Using EnKF Data Assimilation to Improve Predictions of Volcanic Ash Dispersion

Abstract

Accurate forecasts of volcanic ash dispersion patterns as well as the ability to quantify cloud top heights and mass loading properties are critical in mitigating the risks associated with volcanic eruptions. Currently, physical dispersion models and satellite observations are the primary methods used to forecast and monitor volcanic ash clouds, but each has limitations. Leveraging their respective advantages and effectively integrating physical dispersion models and satellite data are key to improving the accuracy of volcanic ash dispersion forecasts. The ensemble Kalman filter (EnKF) can be used to optimize model predictions of volcanic ash dispersion by assimilating observational data, thereby combining the complementary benefits of predictions and observations. In this study, an EnKF-based data assimilation method is presented that combines satellite observations with a volcanic ash dispersion model (HYSPLIT) to further improve the accuracy of volcanic ash dispersion predictions. The eruptions of Eyjafjallajökull (Iceland) during May 2010 were used to evaluate the performance of this approach. The results indicate that the forecasts optimized using data assimilation show marginal improvements in the accuracy of the spatial distribution, mass loading, and cloud top height of the ash cloud at times even outperforming the corresponding satellite retrievals. This method addresses the issue of insufficient model initialization accuracy and continuously optimizes predictions through the ongoing assimilation of satellite observations, thereby enhancing the accuracy of model predictions.