Global retrievals of the spatial, microphysical and optical properties of clouds are urgently needed for assessing cloud radiative forcing at the Earth's surface and top of atmosphere (TOA). Reliable climatologies are also needed to help assess the goodness of global climate models (GCMs) and other applications.

Satellite-based retrievals of cloud water, particle size, cloud top/base and optical thickness are reasonably common in the remote sensing community. However, real-time global retrievals are not. We will present the suite of cloud detection and analysis algorithms that currently are being implemented into operations at the Air Force Weather Agency at Offutt AFB, Nebraska. The algorithm suite consists of two main parts: first, a cloud-mask and phase determination made on a pixel-by-pixel basis; and second, retrieval of cloud microphysics (particle size, cloud-water path), radiative (optical thickness, emissivity), and spatial properties (cloud top and base). We couple the "bulk," total-column retrievals with inferences on normalized cloud-water-content profile shapes to infer vertical profiles of cloud water content and particle size. Retrievals of cloud-top temperature account for the varying optical thickness of thin cirrus, and greatly improve the accuracy of cloud height estimates when compared to the current operational technique.

The retrievals are deterministic and exploit the distinct dependencies that thermal infrared upwelling radiances in several spectral bands have on these properties. The models will be working on the full constellation of operational environmental satellites, including GOES, TIROS, METEOSAT, and MTSAT. The technique makes use primarily of infrared sensor data, so that it is applicable both day and night with no sharp, physically unrealistic analysis boundaries in the vicinity of the terminator. Preparations are also being made at AFWA to process EOS MODIS and NPOESS sensor datasets. Our algorithms automatically adapt to existing/available sensor spectral resolutions and view/illumination conditions. The cloud retrievals compare well with ground-based radar and lidar cloud observations, and with Calipso and Cloudsat datasets as well.

The set of infrared retrieval algorithms are the first step in a logical progression to a one-dimensional variational assimilation (1D-VAR) environment which will help alleviate uncertainties due to noise, the radiative-transfer model itself or in the "goodness" of the retrieval first guess. They also provide a reference for algorithm comparisons, and are being considered as a first guess to accelerate computationally the 1D-VAR processing.