The details of volcanic plume source conditions or internal structure cannot readily be revealed by simple visual images or other existing remote imaging techniques. For example, one predominant observable quantity, the spreading rate in steady or quasi-steady volcanic plumes, is independent of source buoyancy flux. However, observable morphological features of short-duration unsteady plumes appear to be strongly controlled by volcanic source conditions, as inferred from our recent work. Here we present a new technique for using simple morphological evolution to extract the temporal evolution of source conditions of short-lived unsteady eruptions. In particular, using examples from Stromboli (Italy) and Santiaguito (Guatemala) volcanoes, we illustrate simple morphologic indicators of (1) increasing injection rate during the early phase of an eruption; (2) onset of source injection decline; and (3) the timing of source injection cessation. Combined, these observations indicate changes in eruption discharge rate and injection duration, and may assist in estimating total mass erupted for a given event. In addition, we show how morphology may provide clues about the vertical mass distribution in these plumes, which may be important for predicting ash dispersal patterns.


Three volcanic plume classes can be distinguished by the relationship between eruption duration and plume rise time (e.g., Sparks et al., 1997). Eruption duration is the time over which material is injected into the plume. Rise time is the time over which the plume reaches its maximum height. Plumes with short rise times from long-duration eruptions are classified as sustained or steady columns (class 1). Plumes with long rise times from nearly instantaneous explosions are classified as thermals (class 2). Class 3 plumes are characterized by short plume rise times that are comparable to eruption durations. Plumes in this class are short lived and highly unsteady. Each of the three classes exhibits a distinct relationship between the eruption (or injection) conditions and the plume rise dynamics that control plume morphology.

Class 1 plumes tend to have conical geometries (e.g., Wilson, 1976; Sparks et al., 1997), and class 2 plumes tend to have approximately spherical geometries (e.g., Wilson, 1976; Sparks et al., 1997). Both geometries are thought to indicate a self-similar dynamic state in the plume (i.e., the flow morphology does not change in time) that can be approximated well with analytical models of the turbulent dynamics (Morton et al., 1956). According to these models, the rise of sustained volcanic columns is primarily controlled by the rate at which buoyant fluid is discharged into the plume; on the other hand, thermals are controlled by the total amount of buoyant fluid discharged, not the discharge rate (e.g., Morton et al., 1956; Wilson, 1976; Sparks et al., 1997).

Class 3 volcanic plumes exhibit a variety of morphologies with spherical, conical, and cylindrical features (Patrick, 2007). These morphologic characteristics also evolve throughout the plume rise time (e.g., Patrick, 2007; Mori and Burton, 2009; Chojnicki et al., 2015), and they are poorly understood and lack analytical descriptions. Here we seek a method of inferring the dynamic plume conditions and the factors that control them using observations of their morphologic evolution.

During initial rise, volcanic plumes have been modeled as “starting plumes” (Turner, 1962; Wilson and Self, 1980; Sparks and Wilson, 1982; Patrick, 2007). Starting plumes have two prominent features, a spherical head and a conical tail, each with different dynamics. The spherical head contains a starting vortex structure, and the conical tail contains a steady jet structure (Turner, 1969). While this model appears to capture some aspects of class 3 volcanic plume morphology, such as the presence of spherical heads (Chojnicki et al., 2015), it cannot explain the cylindrical geometry of some plumes such as those identified by Patrick (2007).

These cylindrical features were not reproduced in analogue laboratory jets evolving from a steady rate of buoyant discharge, prompting Kitamura and Sumita (2011) to propose a correlation with discharge unsteadiness. Our previous laboratory work supports this claim, as we generated neutrally buoyant analogue jets with cylindrical geometries from an unsteady discharge rate (Chojnicki et al., 2014). We observed the evolution of both the analogue jet morphology and internal velocity structure and found morphology to be a good indicator of internal velocity structure and dynamics. Furthermore, and most importantly, we found that changes in the analogue jet morphology correlated well with changes in the discharge rate. We therefore apply our laboratory analysis to observations of volcanic plumes to show how morphology may be used as an indicator of discharge conditions for class 3 plumes. Because ground-based observations of volcanic plumes from short eruptions (tens to hundreds of seconds) are becoming increasingly common (Patrick, 2007; Mori and Burton, 2009; Lopez et al., 2013; Valade et al., 2014; Webb et al., 2014), we anticipate that our approach will be applicable in a wide range of field settings.


The analogue jet experiments reported by Chojnicki et al. (2014, 2015) were performed under idealized laboratory conditions that allowed the jet and source conditions to be measured simultaneously. Turbulent laboratory jets were generated by injecting water at high speeds into a tank of still water through a circular vent. Injection durations were comparable to, but shorter than, rise times, consistent with conditions for class 3 eruption plumes (Clarke et al., 2009). The experimental injection rates varied in a Gaussian-like temporal evolution, consistent with conditions for short-lived eruptions (Clarke et al., 2002), with a total duration of ∼0.40 s. We measured the resultant jet morphology as well as the structure of the internal flow field using particle image velocimetry (PIV) (Chojnicki et al., 2014, following Adrian, 1984). Analogue jets were seeded with silver-coated particles and illuminated by a laser light sheet, producing the unprocessed PIV images presented in Chojnicki et al. (2014). The tank water appears black relative to the seeded jet water, creating a visualization similar to a dyed jet (Figs. 1A, 2A, and 3A). These images are the best way to simultaneously visualize jet morphology and velocity structure.

The analogue jets created by the Gaussian-form injection rates evolved in three phases (Chojnicki et al., 2014). Phase 1 is characterized by increasing injection rate. The resultant jet forms two main regions: a spherical head and a roughly cylindrical tail (Fig. 1A). The spherical head consists of a starting vortex structure (labeled V1), common in jets entering still ambient fluid (e.g., Kieffer and Sturtevant, 1984). The head has a larger diameter than the tail. The tail has two sections, labeled 2 and 3, with diameters similar to the starting vortex (section 2) and the vent (section 3), respectively.

Phase 2 is characterized by decreasing injection rate. The arrival of phase 2 is indicated by the appearance of a narrow “neck” region between the starting vortex and section 2 of the cylindrical tail (Fig. 2A) that develops as the starting vortex pulls away from the more slowly injected tail. While section 2 slows as it moves away from source, section 3 continues at the same velocity due to inertia; the two sections thus start to combine into a conical form in phase 2.

After injection ends, phase 3 begins (Fig. 3A). The arrival of this phase is indicated by another change in jet shape. Section 2 has formed a vortex and has become the wide head of the jet, while sections 3 and 4 form the jet tail, which is now cylindrical (section 3) to conical (section 4) in shape. The tail is disconnected from the vent, as evidenced by the presence of ambient fluid between the vent and jet tail. The internal velocity structure of the jet reorganizes during this phase from the elongated pattern characterizing the cylindrical geometry to the compact radial pattern characterizing a spherical geometry (not shown; Chojnicki et al., 2014). The original starting vortex V1 has moved completely independently of the tail, and left the field of view.


Although the internal velocity fields from Chojnicki et al. (2014) cannot be readily compared with observations of opaque volcanic plumes, we can compare the flow morphologies. One study in the literature was suitable for this type of analysis because it provided information about plume evolution and estimates for ash and gas contents. The study observed a hornito event at Stromboli (Italy) with an ultraviolet camera (Mori and Burton, 2009). We modified the plume observations (Figs. 1B, 2B, and 3B) for our analysis and used a gray scale, where light gray represents high concentrations of sulfur dioxide and/or volcanic ash, dark gray represents lower concentrations, indicative of mixing of the plume fluid and ambient air, and black represents pure ambient air with no sulfur dioxide or ash.

Mori and Burton (2009) classified this event as type 2 in the scheme of Patrick (2007), an ashy plume that decelerates from an initially high velocity and then rises at a constant rate. Momentum is assumed to be the primary dynamic driver when the front is decelerating (Patrick, 2007). Thus, we assume that buoyancy does not play a dominant role in the near-vent dynamics. Mori and Burton (2009) noted the presence of an ambient wind during this event, but given that the plume axis is nearly vertical, we argue that wind had at most a secondary effect in the initial rise process. We therefore assume that our experimental results, with a neutrally buoyant jet rising into a still ambient, are analogous, at least to a first approximation.

For the first 8 s of the Stromboli eruption (Fig. 1B), a round head and a cylindrical tail characterize the plume morphology. The cylindrical tail can be subdivided into two sections with diameters similar to (section 2) and smaller than (section 3) the head. This morphology is present in the Stromboli plume despite its greater complexity relative to the analogue jets. This morphology corresponds with phase 1 in analogue jet evolution (Fig. 1A), the source acceleration stage in Chojnicki et al. (2014). Thus, we infer an increase in vent discharge for the first 8 s of this event. Our inference is consistent with interpretations of increasing gas flux made by Mori and Burton (2009). We also observe during phase 1 of the laboratory experiments that section 3 of the cylindrical tail has a diameter similar to that of the vent. We thus infer that the diameter of the volcanic plume tail (section 3) of ∼6 m is similar to the diameter of the volcanic vent. This inference is reasonably consistent with independent evidence that the hornito vents are ∼2–5 m wide (Chouet et al., 1974; Vergniolle and Brandeis, 1996; Del Bello et al., 2012).

We are uncertain whether the cylindrical shape of the jets during phase 1 is uniquely indicative of an increasing discharge rate. However, we do assert that the cylindrical shape may be a good indicator of source unsteadiness, given that the cylindrical geometry appears in our analogue jets when the injection is Gaussian in time, but not when the source is steady as in Kitamura and Sumita (2011). Given this ambiguity, future work should examine jet morphologies from a wide range of temporally varying discharge histories to determine if the cylindrical shape is unique to the discharge condition inferred here.

A snapshot of the Stromboli plume at 12 s is shown in Figure 2B. At this time, a narrow neck of fluid begins to form between the head and tail. We infer this narrow region to be similar to the neck that appears in the analogue jets during phase 2 (Fig. 2A). This neck first appears when the discharge rate begins to decrease in the laboratory jets (start of the falling edge of the Gaussian injection; Chojnicki et al., 2014). We therefore infer a decrease in discharge rate at this point in the volcanic eruption. Mori and Burton (2009) were unable to infer this instantaneous change in discharge rate, as their data represent the cumulative, not instantaneous, mass emitted. In contrast, the change in plume shape is an instant indicator of discharge rate change, according to our laboratory data.

Snapshots of the plume from 14 s to 20 s are shown in Figure 3B. Although the shape of the plume near the vent is difficult to see, we note a rounded feature near the base, designated as the end vortex (labeled EV), below which the plume narrows and nearly pinches out. Development of an EV is also observed in “stopping jets” that are buoyant (Kattimeri and Scase, 2014). Although the EV is more difficult to observe in the analogue jets, there is a gap between the base of the tail and the vent during phase 3. In the volcanic plume, this gap appears between the EV and vent (Fig. 3B), and, thus, we interpret the EV to mark the end of injection. The EV first appears at 12 s (Fig. 2B) in the volcanic eruption, suggesting a decrease and end to the discharge rate around this time. Based on our work, we note that the EV and a clear gap may only appear in situations like our laboratory experiments wherein the injection ends rapidly, a condition that may characterize some but not all class 3 volcanic eruptions.

These combined observations indicate that the volcanic plume entered phase 3 by ∼14 s after onset. In this phase of the analogue jets (Fig. 3A), the starting vortex evolves independently of the tail. Evidence for differential motion in the volcanic plume is found in the concentration difference between section 2 and V1 (Fig. 3B); the concentration decreases in V1 but remains similar over time in section 2. V1 appears to be moving or “stretching” away from section 2. Furthermore, the concentration appears to decrease in sections 3 and 4, but at a slower rate than in V1. Analogous to the laboratory jets, the dynamics in different parts of the volcanic plume evolve somewhat independently in the later stages. The spatial variations in plume evolution, and corresponding variations in plume dilution, are important considerations when modeling the dynamics of these plumes and resultant ash dispersal.

Phase 3 is marked by the evolution of section 2 in the analogue jets, which seems to hold for the volcanic plume as well (Fig. 3). In Figure 3, section 2 appears to contain the highest fluid concentrations in both the laboratory (Fig. 3A) and Stromboli (Fig. 3B) cases. This pattern has important implications for the transport of mass by the volcanic plume as it dissipates. In the laboratory case, and possibly also in the Stromboli case, section 2 appears to contain most of the mass within the plume, making the plume height an unreliable indicator of the largest concentration of ash released to the atmosphere for subsequent downwind transport. Although the difference in position between the second structure and flow front is small in this hornito event, it could be larger and more significant in larger class 3 events or as the plume evolves in time and ascends.

In addition to the hornito event at Stromboli, we document the evolution of a class 3 plume at Santiaguito volcano in Guatemala (Fig. 4). Although the magma composition and thus exact eruption mechanisms are thought to vary between Santiaguito and Stromboli, both generate class 3 plumes. The images in Figure 4 were collected on 8 January 2012 using a PlotWatcher Pro time-lapse camera sampling at 1 frame/s from the summit of Santa María volcano. The plume front first appeared at 07:33:56 h, and we consider this to be the eruption start time.

We observe similar morphological evolutions in the class 3 plumes at Santiaguito and Stromboli. The round front and tail with a variable diameter supply evidence for several features documented in the analogue jets: the starting vortex as well as sections 2, 3, and 4 (as labeled in Fig. 4). Unlike the Stromboli plume and analogue jets, however, no neck forms in the Santiaguito plume. The neck may not form for a variety of reasons, including the absence of a rapid decrease in discharge during this event, or perhaps the strong influence of buoyancy forces in the plume that cause section 2 to accelerate into V1, preventing the separation of V1 from section 2. Another characteristic unique to the Santiaguito plume is the cylindrical rather than conical geometry of section 4. The visible boundary of section 4 is less clearly defined as compared with regions V1 through 3, possibly suggesting different dynamics at work in that region.


Under unsteady discharge conditions, analogue jets (Chojnicki et al., 2014) and class 3 volcanic plumes evolve as a sequence of distinct flow segments. The segments grow in height and width, at various rates, as they rise. For analogue jets, changes in morphology correlate with changes in discharge rate, motivating the use of volcanic plume morphological changes to infer eruption discharge rate changes and estimate injection durations (and eventually mass eruption rate or total mass erupted) (Chojnicki et al., 2015). Inferring source conditions from relatively easy- and safe-to-collect plume observations may also reduce ambiguity in interpreting geophysical observations of eruption activity when geophysical data are available, and provide a means of monitoring the evolution of eruption source conditions when geophysical data are not available or not available in real time. Similar patterns of morphological evolution are observed for class 3 volcanic plumes from Santiaguito and Stromboli volcanoes during inertial-dominated rise. We therefore suggest that class 3 plumes with transient discharge conditions will generally evolve in a similar way while dominated by inertial momentum. Understanding of the effects of buoyancy and complex discharge histories on morphological evolution is needed to improve these inferences for a greater range of plumes generated by short-duration eruptions.

This work was supported by the National Science Foundation under grants EAR0810258 and EAR0930703, and by the Fulton Endowment. Contact K. Chojnicki or A. Clarke to request data. Chojnicki acknowledges Jeff Johnson and Ben Andrews for assistance in collecting the Santiaguito plume images. We thank Larry Mastin and two anonymous reviewers for comments that improved the manuscript.