On the determination of a cloud condensation nuclei from satellite: Challenges and possibilities

We use aerosol size distributions measured in the size range from 0.01 to 10+ μm during Transport and Chemical Evolution over the Pacific (TRACE‐P) and Aerosol Characterization Experiment–Asia (ACE‐Asia), results of chemical analysis, measured/modeled humidity growth, and stratification by air mass types to explore correlations between aerosol optical parameters and aerosol number concentration. Size distributions allow us to integrate aerosol number over any size range expected to be effective cloud condensation nuclei (CCN) and to provide definition of a proxy for CCN (CCN proxy ). Because of the internally mixed nature of most accumulation mode aerosol and the relationship between their measured volatility and solubility, this CCN proxy can be linked to the optical properties of these size distributions at ambient conditions. This allows examination of the relationship between CCN proxy and the aerosol spectral radiances detected by satellites. Relative increases in coarse aerosol (e.g., dust) generally add only a few particles to effective CCN but significantly increase the scattering detected by satellite and drive the Angstrom exponent ( α ) toward zero. This has prompted the use of a so‐called aerosol index (AI) on the basis of the product of the aerosol optical depth and the nondimensional α , both of which can be inferred from satellite observations. This approach biases the AI to be closer to scattering values generated by particles in the accumulation mode that dominate particle number and is therefore dominated by sizes commonly effective as CCN. Our measurements demonstrate that AI does not generally relate well to a measured proxy for CCN unless the data are suitably stratified. Multiple layers, complex humidity profiles, dust with very low α mixed with pollution, and size distribution differences in pollution and biomass emissions appear to contribute most to method limitations. However, we demonstrate that these characteristic differences result in predictable influences on AI. These results suggest that inference of CCN from satellites will be challenging, but new satellite and model capabilities could possibly be integrated to improve this retrieval.