Integrating Multiple Transport and Diffusion Codes Into A Unified Software Architecture

Transport and Diffusion (T&D) models can be used to predict the path and distribution of chemical and biological agents in both civilian and military settings. There are a variety of T&D models in wide use including SCIPUFF and ALOHA. The National Center for Atmospheric Research (NCAR) Mesoscale Model Version 5 (MM5) can be used to provide such T&D models with high resolution wind and weather information. Coupled together, MM5 and a T&D model can be put to use operationally for various emergency response applications.

Taking advantage of such a coupling, TASC has developed a unique application called BestTrack. MM5 is used to feed detailed weather information simultaneously to several T&D models. The models are used to simulate the evolution of one or more chemical plumes, each model using its own numerical schemes for the prediction. BestTrack takes all of the predictions and combines them into an ensemble prediction. The ensembling process may be purely statistical (50% Model A + 50% Model B). Alternatively, the process may involve an evaluation of model strengths. That is, Model A may be favored over Model B because it performs well in complex terrain. Finally, the forecast may be based on a "best case" or "worst case" scenario. Each scheme demonstrates the ability to apply rule-based decision making to provide an enhanced forecast.

Graphical User Interfaces (GUIs) have been designed to accept inputs, so that all models can be run on common weather data. Technology has been developed to allow for chemical and biological data from one model to be transferred to another. For example, Model A may have the data it needs to predict the transport of anthrax but no data for VX. Model B may have the characteristics of both anthrax and VX in its database. Using BestTrack, VX data from Model B is transferred to Model A so that both models can provide a common prediction used for a VX ensemble forecast. This not only enhances the functionality of each BestTrack model, it provides a common database from which all models obtain agent information.