A Laboratory Experiment for the Statistical Evaluation of Aerosol Retrieval (STEAR) Algorithms

摘要

We have developed a method for validating aerosol retrieval algorithms by mimicking atmospheric extinction and radiance measurements in a laboratory experiment with real aerosols. This enables radiometric retrievals that utilize the same sampling volumes, relative humidities, and particle size ranges as observed by other in situ instrumentation in the experiment. We utilize three Cavity Attenuated Phase Shift (CAPS) monitors for extinction and UMBC's three-wavelength Polarized Imaging Nephelometer (PI-Neph) for angular scattering measurements. We subsample the PI-Neph radiance measurements to angles that correspond to Aerosol Robotic Network (AERONET) almucantar scans with solar zenith angles ranging from 50° to 77°. These measurements are then used as input to the Generalized Retrieval of Aerosol and Surface Properties (GRASP) algorithm, which retrieves size distributions, complex refractive indices, single-scatter albedos (SSA), and lidar ratios for the in situ samples. We tested 285 aerosol samples in our experiment; the PI-Neph provided quality radiances for 232 of those samples, and the GRASP retrieval code provided retrievals with residuals R ＜ 10% for 93-100 of those samples, depending upon the simulated solar zenith angle. The samples that we tested include Arizona Test Dust, Arginotec NX, Senagal clay, Israel clay, Montmorillonite, Hematite, Goethite, volcanic ash, ammonium nitrate, ammonium sulfate, and fullerene soot. Samples were alternately dried or humidified, and size distributions were limited to diameters of 1.0 or 2.5 μm by using a cyclone. The SSA at 532 nm for these samples ranged from 0.59 to 1.00 when computed with CAPS extinction and PSAP absorption measurements. The GRASP retrieval provided SSAs that are highly correlated with the in situ SSAs, and the correlation coefficients ranged from 0.955 to 0.976, depending upon the simulated solar zenith angle. The GRASP SSAs exhibited an average absolute bias of +0.023±0.01 with respect to extinction and absorption measurements for the entire dataset. Similarly, bistatic lidar ratios at 532 nm (θ_(sca) = 173°) computed from the extinction & PI-Neph scattering measurements ranged from 21 to 144 sr; the correlations of GRASP bistatic lidar ratios with these direct measurements ranged from 0.488 to 0.735, and monotonically increased with the simulated solar zenith angles. The GRASP bistatic lidar ratio retrievals exhibited average positive relative biases of 6-10% and average absolute biases of 4.0-6.6 sr with respect to the direct measurements (again, depending upon the simulated solar zenith angle).