modeling system version 4.4. The 12–28 June 1999 Southern Oxidants Study episode is selected as a testbed to evaluate simulated particulate matter (PM) number and size predictions of CMAQ with different nucleation parameterizations. The evaluation shows that simulated domain‐wide maximum PM2.5 number concentrations with different nucleation parameterizations can vary by 3 orders of magnitude. All parameterizations overpredict (by a factor of 1.4 to 1.7) the total number concentrations of accumulation‐mode PM and significantly underpredict (by factors of 1.3 to 65.7) those of Aitken‐mode PM, resulting in a net underprediction (by factors of 1.3 to 13.7) of the total number concentrations of PM2.5 under a polluted urban environment at a downtown station in Atlanta. The predicted number concentrations for Aitken‐mode PM at this site can vary by up to 3 orders of magnitude, and those for accumulation‐mode PM can vary by up to a factor of 3.2, with the best predictions by the power law of Sihto et al. (2006) (NMB of −31.7%) and the worst predictions by the ternary nucleation parameterization of Merikanto et al. (2007) (NMB of −93.1%). The ternary nucleation parameterization of Napari et al. (2002) gives relatively good agreement with observations but for a wrong reason. The power law of Kuang et al. (2008) and the binary nucleation parameterization of Harrington and Kreidenweis (1998) give better agreement than the remaining parameterizations. All the parameterizations fail to reproduce the observed temporal variations of PM number, volume, and surface area concentrations. The significant variation in the performance of these parameterizations is caused by their different theoretical bases, formulations, and dependence on temperature, relative humidity, and the ambient levels of H2SO4 and NH3. The controlling processes are different for PM number, mass, and surface areas. At urban/rural locations, some PM processes (e.g., homogeneous nucleation) and/or vertical transport may dominate the production of PM2.5 number, and emissions, or PM processes, or vertical transport or their combinations may dominate the production of PM2.5 mass and surface area. Dry deposition or some PM processes such as coagulation may dominate PM2.5 number loss, and horizontal and vertical transport, and cloud processes (e.g., cloud scavenging and wet deposition) may dominate the loss of PM2.5 mass and surface area concentrations. Sensitivity simulations show that the PM number and size distribution predictions are most sensitive to prescribed emission fractions of Aitken and accumulation‐mode PM and the assumed initial PM size distribution, in addition to different nucleation parameterizations.