Physical basis, premises, and self-consistency checks of aerosol retrievals from TRMM VIRS

This paper outlines the processing stream for aerosol retrievals over ocean s from the visible and infrared scanner [VIRS; a five-channel radiometer si milar to the National Oceanic and Atmospheric Administration's Advanced Ver y High Resolution Radiometer (AVHRR)] aboard the Tropical Rainfall Measurin g Mission (TRMM) satellite, launched in November 1997. Emphasis is on 1) th e applying the previously developed AVHRR second-generation aerosol retriev al algorithm to VIRS data to derive an aerosol parameter, indicative of par ticle size; 2) removing the unwanted "thermal leak'' signal in the 1.61-mum channel; 3) giving examples of the first aerosol retrievals from space at this wavelength; and 4) assessing the accuracy of the retrievals with theor etical error analyses and empirical self- and interconsistency checks. Aero sol optical depths tau (A)(1) and tau (A)(2) are retrieved from reflected s olar radiances in VIRS channels 1 and 2 centered at wavelengths lambda (1) = 0.63 and lambda (2) = 1.61 mum, using two independent lookup tables. When tau (A)(1) and tau (4)(2) exceed a certain threshold tau (A)(min) an effec tive Angstrom exponent alpha related to particle size is derived as alpha = -ln(tau (A)(1)/tau (A)(2))/ln(lambda (1)/lambda (2)). Channel 2 is contami nated by a thermal leak,originating from a secondary spectral response peak centred at similar to5.2 mum. If uncorrected, it leads to errors in tau (A )(2) of 100% or more. To minimize this error, nighttime VIRS "dark" radianc es in channel 2 have been related empirically to radiances in channels 4 an d 5 (10.8 and 12 mum, respectively), and view angle through regression anal yses. The reflected component in channel-2 daytime measurements is estimate d by subtracting the empirically derived thermal component from the total s ignal and is used in the retrieval of tau 24. Theoretical error analysis is used to identify the limitations of the VIRS retrieval algorithm, whereas actual retrievals are preliminarily evaluated using a set of specially deve loped empirical checks. The checks show, on average, a high degree of self- and interconsistency but also identify problems with the retrievals, the m ost noteworthy being trends in retrieved optical depths with viewing and il lumination angles. These problems will be tackled in the next-generation ae rosol retrieval algorithm.