CAREER: Uncertainty Propagation and Data Assimilation for Toxic Cloud Prediction

This Faculty Early Career Development (CAREER) award focuses on developing mathematical tools for accurate characterization and propagation of uncertainty in mathematical models, and fusion of model output with sparse, noisy data to determine estimates of the actual physical phenomenon and statistical measure of confidence in those estimates. The principal goals of this CAREER award are twofold: (i) to understand how the uncertainty of input variables and the random forcing of winds affect the output of the dispersion model, and (ii) to provide a prediction of toxic cloud movement, together with quantitative measures of confidence in that prediction. Uncertainty analysis of dispersion of toxic clouds is becoming one of the most important components for timely and accurate threat assessment from natural or man-made incidents (such as Chernobyl or Eyjafjallajökull incidents). Consequent to the toxic material release, response organizations and industries make decisions based on predictions of the cloud motion, with little knowledge or appreciation of the reliability of those predictions. The quantitative understanding of uncertainty is essential when predictions are to be used to inform policy making or mitigation solutions where significant resources are at stake. For example, an understanding of uncertainty in model predictions can play an essential role in the acceptance of the need to vacate a city in case of toxic material release where the cost of different choices varies by millions of dollars and human lives.

If successful, this research work will provide means for evaluating hazard risks in space and time. An important practical outcome will be the ability to generate toxic material hazard maps. This CAREER project also integrates educational outreach efforts into the research plan with the goal to increase the diversity and the number of students from minority groups in science and engineering in association with existing programs on campus. In addition, the PI will design a ``Self-Help'' tutorial web-site to inform and educate both students and researchers about nonlinear filtering and uncertainty characterization concepts. Research and education accomplishments completed, as part of the plan will be described in journal publications, and conference presentations.