Experiments In Automatic Cloud Tracking Using SMS-GOES Data

Abstract

A description is given of the component parts of a computer system for automatically tracking clouds shown by sequences of pictures obtained by geostationary weather satellites. The component programs perform the following functions: separation of clouds from background; subdivision of the cloud data into groups (which are potential tracers); computation of the location, size, brightness and infrared values of each group: matching groups at two different times to give cloud motion vectors;and conversion of motions from the row and column coordinates of recorded data to earth coordinates. In this paper, the emphasis is on the automatic grouping and tracking functions. Some recent tests of the method are described and illustrated. These tests used SMS-GOES data in both visible and infrared channels for several different types of clouds, including low clouds, convective clouds and cirrus. Using 4 mi resolution data at one-half hour intervals, and with no initial information about the expected motions, the system was able to determine accurate motion vectors for clouds moving at low or moderate speeds. For high speed flow, it was found desirable to have a first guess of the motion or more frequent observations; otherwise it is necessary to use lower resolution data, thereby sacrificing some detail and accuracy in the computed motion fields. For middle and high clouds, superior results were obtained when clustering was based on infrared rather than visible data. For low clouds, the reverse was true. In terms of the accuracy and number of cloud motion vectors obtained, the overall results are considerably superior to those that we achieved previously using automatic methods. Our experience indicates that the system can process satellite data to obtain cloud motion vectors for certain research studies, or for input to numerical prediction models.