Estimation of multiple-aerosol concentration and backscatter using multi-wavelength range-resolved lidar

Previous work by the authors has produced statistically based methods for detecting, estimating and classifying aerosol materials in the atmosphere using multiple-wavelength range-resolved CO2 lidar. This work has thus far been limited to the presence of a single aerosol material at a given time within the lidar line-of-sight. Practical implementation requires the ability to detect and discriminate multiple aerosol materials present simultaneously such as smoke and dust in addition to hazardous materials. Treating mixtures of materials necessitates fundamentally different approaches from the single-material case since neither the aerosol backscatter wavelength-dependence nor the concentrations as a function of range are known. Because of this, linear processing cannot resolve the mixture data into its components unambiguously, and non-linear methods must be considered. In this paper we describe an empirical Bayes (EB) approach for resolving mixtures of aerosol into their components. The basic idea of EB is to use the same data to estimate the prior distribution of a set of parameters as that used to estimate the parameters themselves. In our case the concentration and backscatter are the parameters that are estimated with the help of a prior distribution of the backscatter. We implement the EB estimator through the EM (Expectation Maximization) algorithm. The resulting processor is applied to injections of interferent dust into data sets collected by ECBC during JBSDS testing at Dugway Proving Ground, UT in 2006.