Multi-grid-cell validation of satellite aerosol property retrievals in INTEX/ITCT/ICARTT 2004

Aerosol transport off the US Northeast coast during the Summer 2004 International Consortium for Atmospheric Research on Transport and Transformation ( ICARTT) Intercontinental Chemical Transport Experiment ( INTEX) and Intercontinental Transport and Chemical Transformation ( ITCT) experiments produced a wide range of aerosol types and aerosol optical depth ( AOD) values, often with strong horizontal AOD gradients. In these conditions we flew the 14-channel NASA Ames Airborne Tracking Sun photometer ( AATS) on a Jetstream 31 ( J31) aircraft. Legs flown at low altitude ( usually <= 100 m ASL) provided comparisons of AATS AOD spectra to retrievals for 90 grid cells of the satellite radiometers MODIS-Terra, MODIS-Aqua, and MISR, all over the ocean. Characterization of the retrieval environment was aided by using vertical profiles by the J31 ( showing aerosol vertical structure) and, on occasion, shipboard measurements of light scattering and absorption. AATS provides AOD at 13 wavelengths 1 from 354 to 2138 nm, spanning the range of aerosol retrieval wavelengths for MODIS over ocean ( 466-2119 nm) and MISR ( 446-866 nm). Midvisible AOD on low-altitude J31 legs in satellite grid cells ranged from 0.05 to 0.9, with horizontal gradients often in the range 0.05 to 0.13 per 10 km. When possible, we used ship measurements of humidified aerosol scattering and absorption to estimate AOD below the J31. In these cases, which had J31 altitudes 60-110 m ASL ( typical of J31 low-altitude transects), estimated midvisible AOD below the J31 ranged from 0.003 to 0.013, with mean 0.009 and standard deviation 0.003. These values averaged 6% of AOD above the J31. MISR-AATS comparisons on 29 July 2004 in 8 grid cells ( each similar to 17.6 km x 17.6 km) show that MISR versions 15 and 16 captured the AATS-measured AOD gradient ( correlation coefficient R-2 = 0.87 to 0.92), but the MISR gradient was somewhat weaker than the AATS gradient. The large AOD ( midvisible values up to similar to 0.9) and differing gradients in this case produced root-mean-square ( RMS) MISR-AATS AOD differences of 0.03 to 0.21 ( 9 to 31%). MISR V15 angstrom ngstrom exponent alpha ( = -dlnAOD/dln lambda) was closer to AATS than was MISR V16. MODIS-AATS AOD comparisons on 8 overpasses using 61 grid cells ( each nominally 10 km x 10 km) had R-2 x 0.97, with RMS AOD difference similar to 0.03 (similar to 20%). About 87% of the MODIS AOD retrievals differed from AATS values by less than the predicted MODIS over-ocean uncertainty, Delta(tau) = +/- 0.03 +/- 0.05 tau. In contrast to the small MODIS-AATS differences in AOD, MODIS-AATS differences in A ngstrom exponent a were large: RMS differences for alpha( 553, 855 nm) were 0.28 for MODIS-Terra and 0.64 for MODIS-Aqua; RMS differences for alpha( 855, 2119 nm) were larger still, 0.61 for MODIS-Terra and 1.14 for MODIS-Aqua. The largest MODIS-AATS A ngstrom exponent differences were associated with small AOD values, for which MODIS AOD relative uncertainty is large. Excluding cases with AOD( 855 nm) < 0.1 reduced MODIS-AATS a differences substantially. In one grid cell on 21 July 2004, smoke over cloud appeared to impair the MODIS-Aqua cloud mask, resulting in retrieved AODs that significantly exceeded AATS values. Experiments with extending MODIS retrievals into the glint mask yielded MODIS AODs consistently less than AATS AODs, especially at long wavelength, indicating that the current MODIS glint mask limits should not be reduced to the extent tried here. The sign of the AOD differences within the glint mask ( MODI AOD < AATS AOD) is consistent with ship-measured wind speeds there.