A numerical testbed for the characterization and optimization of aerosol remote sensing

Remote sensing of aerosols from satellite and ground-based platforms provides key datasets to help understand the effect of air-borne particulates on air quality, visibility, surface temperature, clouds, and precipitation. However, global measurements of aerosol parameters have only been generated in the last decade or so, with the advent of dedicated low-earth-orbit sun-synchronous satellite sensors such as those of NASA's Earth Observation System (EOS). Many EOS sensors are now past their design lifetimes. Meanwhile, a number of aerosol-related satellite missions are planned for the future, and several of these will have measurements of polarization. A common question often arises: How can a sensor be optimally configured (in terms of spectral wavelength ranges, viewing angles, and measurement quantities such as radiance and polarization) to best fulfill the scientific requirements within the mission's budget constraints? To address these kind of questions in a cost-effective manner, a numerical testbed for remote sensing aerosols is an important requirement. This testbed is a tool that can generate an objective assessment of aerosol information content anticipated from any (planned or real) instrument configuration. Here, we present a numerical testbed that combines the inverse optimal estimation theory with a forward model containing linearized particle scattering and radiative transfer code. Specifically, the testbed comprises the following components: (1) a linearized vector radiative transfer model that computes the Stokes 4-vector elements and their sensitivities (Jacobians) with respect to the aerosol single scattering parameters at each layer and over the column; (2) linearized Mie and T-matrix electromagnetic scattering codes to compute the macroscopic aerosol single scattering optical properties and their sensitivities with respect to refractive index, size, and shape; (3) a linearized land surface model that uses the Lambertian, Ross-Thick, and Li-Sparse kernels to compute the reflectance and the sensitivities to the kernel weighting factors; a linearized BPDF computes the angular polarized reflectance; (4) a linearized ocean surface model integrating the Cox-Munk glitter model with the chlorophyll-dependent water-leaving contribution; (5) a HITRAN-based gas absorption calculation of trace species cross sections (also linearized with respect to temperature and pressure); (6) the Levenberg-Marquardt inverse algorithm for cost-function minimization and optimal derivation of a posteriori solutions. In this presentation, we introduce our testbed and demonstrate applications to several sensor design and algorithm formulation concepts. This includes an evaluation of the use of polarization in the O2 A band for the retrieval of aerosol height information from space, and an assessment of the potential improvement in the characterization of aerosol scattering properties through the addition of more polarization channels to the AERONET sensors.