How accurate are volcanic ash simulations of the 2010 Eyjafjallajökull eruption?

Abstract

In the event of a volcanic eruption the decision to close airspace is based on forecast ash maps, produced using volcanic ash transport and dispersion models. In this paper we quantitatively evaluate the spatial skill of volcanic ash simulations using satellite retrievals of ash from the Eyjafjallajökull eruption during the period from 7 to 16 May 2010. We find that at the start of this period, 7-10 May, the model (FLEXible PARTicle) has excellent skill and can predict the spatial distribution of the satellite-retrieved ash to within 0.5 ∘ × 0.5 ∘ latitude/longitude. However, on 10 May there is a decrease in the spatial accuracy of the model to 2.5 ∘ × 2.5 ∘ latitude/longitude, and between 11 and 12 May the simulated ash location errors grow rapidly. On 11 May ash is located close to a bifurcation point in the atmosphere, resulting in a rapid divergence in the modeled and satellite ash locations. In general, the model skill reduces as the residence time of ash increases. However, the error growth is not always steady. Rapid increases in error growth are linked to key points in the ash trajectories. Ensemble modeling using perturbed meteorological data would help to represent this uncertainty, and assimilation of satellite ash data would help to reduce uncertainty in volcanic ash forecasts.

Estimation of optimal dispersion model source parameters using satellite detections of volcanic ash

“…This is the approach that would be taken operationally during such an event. Recently, the problem of initializing ash dispersion models from satellite data has received some attention [ Wilkins et al , , ; Dacre et al , ; Crawford et al , ]. However, it is not the intention to address these difficulties here but merely to illustrate what a typical result would be using the tools currently available at the Darwin VAAC and to illustrate later on in this paper that the height of the ash cloud can be inferred from the observed ash transport patterns.…”

Epistemic uncertainties and natural hazard risk assessment – Part 1: A review of different natural hazard areas

“…radar) are subject to uncertainties (Arason et al, 2011;Folch et al, 2012), and plumes from weak eruptions such as Eyjafjallajökull can become distorted by local winds, increasing plume height measurement uncertainty and thus affecting the MER calculation (Webster et al, 2012). Meteorological data can also introduce uncertainty to dispersion forecasts, and can lead to cumulative transport errors (Dacre et al, 2016). All of these factors represent primary epistemic uncertainties in the application of such models.…”

Characterizing the Atmospheric Conditions Leading to Large Error Growth in Volcanic Ash Cloud Forecasts

“…The diagnostic used to characterize the synoptic-scale flow conditions is the 2D horizontal flow-separation diagnostic introduced in Dacre et al (2016). This flow separation is calculated as the velocity gradient perpendicular to the flow:…”

“…To represent dispersion on scales smaller than this horizontal diffusion is applied. The diffusion represents the dispersion by unresolved eddies and acts to increase the vertical and lateral spread of volcanic ash clouds (Dacre et al 2015). This approach assumes that the small-scale dispersion processes are of an eddy viscosity character and thus can be represented using a Gaussian description (Pasquill and Smith 1983).…”

“…These particles experience an accumulation of errors in the wind field leading to larger positional errors on average than particles with shorter residence times. Dacre et al (2016) examined the second of these sources of error by performing hindcast simulations of the Eyjafjallajökull eruption (using analysis wind fields). They showed that generally skill decreases as the residence time of ash increases but that the rate of skill decrease depends on the meteorological situation.…”