13.8
Interactive validation and tuning of satellite cloud property retrieval algorithms
Gary Gustafson, AER, Cambridge, MA; and D. Peduzzi, M. Plonski, and B. Shaw
Use of environmental satellites to characterize the global distribution of cloud properties is a fundamental application of these systems. A common problem when validating automated cloud-property retrieval algorithms that operate on the satellite sensor data is the general lack of ground truth. Even when coincident surface-based or aircraft observations are available, differences in sampling, observing geometry, and temporal extent often make direct comparisons problematic. Consequently, algorithm validation is frequently performed through comparisons with subjective classification by expert image interpretation analysts viewing imagery produced from the raw sensor data. A series of interactive visualization and algorithm-tuning software tools have been developed to support testing and maintenance of semi-operational cloud analysis algorithms operated at the Air Force Weather Agency (AFWA). At AFWA, multispectral visible and infrared channel data from DMSP, TIROS, GOES, Meteosat and GMS satellites are routinely processed through automated algorithms to produce depictions of the horizontal and vertical distribution of cloud cover at a resolution of 24 km globally and 6 km regionally. To accommodate testing and monitoring of the large quantity of data routinely processed by these algorithms, the visualization tools are designed to display a full orbit of polar satellite data or a full disk image from a geostationary satellite at a time. Also available are tools to display gridded cloud products for both regional and global coverage. Multispectral data are viewed using full-color techniques coupled with numerous interactive and automatic contrast enhancement functions. Direct visual comparison of sensor data with automated algorithm results is done through the use of color-coded masks that show pixel-level algorithm results co-registered and overlaid (non-destructively) on top of the satellite imagery. Other capabilities include an interactive editor to quickly generate new retrieval algorithms and immediately view results, pixel data tabulation using the cursor, and the ability to adjust algorithm threshold values and recursively run the algorithm to evaluate impact on retrieval accuracy.
Session 13, Applications of IIPS Using Satellites, Other Observation Platforms, and Their Associated Data Processing Systems (Parallel with Sessions 11 & 12)
Thursday, 13 January 2000, 10:30 AM-5:15 PM
Previous paper Next paper