Evaluation of interregional variability in MODIS cloud regimes

Abstract:

Clustering techniques have been used in the last few decades to classify cloud types automatically from satellite observations, most commonly using cloud top pressure and cloud optical depth. The underlying assumption is that the resulting clusters, called "cloud regimes" or "weather states", represent some type of basic states of the atmosphere, and thus that their occurrence can be used as a proxy for related variables such as radiative balance or precipitation.

We have examined the validity of these assumptions by using independent measurements from the CloudSat and CALIPSO satellites. The CloudSat radar yields a reflectivity product that is sensitive to many aspects of the physics of the clouds, while CloudSat together with the CALIPSO lidar can retrieve the vertical structure of the cloud column, including multi-layer clouds. These observations have been separated into groups according to the recently published cloud regimes based on data from the MODIS instrument, deployed on the Aqua satellite orbiting in the same constellation with CloudSat and CALIPSO. The distributions of these observations have been constructed both globally and in a number of regions in different parts of the Earth. By analyzing the differences in the distributions between these regions, we can evaluate the usefulness of the cloud regimes as a proxy for the measured variables.

Some cloud regimes have been found to be rather stable between regions, while others display considerable variability. Moreover, some cloud regimes appear much more similar to each other in CloudSat observations than they do using the MODIS regimes. We analyze the implications of these differences for the usability of the cloud regimes as climate indicators. We also explore various filtering techniques and different clustering methods that can potentially be used to reduce these differences, and thus to improve the universality of the cloud regimes.