Application of Optimal Spectral Sampling for a Real-Time Global Cloud Analysis Model

Abstract

he Cloud Depiction and Forecast System version 2 (CDFS II) is the operational global cloud analysis and forecasting model of the 557th Weather Wing, formerly the U.S. Air Force Weather Agency. The CDFS II cloud-detection algorithms are threshold-based tests that compare satellite-observed multispectral reflectance and brightness temperature signatures with those expected for the clear atmosphere. User-prescribed quantitative differences between sensor observations and the expected clear-scene radiances denote cloudy pixels. These radiances historically have been modeled at 24-km resolution from a running 10-day statistical analysis of cloud-free pixels that requires the entire global cloud analysis to be executed twice in real time: once in operational cloud detection mode and a second time in a cloud-clearing mode that is designed explicitly for generating clear-scene statistics. Having to run the cloud analysis twice means the availability of fewer compute cycles for other operational models and requires costly interactive maintenance of distinct cloud-detection and cloud-clearing threshold sets. Additionally, this technique breaks down whenever a region is persistently cloudy. These problems are eliminated by means of the optimal spectral sampling (OSS) radiative transfer model of Moncet et al., optimized for execution in the CDFS run-time environment. OSS is particularly well suited for real-time remote sensing applications because of its user-tunable computational speed and numerical accuracy, with respect to a reference line-by-line model. The use of OSS has cut cloud model processing times in half, eliminated the influence of cloudy pixel artifacts in the statistical time series prescription of cloud-cleared radiances, and improved cloud-mask quality.