Optimal Estimation Retrieval of Aerosol Microphysical Properties in the Lower Stratosphere from SAGE II Satellite Observations
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
A new retrieval algorithm has been developed based on the Optimal Estimation (OE) approach, which retrieves lognormal aerosol size distribution parameters from multiwavelength aerosol extinction data, as measured by the Stratospheric Aerosol and Gas Experiment (SAGE) II in the lower stratosphere. Retrieving these aerosol properties becomes increasingly more difficult under aerosol background conditions, when tiny particles (« 0.1 µm) prevail, to which the experiment is nearly or entirely insensitive. A successful retrieval algorithm must then be able (a) to fill the 'blind spot' with suitable information about the practically invisible particles, and (b) to identify 'the best' of many possible solutions. The OE approach differs from other previously used aerosol retrieval techniques by taking a statistical approach to the multiple solution problem, in which the entire range of possible solutions are considered (including the smallest particles) and characterized by probability density functions. The three main parts of this thesis are (1) the development of the new OE retrieval algorithm, (2) the validation of this algorithm on the basis of synthetic extinction data, and (3) application of the new algorithm to SAGE II measurements of stratospheric background aerosol. The validation results indicate that the new method is able to retrieve the particle size of typical background aerosols reasonably well, and that the retrieved uncertainties are a good estimate of the true errors. The derived surface area densities (A), and volume densities (V ) tend to be closer to the correct solutions than the directly retrieved number density (N), median radius (R), and lognormal distribution width (S). Aerosol properties as retrieved from SAGE II measurements (recorded in 1999) are observed to be close to correlative in situ data. In many cases the OE and in situ data agree within the (OE and/or the in situ ) uncertainties. The retrieved error estimates are of the order of 69% (σN), 33% (σR), 14% (σS), 23% (σA), 12% (σV), and 13% (σReff ). The OE number densities are generally larger, and the OE median particle sizes are generally smaller than those N and R retrieved by Bingen et al. (2004a), who suggest that their results underestimate (N) or overestimate (R) correlative in situ data due to the 'small particle problem'. The OE surface area estimates are generally closer to correlative in situ profiles (courtesy of T. Deshler, University of Wyoming), and larger than Principal Component Analysis (PCA) retrieval solutions of A (courtesy of L. W. Thomason, NASA LaRC) that have been observed to underestimate correlative in situ data by 40-50%. These observations suggest that the new OE retrieval algorithm is a successful approach to the aerosol retrieval problem, which is able to add to the current knowledge by improving current estimates of aerosol properties in the lower stratosphere under low aerosol loading conditions.