Cloud detection using a combination of hyperspectral infrared channels and microwave-derived precipitable water

We present a technique to detect clouds using brightness temperature differences (BTD) of Atmospheric Infrared Sounder (AIRS) window channels, while constraining the BTD due to the water vapor continuum using precipitable water (PW) estimates derived from Advanced Microwave Sounding Unit (AMSU) channels. Different combinations of shortwave (SW) and longwave (LW) channels are investigated for daytime and nighttime applicability. Radiative transfer (RT) simulations and observations of SW-LW BTDs for stratus and cirrus clouds indicate a larger sensitivity in BTD than previous LW split-window cloud detection methods. Both simulations and observations show stratus clouds often have BTD < 0, while a large number of cirrus have BTD > 0.

Our approach is to define the range of BTD that can occur for a clear-sky atmosphere given a fixed value for total column PW under a myriad of T(z) and RH(z) profiles. The range of BTD is not only sensitive to T(z) and RH(z) variability, but to surface emissivity, differences between surface temperature (Ts) and near surface temperature (Ta), channel noise properties, uncertainties of sub-AMSU footprint PW variability on the scale of an AIRS footprint, and measurement scan angle. Once the range in BTD for clear-sky is defined, a dynamic threshold for cloud is set.

A set of radiosondes taken from tropical, subtropical, midlatitude, and subpolar latitudes over the full seasonal cycle are used to simulate a realistic range of BTDs given observed ranges in T(z) and RH(z). The initial analysis is limited to oceanic regions to minimize emissivity and Ts-Ta effects on the BTD. The aforementioned physical quantities and their uncertainties, and channel noise properties are incrementally added to the RT simulations to see the effect on the BTD. Global cloud detection maps will be presented, and will be validated against cloud climatologies from other methods and data sources, including clouds as detected by cloud radar and lidar at ARM sites.