Producing coherent and accurate diagnoses and predictions is the core of the weather forecasting process. Forecasters integrate information from a range of sources including raw observations and output from meteorological forecasting software/packages, and make decisions regarding proper reporting. The quality of their decisions depends on both their meteorological and local-empirical understanding of the atmosphere. The complexity and diversity of the domain knowledge, as well as the dynamic nature of the process itself, are major obstacles for a centralised and a comprehensive approach to the development of Decision Support Systems for this process.
Existing tools deal with separate small decision steps and are developed independently in each weather forecasting office. The resulting decision support systems are at the code level with no explicit knowledge level description, which limits the re-usage by different technologies. In addition, they are location dependent, often without proper documentation, which is a hurdle to knowledge sharing.
We present an ontology-based approach to the weather forecasting process, which overcomes these problems. Our approach facilitates both knowledge management (sharing) and the efficient construction of weather forecasting decision support systems, utilising different technologies (reusing). Our ontology is used to structure and store knowledge about the weather forecasting process. We implement it by using Protege - a frame-based tool. We illustrate the power and flexibility of the approach by describing how it can be used to support the construction of Bayesian Networks and rules for rule- based (decision tree)