Building an Uncertainty Modeling Framework for Real-Time VATD

When forecasting the future location of volcanic-ash clouds, uncertainties exist in the input parameters used in dispersion modeling and in the weather prediction data used for modeling the advection terms. Recent developments have shown that probabilistic modeling provides the tools to assess the variability in downwind ash concentrations. We show a probabilistic modeling approach where ensembles of forecasts are generated from a suite of simulations using a coupled one-dimensional plume model and a Lagrangian dispersion model. This approach produces charts of the probability of ash-cloud concentrations and mass loadings exceeding user-defined thresholds. We focus on the initial plume uncertainties and discuss how uncertainties in numerical weather prediction data could also be applied within our approach. Our results show how, by assigning the initial likelihoods of input parameters, the probabilistic approach can produce mean ash concentrations and mass loadings as well as probabilities of breaching a defined threshold. We show how, given the variability in the inputs, the probabilistic modeling can be used to assess the confidence in the ash-mass loadings. This is critical for real-time volcanic-hazard assessment and our approach illustrates how a new tool could be developed for those in decision support.