Algorithm Development and Mining (ADaM) System for Earth Science Applications

The ADaM (Algorithm Development and Mining) system, developed at the Information Technology and Systems Center at the University of Alabama in Huntsville, is one such mining tool. The system provides knowledge discovery and data mining capabilities for data values, as well as for metadata, and catalogs the information discovered. It contains algorithms for detecting a variety of geophysical phenomena to address the needs of the earth science community. The portable client-server architecture is also network accessible, which allows the system to be used as an application at a data archiving center or on a user's desktop workstation. The ADaM system architecture consists of three basic types of modules: input filters (readers for different data formats), processing modules (general-purpose algorithms and user-defined algorithms) and output filters (writers for different data formats). New modules can be easily added to this extensible system. These modules can be arranged as needed for a particular job, with results from each step passed to the next one in line. The use of data input filters, specialized for a variety of data types, is instrumental in simplifying the development of the processing and output operations. The selected input filter translates the data into a common internal structure so that the processing operations can all be written to a single data representation. This allows the addition of new operations to the system without having to address input data format problems. Similarly, the addition of a new input filter provides access to the entire suite of processing operations for the datatype in question. The mining system currently allows over 80 different operations to be performed on the input data stream. These operations vary from specialized atmospheric science data set specific algorithms to different digital image processing techniques (ranging from generic spatial filters to specialized texture feature extraction modules), processing modules for automatic pattern recognition (different classifiers), machine perception, neural networks and genetic algorithms. ADaM has been utilized in a variety of earth science applications. A short summary of some of these different applications is given below:

* Cumulus Cloud Detection: Boundary layer cumulus clouds over land are difficult to detect in satellite data, owing to low contrast in both visible and infrared channels. For Geostationary Operational Environmental Satellite (GOES) satellite data the problem becomes severe, as the infrared channel resolution is 4km, compared to 1km in the visible channel. A study was conducted analyzing three of the different image processing and pattern recognition techniques available in ADaM for cumulus cloud detection. These were classifiers based on 1) texture and spectral features, 2) edge detection and spectral features, and 3) purely spectral features.

* Phenomena Detection: The ability of ADaM to search or mine for particular data values or geophysical phenomena within a specified data product has been actively utilized on several projects. Special Sensor Microwave/Imager (SSM/I) data is being mined during ingest to detect and locate Mesoscale Convective Systems. In a different project, Advanced Microwave Sounding Unit (AMSU) data is mined in real time to locate tropical storms and estimate their maximum wind speeds. Operational forecasters at National Hurricane Center are utilizing this information to aid in their analysis and tracking of hurricanes.

* Custom Order Processing: An innovative data storage and retrieval system is being built around the flexible and extensible ADaM processing architecture. This system takes advantage of ADaM ability to provide an extensible framework for many types of processing including subsetting, format conversion, gridding and mapping, in addition to data mapping. The main emphasis of this new information system is to allow the end user flexibility and ease in accessing and utilizing data.