Application of a novel ensemble mean technique to probabilistic forecasting of volcanic ash transport using HYSPLIT

Abstract:

Forecasts from an atmospheric transport and dispersion model are often used to respond to release of a hazardous substance to the atmosphere. Such forecasts often contain significant uncertainties which are not conveyed by the single deterministic solution. Thus, ensemble modeling systems are being developed. A critical aspect of such a system is the ability to generate a mean, or central tendency, for all ensemble members that may serve as the best single predictor of the model state. To this effect we investigate the implementation of a novel technique for computing the ensemble mean that minimizes positional differences between members relative to the traditional arithmetic mean. We show how this method preserves extrema in the ash concentrations fields and can be used as a direct measure of ensemble spread/dispersion. For this initial study, we use the HYSPLIT transport and dispersion model to produce ensemble forecasts for retrospective volcanic eruption cases and discuss the potential for operational use.