Polar cloud-detection algorithms for a real-time analysis and forecasting model

We describe in our paper new and modified cloud-detection and background discrimination algorithms developed to improve automated cloud-detection performance in regions poleward of 70 degrees. Several of the new modules are applicable to multiple polar satellite systems including DMSP, TIROS, and the planned European platforms. Each module includes logic to adapt to sensor-specific limitations when and where they exist, obviating the need for developing separate procedures for each sensor system.

The algorithms fall into two fundamental classes: “static” spectral tests and “temporal-difference” tests. Additionally there are two hybrid spectral-temporal tests that identify inversion fogs and moving/forming cloud features both day and night. In all there are 7 multispectral tests: 3 daytime-only, 2 day/night, and 2 nighttime-only algorithms. Daytime tests include obviously bright clouds, snow-cloud discrimination, and highly reflective clouds at ~3.7-micron wavelengths. Nighttime modules test for fog and thin cirrus. In both day and night checks are made for obviously cold clouds, and thinner-cirrus and low-cloud edges.

There are three temporal differencing tests, each of which operate both day and night. A “traditional” single-channel infrared temporal differencing test flags forming and moving cloud systems. This is akin to the techniques used operationally by the US Air Force global cloud model with geostationary data. A second temporal differencing test identifies inversion fogs in the polar night, when such clouds are much warmer than the underlying surface. A third experimental test checks for midwave-longwave temporal signatures.

These polar cloud detection algorithms will be transitioned to the US Air Force's Cloud Depiction and Forecast System (CDFS) operational global cloud models during calendar year 2006.