Download or read book Development and Assessment of a Neural Network Approach for Retrieving Aerosol Properties from Multispectral, Multiangle Polarization Measurements written by Alexandra Tsekeri and published by . This book was released on 2012 with total page 432 pages. Available in PDF, EPUB and Kindle. Book excerpt: Quantifying the microphysical properties of aerosols is crucial for quantifying global climate forcings. Satellite based aerosol retrievals usually rely on intensity measurements of the scattered light, but this approach has been proven inadequate for retrieving the complex refractive index and shape of aerosols, as well as the contamination from the ground surfaces. It is with these limitations in mind that we plan to improve the quality and scope of aerosol retrievals, by making use of the full capabilities of current and future polarimetric sensor systems. In order to utilize the increased information content on aerosol optical thickness (AOT), size distribution, shape and single scattering albedo (SSA), intrinsically available in multispectral-multiangle polarimetric observations, we make use of suitably constructed neural networks (NNs). We focus our analysis initially on simple retrievals over the ocean, in order to best assess the potential of the NNs as a practical approach and to identify any possible limitations. In particular, we find that, by choosing a suitable combination of inputs and outputs, based on principal component analysis (PCA), we can develop a robust NN retrieval trained on synthetic datasets. We further show the value of using cascaded NNs, to improve retrieval accuracy. Consequently, we demonstrate the potential and limitations of this approach on real aircraft instrument data from the Research Scanning Polarimeter (RSP). Discrepancies in the retrievals are found to be due to limitations from the use of spherical particle assumptions and preliminary efforts to overcome this restriction are identified. It is our belief that the value of these methods, in comparison to existing local inversion schemes, will further increase with the expected magnification of data sizes on future missions, such as the Aerosol-Cloud-Ecosystem (ACE) Mission.