Comparison of parameterizations for the atmospheric extinction coefficient in Lin'an, China

Abstract

A more precise atmospheric extinction coefficient would be advantageous for improving air quality (visibility) forecasting. In this study, the size distribution, chemical composition and relative humidity were measured in Lin'an from January 9 to 31, 2015. The merits and weaknesses of three parameterization schemes are discussed in this paper, including the non-linear fitting scheme, the IMPROVE (Interagency Monitoring of Protected Visual Environment) algorithm and the κ-elemental carbon (EC)-Mie model. Comparing the three schemes mentioned above, we find that the non-linear fitting equation requires the least amount of data, and its calculation process is the simplest. However, its calculated values are significantly influenced by specific data and fitting formulas. The uncertainty of the variable coefficients makes it difficult to directly implement this method for other datasets. The calculated values of the three versions of the IMPROVE algorithm strongly correlate with the measured values, with slopes near 1.0 and statistical indexes (R2) of 0.848, 0.858 and 0.866. However, this method is affected by the chemical compositions of the particles in different regions; for example, when the quality of PM2.5 is reconstructed from the measured data, the coefficients of different components change, thus affecting the final results. The κ-EC-Mie model objectively reflects the changes in the law of visibility, as the model considers both the chemical composition and size distribution of particles, and its predominant merit is derived from the fact that it is calculated without an empirical formula, which may eliminate the computational errors caused by the uncertainties of coefficients.