Spurious Signals in TRMM/VIRS Reflectance Channels and Their Effect on Aerosol Retrievals

Abstract

Aerosol optical depths, τ1 and τ2, and the Ångström exponent α = ?ln(τ1/τ2)/ln(?1/?2), are retrieved from daytime measurements (sun zenith angle ?o < 60°) over ocean in reflectance bands 1 (?1 = 0.63 µm) and 2 (?2 = 1.61 µm) of the five-channel visible and infrared scanner (VIRS) on board the Tropical Rainfall Measuring Mission (TRMM) satellite. In band 2, a thermal leak originating from the secondary spectral response peak at ?5.2 µm contributes radiance comparable to the signal scattered by aerosols. In the past two corrections, the thermal signal was parameterized empirically as a linear function of radiances in bands 4 and 5 (centered at 10.8 and 11.9 µm, respectively), R4 and R5, and a quadratic function of view angle ? through multiple regression analyses. The regression coefficients were estimated from a limited amount of all-sky nighttime (100° < ?o < 170°) data over land and ocean, and were used to predict and remove the false signal from daytime data. As a result, retrievals of τ2 and α have been improved, but they still remain seriously flawed. This study reexamines the nighttime signal in VIRS channel 2 using two representative 9-day segments of the TRMM single scanner footprint (SSF) data collected from 4?12 February and 2?10 April 1998. The past parameterizations did not always perform accurately. Their residuals are biased and skewed, and reveal artificial trends with time, latitude, ?, R4, and R5. A new parameterization of the nighttime signal is proposed that makes use of 1) clear-sky ocean data only (rather than previously used all sky, full set); 2) more accurate principal component analyses (PCA) to approximate the ?, R4, and R5 dependencies of the false signal (in place of the formerly used liner/quadratic regressions); and 3) explicit accounting for temporal instability of the spurious signal (rather than assuming it to be stable as was done in the past). The new parameterization substantially relieves the problems found in the previous two parameterizations. A much smaller false signal of unknown origin, found in channel 1, is also analyzed and parameterized in this study, consistently with channel 2. The effects of false signals and residuals of different corrections on retrieved τ and α are preliminarily estimated using an approximate formulation based on a simplified treatment of the radiative transfer equation.