Evaluating the principles of cloud remote sensing with AVHRR and MAS data over SHEBA

A rigorous discrete ordinates radiative transfer formulation has been applied to two Advanced Very High Resolution Radiometer (AVHRR) images extracted from telemetry collected by the CCGS Des Groseilliers satellite tracking system during SHEBA, to estimate cloud optical depth and effective radius of the cloud droplet size distribution. The two cases, from June 2 and 3 of 1998, were chosen for analysis because (1) the images contained mostly liquid water clouds and (2) contemporaneous MODIS Airborne Simulator (MAS) overflight imagery was available for these AVHRR overpasses. The objective is to apply the same detailed radiative transfer formulation to both the MAS and AVHRR data, so that the quality of the retrievals from the latter can be evaluated. Retrievals of cloud optical properties from MAS are assumed to be more reliable, because (1) all MAS channels have direct radiometric calibration, (2) the higher spatial resolution of MAS (50 m nadir versus 1.1 km nadir with AVHRR) should yield smaller uncertainties related to partially cloudy pixels in a given study area, and (3) effective droplet radius can be retrieved directly from the MAS 1.62 micron channel without additional uncertainties involved with subtracting a thermal radiance component.

Examination of the retrievals from both sensors in these two cases reveals considerable spatial variability (more than a factor of two) in cloud optical depth, on a variety of scales ranging from tens of meters to tens of kilometers, even for relatively uniform liquid water clouds. Retrievals of cloud effective droplet radius from AVHRR are generally consistent with those from MAS, suggesting that AVHRR can be reliably used to estimate this quantity. However, AVHRR-based retrievals of cloud optical depth are subject to large errors that result from small uncertainties in the absolute radiometric calibration of AVHRR channel 2. Using recalibration coefficients from one of the more robust AVHRR post-launch calibration efforts, the cloud optical depths we retrieve from NOAA-14 AVHRR channel 2 are consistently 30 - 50% larger than those obtained from MAS. The intercomparison of MAS and AVHRR retrievals of cloud optical depth also revealed errors with AVHRR due to partial cloud cover, and these errors are not immediately apparent when examining the AVHRR retrievals alone. If the AVHRR retrievals are averaged to spatial resolutions of order 10 - 30 km, they appear to become more stable for use in applications such as atmospheric energy budget calculations.