Evaluation of Automated Imagery Analysis Algorithms for Use in the Three-dimensional Nephanalysis Model at AFGWC
Description:... Two different imagery analysis algorithms, one of which is currently in use by the 3DNEPH automated cloud analysis program at the Air Force Global Weather Central (AFGWC), and the other developed at AFGL, were evaluated and compared on a sample set of DMSP IR data (10.2 to 12.8 micrometers) with the aid of the AFGL Man-computer Interactive Data Access System. The 3DNEPH algorithm, MODAL, and the AFGL algorithm, CLUSTER, were designed to separate cloud layers in IR satellite imagery. Such information is used by the 3DNEPH to compute cloud analysis parameters as percent cloud (earth) coverage and cloud heights. Both algorithms were found to perform well under some meteorological circumstances, but not so well under others. In general, however, it was found that CLUSTER performs better. The CLUSTER algorithm found clear/cloud boundaries, numbers of cloud layers, and fractional cloud covers that are closer to estimates made by analysts looking at corresponding IR and visible images. The MODAL algorithm found more cloud layers than were observed, and tended to overanalyze the transition regions between cloud layers. The AFGWC has recognized the limitations of MODAL and the complications in its computer code. Subsequently, a substantially revised version of the 3DNEPH has been written, and is known as the Real-Time Nephanalysis (RTNEPH).
Show description