On Earth, weather arguably has the longest history of observations with extensive coverage at intensive rates. Recent advances in weather observation technologies further accelerate weather data production from in-situ automatic sensors, weather radars, and meteorological satellites. Traditionally, in-situ observations are stored as text files or spread sheets indexed by station identifiers or dates of observations. Recently, relational database technology has been employed to archive observations from ground weather networks, such as the Oklahoma Mesonet. The use of relational database technology (e.g. Oracle, Informix, and Microsoft SQL server) enables support of extracting weather data records by a set of criteria. For example, the user may select stations and days with temperature observations greater than 90oF. Satellite and radar data, on the other hand, are usually archived in binary files using various data formats, such as GRIB, HDF, NetCDF, and NIDS/NEXRAD. Data retrieval is often restricted to area and time of interest.

What has been missing in the above forms of weather data archives is information about the structures and development of weather systems, which we refer to as weather intelligence. Meteorologists often loop weather data or images to examine the structure and development of storms or cyclones. Looping weather data enables meteorologists to perceive the development and movements of weather systems, and therefore to gain insights of how the weather has evolved and furthermore to forecast how the evolution will continue. If we can incorporate weather intelligence into weather databases, then weather data retrieval can be based on spatiotemporal characteristics and structures of a given weather system. For example, it will be able to support the selection of weather data representing lines of convective storms that initiated from the Texas panhandle, moved eastward, and produced supercells in central Oklahoma from 2000-2002. In doing so, meteorologists can retrieve weather data based on the defined weather characteristics to assess weather forecasting, compare weather cases, and, with other environmental data, such as land cover, terrain, and soil moisture, examine regional or local influences on weather development.

We hereby propose a GIS framework to enrich weather databases by incorporating weather intelligence. The use of GIS technology is necessary because information about the structure and development of weather systems cannot be derived without careful examination of spatial and temporal relations in weather data. Distinguished from the other information technologies, GIS offer a suite of functions to analyze and relate spatial data. While GIS lack functions to handle temporal data, we have developed a data structure to enable temporal indexing and tracking. The proposed GIS framework integrates both object and field (grid) representations to capture spatiotemporal characteristics of weather systems in a hierarchical structure. Using precipitation data as an example, we have developed a prototype to demonstrate the proposed GIS framework and its support for spatiotemporal query and analysis of weather systems.