Global aerosol mixtures and their multiyear and seasonal characteristics
The optical and microphysical characteristics of distinct aerosol types in the atmosphere are not yet specified at the level of detail required for climate forcing studies. What is even less well known are the characteristics of mixtures of aerosol and, in particular, their precise global spatial di...
| Published in: | Atmospheric Environment |
|---|---|
| Main Authors: | , , , |
| Format: | Article in Journal/Newspaper |
| Language: | unknown |
| Published: |
Elsevier
2015
|
| Subjects: | |
| Online Access: | https://centaur.reading.ac.uk/77175/ https://doi.org/10.1016/j.atmosenv.2015.06.029 |
| Summary: | The optical and microphysical characteristics of distinct aerosol types in the atmosphere are not yet specified at the level of detail required for climate forcing studies. What is even less well known are the characteristics of mixtures of aerosol and, in particular, their precise global spatial distribution. Here, cluster analysis is applied to seven years of 3-hourly, gridded 2.5 x 2 degree aerosol optical depth data from the Goddard Chemistry Aerosol Radiation and Transport (GOCART) model, one of the most-studied global simulations of aerosol type currently available, to construct a spatial partition of the globe into a finite number of aerosol mixtures. The optimal number of aerosol mixtures is obtained with a k-means algorithm with smart seeding in conjunction with a stopping condition based on applying the ‘law of diminishing returns’ to the norm of the Euclidean distance to provide upper and lower bounds on the number of clusters. Each cluster has a distinct composition in terms of the proportion of biomass burning, sulphate, dust and marine (sea salt) aerosol and this leads rather naturally to a taxonomy for labeling aerosol mixtures. In addition, the assignment of primary colours to constituent aerosol types enables true colour-mixing and the production of easy-to-interpret maps of their distribution. The mean multi-year global partition as well as partitions deduced on the seasonal timescale are used to extract aerosol robotic network (AERONET) Level 2.0 Version 2 inversion products in each cluster for estimating the values of key optical and microphysical parameters to help characterize aerosol mixtures. On the multi-year timescale, the globe can be spatially partitioned into 10 distinct aerosol mixtures, with only marginally more variability on the seasonal timescale. In the context of the observational constraints and uncertainties associated with AERONET retrievals, bivariate analysis suggests that mixtures dominated by dust and marine aerosol can be detected with reference to their single ... |
|---|