Ash cloud detection and forecasting became a research priority and societal concern (Bonadonna et al., 2014) after the 2010 Eyjafjallajökull (Iceland) eruptions caused widespread flight cancellations and an estimated U.S.$5.0 billion impact on global gross domestic product (GDP) (5 day period up to 24 May 2010; [The Economic Impacts of Air Travel Restrictions Due to Volcanic Ash, 2010 [www.oxfordeconomics.com/my-oxford/projects/129051]). In 2011, Grímsvötn volcano (Iceland) again caused widespread cancellations and in 2014 an aircraft sustained ∼AU$20.0 million engine damage after flying through a volcanic cloud from Gunung Kelud (Java) (The West Australian, 22 February 2014 [au.news.yahoo.com/thewest/a/21618743/20m-ash-cloud-damage-bill/]). Since 1953, over 129 incidents involving aircraft and volcanic ash clouds have occurred and more than 79 sustained serious damage (Guffanti et al., 2010). Fundamental to managing risk and economic impacts is forecasting of volcanic ash clouds (e.g., Casadevall, 1994), which depends on accurate models of ash sedimentation.

Volcanic cloud modeling consists of a description of the source (characteristics of the eruption column such as geometry, height, mass eruption rate, vertical separation of ash and gases) (e.g., Mastin et al., 2009; Moxnes et al., 2014) and the sink (cloud microphysical processes and sedimentation) (e.g., Costa et al., 2010). Source and sink terms are coupled to a model describing atmospheric dynamics and microphysics to simulate particle transport (e.g., Stohl et al., 2011). Quantification of the source has been improved, while the sink has received less attention.

Volcanic tephra comprises of particles ranging from meters to sub-micrometer size (Durant et al., 2010). ‘Classical’ models of tephra sedimentation focus on gravitational settling of single particles (e.g., Sparks et al., 1997). Large particles fall out quickly, less-affected by inertial and viscous drag; good agreement is achieved between predicted and observed coarse deposit characteristics within 10–100 km of the volcano (e.g., Pyle, 1989). In distal (100–1000 km) sections, the particle size is much finer and observed and predicted characteristics agree poorly (e.g., Fierstein and Nathenson, 1992). Calculating particle fall based on Reynolds number in different flow regimes (Bonadonna et al., 1998; Rose et al., 1993) and correcting for non-spherical particle shape provides some improvement (e.g., Bursik, 1998; Ganser, 1993; Riley et al., 2003; Wilson and Huang, 1979).

The 18 May 1980 eruption of Mount St. Helens (USA) provided the opportunity to observe distal ash sedimentation (e.g., Sarna-Wojcicki et al., 1981). Particle aggregate fallout was observed over a vast area (e.g., Sorem, 1982) and enhanced fine-ash (<63 μm; Mastin et al., 2009) sedimentation was required to reconcile observations and predictions (Carey and Sigurdsson, 1982). Aggregation (or coagulation) of ash particles (e.g., Brown et al., 2012) into larger composite particles with a higher terminal fall velocity leads to a reduction in the atmospheric lifetime of fine ash. As most operational forecasting models do not account for aggregation, proximal ash deposition may be underestimated and distal airborne ash fraction overestimated, which is particularly relevant for forecasting aviation hazards (e.g., the pan-European airspace closures in 2010 were largely guided by model-based predictions). Even sophisticated source inversion modeling constrained by satellite observations (e.g., Stohl et al., 2011) may have similar limitations: while the source term is optimized, the ash cloud forecast is dependent on the sink description; errors increase as the forecast is extended into the future.

Cloud instabilities reduce fine ash atmospheric lifetime and have been observed on many recent volcanic clouds: ‘Ash veils’ on the base of a 3 h old ash cloud from the 1990 eruption of Mount Redoubt, Alaska, USA (Hobbs et al., 1991); ‘finger-like protrusions’ on a cloud from the 1997 Montserrat (Caribbean) eruptions (Bonadonna et al., 2002); and ice-rich, turbulent mammatus lobes (Schultz et al., 2006) on the 1980 Mount St. Helens eruption cloud (Durant et al., 2009). The properties of an ash deposit, i.e., poor sorting, polymodal size distributions and localized lateral thickness variations, may provide evidence of particle settling influenced by cloud instabilities (e.g., Carazzo and Jellinek, 2013). Fine-ash settling through the ocean water column may also be enhanced by transport in gravitational convective instabilities (Carey, 1997; Manville and Wilson, 2004).

Manzella et al. (2015, p. 211 in this issue of Geology) used high-resolution imagery to link observations of gravitational instabilities on the Eyjafjallajökull eruption cloud to rapid fine-ash sedimentation. Layers at the cloud base peeled away forming ‘fingers’ that carried fine-ash particles downward at ∼1 m/s, orders of magnitude faster than the predicted terminal fall velocities of the smallest particles. The resulting deposit within 10 km of the vent consisted of a coarse, unimodal size population of coarse ash; beyond 10 km, fallout was poorly sorted with bimodal particle size related to aggregate fallout (Bonadonna et al., 2011). The authors develop a gravitationally driven convective settling law for fine ash in volcanic clouds based on laboratory experiments using glass microspheres in a density-stratified aqueous isothermal solution (building on work by Carazzo and Jellinek, 2013; Hoyal et al., 1999). Instability propagation reduced fine-ash particle lifetime by 1–3 orders of magnitude relative to single particle settling. This research offers the basis for simple, effective parameterization in operational modeling .

The role of hydrometeors in aggregation and formation of instabilities should also be considered. Volcanic clouds commonly contain more water than the background atmosphere (e.g., Williams and McNutt, 2004) and ash particles act as nucleation sites for water phases (e.g., Durant et al., 2008). Understanding of mammatus clouds provides a meteorological analog to gain insight into the mechanisms driving ash cloud instabilities (e.g., Schultz et al., 2006). Such features typically occur on the underside of thunderstorm anvils, and there have been numerous sightings on recent volcanic clouds. Within volcanic mammatus lobes, ash-hydrometeors accumulate at the base of the cloud layer and evaporate or sublimate (resulting in cooling of the immediate atmosphere due to the associated latent heat exchange). Gravitational loading, combined with the associated increase in air density, triggers the onset and propagation of instabilities leading to bulk sedimentation (Durant et al., 2009). A more contentious issue concerns the factors leading to the onset of cloud base instabilities. Specifically, does the process of aggregation drive the formation of convective instabilities, or does the development of instability lead to an increase in particle aggregation (through turbulent particle interactions)? A complete model for fine-ash settling and removal will, therefore, need to consider hydrometeor formation and water phase changes within the frame of particle aggregation and bulk cloud instabilities.

Thanks to Costanza Bonadonna, Larry Mastin, and Ellen Thomas for comments.