A Data-Driven Soft Real-Time Expert System for Producing Coral Bleaching Alerts
Date of Award
Doctor of Philosophy (PhD)
Graduate School of Computer and Information Sciences
Maxine S. Cohen
In the Florida Keys there are many physical, chemical and biological events of interest and concern to personnel of the Florida Keys National Marine Sanctuary, marine biologists, oceanographers, fishermen and divers. Large volumes of continuously generated meteorological and oceanographic data from instruments in the SEAKEYS (Sustained Ecological Research Related to Management of the Florida Keys Seascape) network help to understand these events. However, since no one has the time to look at every printout of data from every station, every day, seven days a week, it is highly desirable to have an automated system that can monitor parameters of interest and produce specialized alerts of specific events, as indicated by prescribed or abnormal ranges, or combinations of parameters. A soft real-time expert system was developed to produce such alerts based on data input from the SEAKEYS network. The prototype system collected data from the Sombrero Reef station in the network and produced automated e-mail and World-Wide Web alerts when conditions were thought to be conducive to, or predictive of, coral bleaching, which occurs under environmental conditions stressful to corals. Configuration of the system included a point system for three coral bleaching models (high sea temperature only, high sea temperature plus low winds, high sea temperature plus low winds plus low tide). The approach is an important development in the use of knowledge-based systems to solve environmental problems, as it provides for knowledge synthesis (in the form of data summaries) from any environmental ASCII data stream or table, be it real-time or not.
James C. Hendee. 2000. A Data-Driven Soft Real-Time Expert System for Producing Coral Bleaching Alerts. Doctoral dissertation. Nova Southeastern University. Retrieved from NSUWorks, Graduate School of Computer and Information Sciences. (579)