On the efficient use of satellite data to improve volcanic ash dispersion modelling

Volcanic ash poses a hazard to aviation, with ubiquitous jet powered aircraft being particularly vulnerable to its effects. Consequently, information about the location of ash and its predicted dispersion pattern is crucial to aviation stakeholders during eruption events. The Australian Bureau of Meteorology performs this function over a region that includes Australia, Indonesia, Papua New Guinea, and the southern Philippines. In current practice, information about the location of ash is obtained mainly from satellite imagery. Apart from the difficulty of identifying ash reliably by remote sensing, other properties of the ash cloud such as its altitude are still subject to significant uncertainties even when ash is successfully detected. On the other hand, the future location of ash is predicted by the use of dispersion models. In Lagrangian versions of these models, such as that used operationally by the Bureau, model particles, representing atmospheric pollutants such as ash, are released at specified locations and transported by gridded atmospheric winds and sub-grid turbulence. Information obtained from satellite retrievals of ash, such as its altitude, is frequently used to initialize the dispersion model. However, this is not in general done in a self-consistent and efficient manner by also taking into account how the simulated and observed ash distributions compare at a given analysis time. In this paper, we demonstrate how information about the ash distribution may be optimally integrated within the dispersion model. This is done by running a suite of dispersion model simulations with different values of the altitude of the initial ash cloud in the analysis phase of the algorithm. Pattern correlations are used to compare the simulations with the observed ash cloud. The altitude values which provide the best matches between simulations and observations are then used to initialize the dispersion model in the forecast phase of the algorithm. In this paper, we use the eruption of the Puyehue-Cordon Caulle volcanic complex, located in Chile, which occurred in June 2011 as a case study. Ash from this eruption circumnavigated the globe and caused significant disruption to aviation services in the southern hemisphere mid-latitudes, including southern parts of Australia. We demonstrate that the altitude estimates obtained from the method presented in this paper are consistent with the altitude measurements obtained from lidar instruments and superior to estimates obtained from satellite imagery alone. We also demonstrate the utility of satellite derived probabilistic forecasts of ash dispersion in the presence of significant uncertainty in the ash cloud altitude, as opposed to deterministic forecasts currently in use.