Provision of this service requires both observational and modelling capabilities. The detection and monitoring of ash clouds in the atmosphere is strongly dependent on satellite data, particularly from the Japan Meteorological Agency's Himawari-8 satellite in geostationary orbit. For predicting the near-term positions of volcanic ash clouds, the HYSPLIT dispersion model is utilized. There is an increasing requirement from airlines for guidance on the uncertainties associated with warnings for volcanic ash clouds and for quantitative estimates of the concentration of ash to better enable the management of risk. These are both challenging problems.
High-quality forecasts of ash clouds require estimates of the volcanic source term, which includes the total mass eruption rate within the initial eruption and its distribution in space, time, and amongst particles of different sizes. Dispersion model forecasts also require input from high-quality Numerical Weather Prediction (NWP) model data and an adequate representation of relevant physical processes within the dispersion model. These aspects have been the subject of research and development activity within the Bureau in support of the aviation industry.
For observations, the Volcanic Cloud Analysis Toolkit (VOLCAT) developed by NOAA/NESDIS has been implemented. The algorithm uses multi-spectral satellite imagery to automatically identify and make quantitative estimates of ash cloud properties such as height and mass loading. Case studies of several high-level eruptions since July 2015 (when Himawari-8 became operational) have been used to evaluate the performance characteristics of the detection and retrieval capabilities of the algorithm. During the long-lived Rinjani eruption of November 2015 that disrupted air traffic in the vicinity of Denpasar, Bali for several days, detections of the ash cloud were very good. However, the retrieved heights appeared to be problematic, being too high compared to the heights determined from a coincident CALIPSO satellite overpass. For the eruptions of Soputan in early-January 2016 (in the peak of the wet season), detections were more problematic. At times, deep convective clouds were present, leading to missed detections of ash clouds or otherwise creating incorrect detections. As in the Rinjani example, problems with the inaccurate cloud top heights as retrieved from the satellite were again noted.
Much of the R&D effort around volcanic ash has been on improvements and extensions of dispersion modelling. One area of focus has been an improved representation of the physical processes within the model, particularly around the removal of ash from the atmosphere. This effort has focussed on improving the fall speed parameterization for volcanic ash within the model and the inclusion of more realistic particle size distributions for ash. The effects of the 'wet deposition' process, involving the removal of ash by rain and cloud, have also been investigated. This process can remove significant amounts of ash in a brief period and is a particularly important process to consider in the rain-prone regions of the Maritime Continent.
A significant outcome of this research is the creation of the Dispersion Ensemble Prediction System (DEPS) for volcanic ash. DEPS provides a web-based, demand-driven user interface that defines a set of dispersion model simulations that result in an estimate of the probability that ash amounts exceeding pre-defined thresholds are present within the domain. This approach allows for the effects of meteorological uncertainty to be considered in the forecast. The probability estimates are derived by initializing HYSPLIT with the Bureau's 24-member global NWP ensemble plus deterministic runs from local and international models. Source term parameters for initializing the runs, including the height of the plume, an estimate of the mass eruption rate, the plume type (i.e. column or umbrella cloud) and the particle size distribution within the cloud are customizable, although well-considered default values are also available. With the Bureau's current computing resources, the full ensemble process can complete in approximately 10 minutes. The system is currently being operationalized into the Bureau.
Continuing work on DEPS is focused on adding the effects of source term uncertainty into the probabilistic calculations. Uncertainty in the source term and model initialization can result in significant errors in the resulting forecast. The approach here is to combine real-time observations of the volcanic cloud with inverse modelling techniques to optimize the source term, providing the best match of the dispersion modelled cloud to that observed. This assists in narrowing the large uncertainty inherent in the source term. Initially, the source term optimization will focus on using the spatial boundaries of the ash cloud, as either determined from the automated satellite detections or from the manual analysis by the forecaster. This approach optimizes the top and bottom heights of the initial plume. Later efforts will focus on optimizing the mass eruption rate of the volcano by considering the quantitative mass loading retrievals from satellite. This will be an important step towards providing reliable ash concentration forecasts desired by the aviation industry.