The Problem of Retrieving Atmospheric Parameters from Radiative Observations - Short-Wave Solar Radiation in the Earth's Atmosphere - page 135

thecomparisonbetweentheobservationalresultsandsolutionofthedirect
problem, the inverse problems are solved with the accuracy defined with the
uncertainty of the selected model parameters choice, i. e. with the concrete
choice of operator G . Hence, the stage of the choice of method for the direct
problem solving is the most important part of solving the inverse problem. Be-
sides, as has been mentioned above, the operator of the direct problem solving
is inevitably approximated in any case; hence, the account of the approximation
influence on the results is necessary as well.
In conclusion, the following general scheme for numerically solving the
inverse problems in atmospheric optics could be proposed:
1. Studying the contemporary theory of the physical processes forming the
measured characteristics.
2. Choosing a concrete mathematical model of the observations together
with its parameters, realization of this model on computer.
3. The error analysis of the direct problem.
4. Dividing the parameters of the mathematical model to the known ones
andtothesubjectsoftheretrieval.
5. Choosing the method for solving the inverse problem. Estimating its
accuracy.
6. Realization of the solving algorithm on computer.
7. Observational data processing, the analysis and interpretation.
Excluding the first one, which has been considered in Chap. 1 we will discuss
all listed stages further, applying them to concrete inverse problems. However,
the survey is more appropriate in a different order from that listed above. We
shouldmention that firstly the described scheme has been proposed according
to the results of the accomplished observations, so the actual problem of the
optimal experiment planningwill not be touchedupon. Secondly, the presented
algorithm has a more complicated logic in practice; in particular, returning to
previous stages with the purpose of verifying the model and modernization of
the numerical methods are possible. Thus, the numerous consequent versions
of the processed results presented in the studies by Chu et al. (1989, 1993)
and Steele and Turko (1997) are the standard situation while processing the
observational data of atmospheric optics. In fact, it is well known to specialists:
the results of the field observations in majority is impossible to process once
and for all, there is always something to improve.
We will not review the huge volume and variety of recent inverse problems
of atmospheric optics and methods of their solution. As has been mentioned
hereinbefore a certain classification of these problems was presented in the
study by Timofeyev (1998), and concerning the solution methods there has
been no classification for them yet. Here we will confine ourselves only to
the concrete inverse problems of retrieval of the atmospheric and surface
parametersfromtheresultsoftheairborneandsatelliteobservationsofthe
solar spectral radiance and irradiance in the atmosphere considered inChap. 3.