|
Clouds play an important role in the hydrologic cycle, influence global energy balance, and represent a significant yet poorly understood component of global climate change. As a result, quantitative global observations of liquid and ice cloud microphysical and radiative properties continue to be a focus of a growing number of satellite-based sensors each having an associated suite of retrieval algorithms. While a number of these algorithms have successfully been applied to map clouds, many can only be applied under specific conditions (eg. during the daytime) or over a limited dynamic range (eg. optically thin cirrus) often leading to unphysical discontinuities when one seeks to compile a complete picture of the global distribution of clouds. Furthermore, discrepancies exist between products of different algorithms when they are applied to the same scene by virtue of differences in the information provided by distinct combinations of measurements.
This paper revisits the problem of cloud microphysical property
retrievals from satellite radiance observations at solar and thermal
wavelengths in an effort to quantify their information content with
respect to single layer liquid and ice clouds over an oceanic
background. Using the channels on the Moderate Resolution Imaging
Spectroradiometer (MODIS) as an example, it will be demonstrated that
an entropy-based definition of information content provides a useful
metric for evaluating the utility of a set of observations in a
retrieval problem. This approach is used to objectively determine the
subset of wavelengths that provide the greatest amount of information
for oceanic microphysical property retrievals from the MODIS
instrument. The results show that the combination of a conservative and a non-conservative scattering shortwave channel in concert with a near-infrared channel, an infrared window channel, and one in the wings of the 15 m CO2 band provide the optimal channel combination for the wide variety of liquid and ice clouds examined. With an eye toward developing a coherent representation of the global distribution of cloud microphysical and radiative properties, this combination of channels may be integrated into a suitable multi-channel inversion methodology such as the optimal estimation or Bayesian techniques to provide a means of establishing a common framework for cloud retrievals under varying conditions. Under some circumstances, other channels may provide a small amount of additional information but in most cases the remaining channels only supply redundant information and do not justify the additional computation cost required to integrate them into an algorithm.
|
