A New CCN Number Concentration Prediction Method Based on Multiple Linear Regression and Non-Negative Matrix Factorization: 2. Application to Obtain CCN Spectra in and Around the Korean Peninsula

Abstract

The new cloud condensation nuclei (CCN) number concentration (N-CCN) prediction method that combines multiple linear regression and non-negative matrix factorization, hereafter the MLRNMF method, to the aerosol number size distribution data is evaluated and applied to the Korea-United States air quality (KORUS-AQ) airborne measurement data. The mean fractional bias (MFB) and the mean fractional error (MFE) were used as performance metrics. The performance metrics of the MLRNMF method for the test data set at several supersaturations (S) (0.4%, 0.6%, 0.8%, and 1.0%) met the performance goal values (MFB and MFE being less than or equal to +/- 30% and +50%, respectively), and thus, it implies that the MLRNMF method is appropriate at various S. Applying the MLRNMF method to KORUS-AQ airborne measurement data, it was demonstrated that aerosols in Seoul, Korean Peninsula, East Sea, and South Sea had similar hygroscopicity to each other, and aerosols over the Yellow Sea and the East China Sea showed similar hygroscopicity values during the KORUS-AQ field campaign. Based on the MLRNMF method, two parameterizations of CCN spectra for a full S range representing various regions in and around the Korean Peninsula in springtime were formulated (Twomey relationship and logarithmic fitting curve). Additionally, it was found that N-CCN decreased exponentially with altitude, and the scale height was approximately 1,700 m in the Seoul metropolitan area. The MLRNMF method can be used to create a global distribution map of N-CCN or CCN spectra in many regions around the world.