Validation of ash cloud modelling with satellite retrievals: a case study of the 16-17 June 1996 Mount Ruapehu eruption

Abstract

Volcanic ash can be hazardous to aviation and human health; hence, there is a need to be able to accurately forecast the dispersion of ash clouds in real time. This relies, in part, on the choice of suitable input parameters for the modelling of a given eruption event, often when data from the actual eruption are not yet available. In this paper, the sensitivity of a coupled modelling system, consisting of the Lagrangian particle dispersion model FLEXPART and the Weather Research and Forecasting (WRF) model, namely FLEXPART-WRF, to varying eruption source parameters is examined through comparison of modelled ash clouds with satellite data for the 16-17 June 1996 eruption of Mount Ruapehu, New Zealand. The model is evaluated using the probability of detection, false alarm ratio and Critical Success Index statistical analyses. The eruption source parameters considered in the ash cloud modelling are: the eruption duration, the mass eruption rate, the particle density, the number of particles, the plume height, the particle size distribution and the plume ratio (defined as the ratio of the thickness of the laterally spreading ash cloud at the plume top to the height of the plume). We allowed the plume height, plume ratio and particle size distribution to vary in order to carry out the sensitivity experiments. Our results indicate that for this eruption: (1) the plume ratio has a large effect on the model prediction of ash clouds dispersion; (2) the uncertainties associated with the plume heights retrieved from satellite data do not have a significant impact on the model prediction of ash cloud dispersion; and (3) the particle size distribution of fine ash has little effect on the model prediction of ash cloud dispersion. Moreover, the results show that the ash cloud forecasts are only accurate for a comparatively short period of time (up to 11 h) due to the limitations associated with WRF model. Further analysis is needed to investigate the general applicability of the method.