Mdbs: a Modeling and Database System to Support Research in the Earth Sciences 1 Computational Support for Scientiic Investigations 1.1 Goals and Strategy We Gratefully Acknowledge the Work of Our Graduate Students

An interdisciplinary team of computer scientists and EOS earth-science investigators is jointly investigating the concept of a modeling and database system (MDBS). MDBS is intended to support large-scale earth science investigations by complementing EOS-DIS in terms of support for high-level modeling and data management. In particular, MDBS is intended to facilitate the iterative development of scientiic models in data-intensive and computationally intensive applications. It is based on a simple characterization of scientiic activity and on an advanced data model that together capture the essence of scientiic modeling and database activities in terms of large lattices of representational domains and associated transformations that are employed to organize scientiic knowledge, data and algorithms. A high-level, declarative modeling and database language (MDBL) that is based on this data model and on four fundamental sets of scientiic operations provides support for scientiic modeling activities and for easy access to distributed databases of datasets and algorithms. It permits the eecient resuse of code for building new algorithms and supports key aspects of scientiic modeling activities, such as lineage and error tracking in datasets, and for the concepts of model and project. We are applying a version of MDBS in order to solve several earth science research problems that have arisen in EOS investigations. The overall goal of our scientiic database research project is to investigate, design and develop computational support that permits scientiic investigators to achieve their scientiic goals more eeciently. Our investigations of several earth science research projects have indicated that many scientiic teams focus up to 50% of their attention on computational issues that are irrelevant to their scientiic research. Hence we believe that signiicant increases in eeciency are possible for applications in which computation is a major activity. It is critical, however, to provide support that scientists will adopt and use. In particular, our support takes the form of a Modeling and Database System (MDBS) which incorporates support for: 1. the scientiic process of iterative model development;