C. Gautier
Mission Research Corporation
In the ARM context, satellite observations can be used in two important ways: 1) to validate radiative transfer models and 2) to contribute to the four-dimensional description of the atmosphere within a data assimilation system. These uses are important because surface observations are insufficient to validate radiative transfer models-satellite data more or less double the amount of information-and also because data assimilation is needed to provide initial conditions for numerical models and verification data for validation of parameterization.
To validate radiative transfer models, forward radiative transfer models are used to compute radiances at the top of the atmosphere, using a description of the atmospheric properties based on in-situ observations of the input parameters needed by the radiative transfer model. Because both the in-situ and the satellite observations contain errors, this validation is only performed within a certain degree of accuracy that depends on both sets of errors.
For data assimilation, satellite observations of the atmosphere have mostly been used in an inverted mode; that is, an inversion scheme has been applied to satellite observations (usually multispectral) to transform them into geophysical parameters. While having the advantage of providing data that look like radiosonde data and therefore are readily usable by numerical weather or climate forecast models, the inversion process faces a number of mathematical problems which can only be approximately solved, and the number of possible solutions of the inversion always remains infinite. This ill-posed mathematical inversion can be circumvented by directly assimilating the satellite data into the model or data assimilation system that requires them.
The direct assimilation approach has several steps. First, satellite radiances are computed using a forward radiative transfer model and numerical forecast model data at the model's grid point as inputs. Then, these simulated radiances are compared with radiances actually observed from space. When discrepancies between the computed and the observed radiances exist, they are minimized by adjusting the forecast model parameters. The main issue here is how to modify these parameters to improve the satellite radiance simulations. While, mathematically, the same infinite number of solutions exists, as in the case of the inversion, the inclusion of physics introduced in the forward computations helps chose a better solution. Whereas this overall approach can become cumbersome on a global basis because of the large number of computations required, this approach might be particularly well suited for ARM sites for, at least, two reasons. First, there already will be a large number of observations made over the ARM sites for other research purposes, and these measurements could be assimilated in the forecast model. Second, because the ARM sites are fixed, a number of parameters describing the state of the ARM site will remain constant, at least over some time periods, thus helping to describe the four-dimensional fields.
In the following section, we describe the post processing tools and models available to perform satellite radiance computations and assess the quality of the state-of-the-art radiative transfer models and discuss their expected accuracies. In the next section, we review the type of assimilation technique available for providing the best state-of-the-atmosphere description, using satellite observations together with a forecast model. Finally, we discuss the additional measurements needed to fully describe the state of the ARM site to accomplish the two above-mentioned objectives.
With the operational satellites currently in orbit, the simulated and observed satellite radiances can be either in visible or in infrared spectral ranges of the AVHRR, VISSR and TOVS channels or eventually, broadband shortwave and longwave radiances such as those that could be measured by an ERB-like satellite.