This grant provides support to develop a methodology for the quantitative assessment of confidence in predictions of the motion of ash clouds caused by volcanic eruptions. By combining statistical modeling, stochastic analysis, and the tools of computational science together with a widely-used volcanic particle dispersion code, this grant will provide tools to assess real time predictions of ash cloud motion that accounts for varying wind conditions and a range of model variables. More specifically, the PUFF computer code will be used as the ash cloud simulation tool. PUFF, which propagates an airborne ash plume, will be integrated with a volcanic eruption model called BENT. BENT outputs ash distribution as a function of time and height. This integration substitutes for the unknown PUFF input parameters several BENT inputs that are better (but not entirely) constrained by physics. This integrated model will be used to develop a framework for analyzing the effects of uncertain parameters and of differing windfield models, on the output distribution for the cloud location. Polynomical chaos quadrature and stochastic integration techniques will be used to provide a quantitative measure of the reliability (i.e. error) of those predictions.
Understanding the effects of uncertainty in dispersion model predictions are important to assess and mitigate potential hazards associated with volcanic ash clouds. There are economic and sociologic impacts both near the volcano and far downwind from the eruption. For example, the airline industry decides flight schedules, routes, and fuel consumption estimates, based in part on dispersion model forecasts. The results of this research will provide decision makers and civil protection authorities with a framework for evaluating hazard risk. Importantly, this methodological development is independent of the specific PUFF code, so the framework for prediction and reliability analysis can be directly applied to other ash cloud codes, and more generally to other hazard models -- such as radiological and chemical plume dispersion, and gas release from truck or rail accidents.
Intellectual Merit This project brought together volcanologists, remote sensing scientists, engineers and statisticians in a unique team to understand the complexities of probabilistic modeling and develop a workflow system to classify, quantify and ultimately reduce uncertainties in volcanic cloud dispersion modeling. The work was a collaboration of scientists from University of Alaska Fairbanks and the State University of New York-Buffalo. The team built a system that incorporated the dispersion model input variability and numerical weather prediction ensembles to develop a full probabilistic modeling of ash clouds. This system was developed to be applicable for any volcanic eruption worldwide. A new coupled system was built, through collaborative research at the two universities, called Puffin or Bent-Puff, integrating a one-D plume rise model and Lagrangian dispersion model. This tool can be used for both deterministic and probabilistic dispersion modeling and is available online through the NSF funded Vhub website, https://vhub.org/. The project workflow can be used to improve dispersion model input uncertainty and be compared to probabilistic satellite observations of ash loading to increase confidence in each technique. The final product from the project can be used to assist in volcanic ash cloud decision making and hazard assessment. Broader impacts The NSF IDR project involved graduate students at each collaborative institution. Students developed new research as part of their Masters and PhD programs, publishing in peer reviewed literature and their final graduate thesis. The work developed in these graduate programs was a collaboration of each of the groups involved in the NSF IDR project, as can be seen from the author list of the papers with the students as lead authors. The developed tools are available for use on the NSF funded Vhub, https://vhub.org/, including the coupled one-D plume rise model and Lagrangian dispersion model. As part of the research, both faculty and students attended U.S. based and international research meetings, to present their work and discussion the project work with the larger research community. Also, a workshop was included in the project to bring together researchers to increase the broader impact of the projects' tools and to also increase the capacity of these researchers in probabilstic modeling and build upon and use the tools from this NSF IDR project. Finally, numerous peer reviewed papers were published, with many having a student as the lead author, thus building the next generation of geoscience, engineering and mathematical modeling scientists.
