This paper reports on a study that was performed as part of a larger F
rench program to investigate radiation fog development. It concentrate
s on the derivation of fog optical thicknesses from satellite observat
ions. The technique, which employs Advanced Very High Resolution Radio
meter (AVHRR) measurements from channels 1 and 2, is applied to three
satellite images corresponding to a fog event that occurred over North
ern France on February 8, 1989. The paper first investigates the sensi
tivity of the retrieved optical thickness to the various parameters af
fecting the inversion method, namely the effects of ozone, water vapor
, aerosols, droplet size, surface albedo as well as the calibration of
the AVHRR sensor. These effects are simulated using a radiative trans
fer code, based upon the Discrete Ordinate Method, that calculates the
radiances reflected by a vertically inhomogeneous scattering atmosphe
re. The optical thicknesses obtained by the code, through inversion of
the top of atmosphere reflectances, are then compared to independent
in situ pyranometer measurements. It is found that accurate calibratio
n of the AVHRR is critical to the inversion technique. Consequently ne
w calibration coefficients were computed using in situ pyranometer mea
surements. These coefficients were found to be very close (< 2%) to th
ose derived from a collection of NOAA 11 satellite data. Moreover, the
very good agreement between the calibration coefficients derived usin
g two successive images, corresponding to different viewing angles, se
ems to indicate that using a plane-parallel model is justified for inv
ersion of NOAA 11 data over a fog layer. Independent of the calibratio
n errors, the accuracy in computing the optical thicknesses is estimat
ed to be about 10-20% using channel 1 data and 15-30% using channel 2
data, in the case of dense fog.