The Environmental Scenario Generator

The Environmental Scenario Generator (ESG) is a distributed software system designed to allow a modeling and simulation customer to intelligently access distributed environmental data archives for inclusion and integration with model runs. The ESG is built to solve several key problems for the modeler. The first is to provide access to an intelligent "data mining" tool so that key environmental data can not only be retrieved and visualized but, in addition, user defined conditions can be searched for and discovered. As an example, a user modeling troop landing might want to model the result of an extreme rain event prior to deployment. Without a tool such as ESG the simulation coordinator would be required to know:

· For my landing region what constitutes an extreme rain? · How can I find an example in the real data of when such an event occurred? · What about temporal or spatial variations to my scenario such as finding the wettest week, month or year?

If we consider combining these questions across multiple parameters, such as temperature, pressure, etc. and then add multiple regions and seasons the problem reveals itself to be quite daunting.

The second hurdle facing a modeler who wants to include real environmental effects in the simulation is how to manage many discrete data sources. Often simulation runs face tight time deadlines and lack the manpower necessary to retrieve data from across the network, reformat it for ingest, regrid or resample it to fit the simulation parameters, then incorporate it in model runs. Even if this could be accomplished, what confidence can the modeler have in the different data sources and their applicability to the current simulation without becoming expert in each data type? The unfortunate side effect of this is that the environment is often forgotten in simulations or a single environmental database is created and "canned" to be replayed again and again in the simulation.

The ESG solves this problem for the modeler by providing a 100% Java platform independent client with access to both data mining and database creation capabilities on a network distributed parallel computer cluster with the ability to perform fuzzy logic based searching on a global array of environmental parameters. The system is designed to allow the user to specify the desired spatial, temporal, and parameter conditions in fuzzy linguistic and/or numeric terms and to receive a ranked list of events best matching the desired conditions in the historical archive. Once discovered the client application can request temporal and spatial visualization products from the data, browse the climatology of the particular region and parameters selected, and request delivery of the data formatted for inclusion model runs. By providing intelligent instantaneous access to real data it ensures that the modeler is able to include realistic, reliable and detailed environments in their simulation applications.

This paper will present the results of data-mining, visualization, and a domain integration tool developed in a network distributed fashion and applied to environmental modeling.