Estimating aerosol concentration and spectral backscatter with multi-wavelength range-resolved lidar

Abstract

Algorithm development for detecting and discriminating atmospheric aerosols using range-resolved lidar is a straightforward, if non-trivial, application of well-established techniques of statistical signal processing assuming the aerosol backscatter coefficients are known as a function of wavelength. Unfortunately, in contrast to the analogous case of vapors, in most aerosol applications those coefficients are rarely known accurately. This is due to a combination of factors: (1) unknown refractive index dependence on wavelength, particularly for bioaerosols; (2) unknown particle size distribution; and (3) lack of particle sphericity making M e calculations unreliable. Uncertainties in any of these factors can distort the backscatter cross-section spectral dependence to the extent that aerosol identification becomes impossible. This paper presents a sequential algorithm for estimating both the aerosol concentration dependence on range and time and backscatter coefficient spectral signatures for a set of materials using M wavelengths with data available prior to the aerosol release for estimating the ambient lidar return. The rangedependence of the aerosol is modeled as an expansion of the concentration in an orthonormal basis set whose coefficients carry the time dependence. The basic idea is to run two estimators in parallel: a Kalman filter for the expansion coefficients, and a maximum likelihood estimator for the set of aerosol backscatter coefficients. These algorithms exchange information continuously over the data processing stream. The approach is illustrated on atmospheric backscatter lidar data collected by the U.S. Army multi-wavelength lidar from aerosol releases at the recent JBSDS trials at Dugway Proving Ground, UT.