Query Manager (ldl++) Information Repository (postgres) Visualization Manager (idl) Extracting Spatio-temporal Patterns from Geoscience Datasets

distribution of a coupled atmosphere-ocean general circulation model. \Comparison of observed and simulated cyclone frequency distribution as determined by an objective method", At-Figure 10: QUEST visualization manager animation of SLP elds during the life of the cyclone track shown in Figure 9. 6 Conclusion Currently, scientists have access to large repositories of observational and model data, but lack high level analysis tools with which to investigate the data. Rather than being able to issue queries such as \Show me motion sequences that include hurricane (cyclone formation or temperature gradient patterns) that developed in the Southern Hemisphere", scientists are forced to issue queries like \Retrieve all temperature records produced between January 1, 1980 and De-cember 25, 1991". We have developed a prototype system called QUEST to provide content-based query access to massive datasets. To demonstrate the utility of spatial-temporal features as high-level indexes into terabyte datasets, we presented an algorithm for extracting cyclone tracks from sea level pressure data generated by a AGCM. A user scenario was presented to illustrate the interaction between a scientist and QUEST beginning with a query to select a subset of cyclones, and ending with the visualization of sea level pressure elds associated with the selected cyclone. Much work remains to be performed to extend the functionality of QUEST. Work is underway to employ massively parallel processors such as the Intel Paragon, to extract spatio-temporal patterns from existing geoscience datasets, and to extract these features on-line as the model is executing. On-line extraction of features also provides the ability to com-putationally steer the model's execution in the event that desired features are not being produced by the model. A nal area of current research is the use of Distributed Object Management Systems (DOMS) to provide support for integrating data analysis, visual-ization and data management in a heterogeneous distributed environment. Acknowledgements We sincerely acknowledge support from NASA HPCC grants #NAG 5-2224 and #NAG 5-2225. We thank J. Sphar for supplying the AGCM output and C.C. Ma for valuable discussions on cyclogenesis. We also thank William Cheng for programming support on the QUEST GUI. Computer time for the AGCM integration was provided by the DOE AMIP.Figure 8: QUEST query hit browser. Figure 9: QUEST visualization manager plot of a selected cyclone track. Figure 6: QUEST GUI for selecting a feature query icon. Figure 7: QUEST panel for specifying the spatial, temporal, and magnitude extents for desired cyclones. QUEST …