Satellite remote sensing deals with a complex system coupling atmosphere an
d surface. Any physical model with reasonable precision needs several to te
ns of parameters. Without a priori knowledge of these parameters, Propositi
on 3 of Verstraete et al. requires the number of independent observations t
o be greater than the number of unknown parameters. This requirement can ha
rdly be satisfied even in the coming EOS era. As Tarantola pointed out, the
inversion problems in geoscience are always underdetermined in some sense.
In order to make good use of every kind of a priori knowledge for effectiv
ely extracting information from remote sensing observations, the right ques
tion to set is as follows: Given an imperfect model and a certain amount of
a priori information on model parameters, in which sense should one modify
the a priori information, given the actual observation with noise? A prior
i knowledge of physical parameters can be presented in different ways such
as physical limits, global statistical means and variance for a certain lan
dcover type, or previous statistics and temporal variation of a specific ta
rget. When such a priori knowledge can be expressed as joint probability de
nsity, Bayessian theorem can be used in the inversion to obtain posterior p
robability densities of parameters using newly acquired observations. There
is no prerequirement on how many independent observations must be made, an
d the knowledge gained merely depends on the information content of the new
observations. Some specific problems about knowledge accumulation and rene
wal are also discussed.