*Manuscript Click here to view linked References 3-D numerical simulations of eruption column collapse: Effects of vent size on pressure-balanced jet/plumes Y.J. Suzukia,∗, T. Koyaguchia a Earthquake Research Institute, University of Tokyo, Yayoi 1-1-1, Bunkyo-ku, Tokyo 113-0032, Japan Abstract Buoyant columns or pyroclastic flows form during explosive volcanic eruptions. In the transition-state, these two eruption styles can develop simultaneously. We investigated the critical condition that separates the two eruption styles (referred to as “the column collapse condition”) by performing a series of three-dimensional numerical simulations. In the simulation results, we identify two types of flow regime: a turbulent jet that efficiently entrains ambient air (jet-type) and a fountain with a high-concentration of the ejected material (fountain-type). Hence, there are two types of column collapse (jet-type and fountain-type). Which type of collapse occurs at the column collapse condition depends on whether the critical mass discharge rate for column collapse (MDRCC ) is larger or smaller than that for the generation of a fountain (MDRJF ) for a given exit velocity. Temperature controls the relative magnitude of MDRCC relative to MDRJF , and hence the type of collapse. For given magma properties (e.g., temperature and water content), Corresponding author Email addresses: [email protected] (Y.J. Suzuki), [email protected] (T. Koyaguchi) ∗ Preprint submitted to Journal of Volcanology and Geothermal Research January 25, 2012 the column collapse condition is expressed by a critical value of the Richardson number (RiCC ≡ g0′ L0 /w02 , where g0′ is the source buoyancy, L0 is the vent radius, and w0 is the exit velocity). When the jet-type collapse occurs at the column collapse condition, RiCC is independent of the exit velocity. When the fountain type collapse occurs at the column collapse condition, on the other hand, RiCC decreases as the exit velocity increases. As the exit velocity exceeds the sound velocity, a robust flow structure with a series of standing shock waves develops in the fountain, which suppresses entrainment of ambient air and enhances column collapse. Keywords: eruption cloud, column collapse condition, numerical simulation, flow-regime map, turbulent mixing 1 1. Introduction 2 During explosive volcanic eruptions, a mixture of hot ash (pyroclasts) 3 and volcanic gas is released from the vent into the atmosphere. The mixture 4 generally has an initial density several times larger than the atmospheric 5 density at the vent and its ascent is driven only by its momentum. As the 6 ejected material entrains ambient air, the density of the mixture decreases 7 because the entrained air expands by heating from the hot pyroclasts. If 8 the density of the mixture becomes less than the atmospheric density before 9 the eruption cloud loses its upward momentum, a buoyant plume rises to 10 form a plinian eruption column. On the other hand, if the mixture loses its 11 upward momentum before it becomes buoyant, the eruption column collapses 12 to generate a pyroclastic flow. We refer to the condition that separates these 13 two eruption regimes as “the column collapse condition” (e.g., Sparks and 2 14 Wilson, 1976). Because the impact and type of volcanic hazards are largely 15 different between the two eruption regimes, predicting the column collapse 16 condition is of critical importance in physical volcanology. 17 The column collapse condition has previously been predicted by one di- 18 mensional (1-D) steady eruption column models (e.g., Wilson et al., 1980; 19 Bursik and Woods, 1991; Kaminski and Jaupart, 2001; Carazzo et al., 2008; 20 Koyaguchi et al., 2010). First generation 1-D steady models are based on 21 a classical theory of turbulent jet/plumes from a point source, in which the 22 efficiency of turbulent mixing between the eruption cloud and ambient air 23 is expressed by a single parameter, the “entrainment coefficient” (e.g., Mor- 24 ton et al., 1956; Taylor, 1945). These models captured the basic physics 25 of eruption column dynamics and enabled us to calculate the upward mo- 26 mentum and buoyancy of eruption columns as a function of the downstream 27 distance from the vent, and thus, to predict the maximum heights of the 28 eruption column and column collapse condition for given magma properties 29 (e.g., temperature and water content) and source conditions (e.g., vent ra- 30 dius and velocity). The models of column collapse show a quasi-quantitative 31 agreement with field observations of witnessed eruptions (e.g., Carey et al., 32 1990). Recently, the agreement between the model predictions and field ob- 33 servations has been improved by a more sophisticated 1-D model in which 34 the variation of entrainment coefficient with height is taken into account 35 (Carazzo et al., 2008). 36 Since the 1980s, 2-D and 3-D models have been developed for eruption 37 column dynamics. They have shown a number of complex fluid dynamical 38 features that cannot be described by 1-D models (e.g., Wohletz et al., 1984; 3 39 Valentine and Wohletz, 1989; Neri and Dobran, 1994; Oberhuber et al., 1998; 40 Suzuki et al., 2005; Ogden et al., 2008a,b; Suzuki and Koyaguchi, 2009). Gen- 41 erally, the flow of a gas-pyroclast mixture near the vent is characterized by an 42 annular structure consisting of an outer shear region and dense inner core (re- 43 ferred to as a “potential core” in aerodynamics literature (e.g., Rajaratnam, 44 1976)). The presence of the dense inner core affects the flow pattern dur- 45 ing column collapse (e.g., Suzuki et al., 2005). When the mixture is ejected 46 from a large vent (more than a few hundred meters), the outer shear region 47 cannot reach the central axis before the initial momentum is exhausted, so 48 that the dense inner core generates a fountain structure, called “a radially 49 suspended flow” (Neri and Dobran, 1994). On the other hand, for eruption 50 from a relatively narrow vent, the dense inner core disperses due to turbulent 51 mixing and the eruption cloud collapses without a fountain structure. Ogden 52 et al. (2008a) pointed out that the overpressure at the vent generates another 53 type of annular structure in the flow near the vent, which causes complex 54 flow behavior during column collapse; an oscillatory collapse with a regular 55 periodicity occurs even when a steady condition is applied at the source. 56 In this paper and some accompanying papers, we systematically inves- 57 tigate how these 2-D and 3-D effects modify the column collapse condition 58 on the basis of a number of 3-D numerical simulations of an eruption cloud 59 with a high spatial resolution. We carried out an extensive parameter study 60 to make regime maps of different flow patterns and to determine the column 61 collapse condition for each flow pattern. From the features of the column 62 collapse conditions, as well as those of the boundaries between the different 63 flow patterns, we can infer the governing factors which control the dynamics 4 64 of column collapse. In this particular paper, we focus on the flow whose 65 pressure balances ambient atmospheric pressure (referred to as “a pressure- 66 balanced jet/plume”). We show that even for this simplest case, the 2-D 67 and 3-D effects of flow (e.g., the presence of a potential core with or without 68 shock waves) substantially modify the column collapse condition. 69 2. Model Description 70 Our main concern is to clarify how the variation in the vent radius in- 71 fluences the global features of eruption clouds such as the formation of an 72 eruption column and/or pyroclastic flow. The most essential physics which 73 governs the dynamics of an eruption cloud is the efficiency of turbulent mix- 74 ing between the ejected material and ambient air. In order to minimize 75 complexity and to focus on the problem of turbulent mixing, a pseudo-gas 76 model is applied in this study; we ignore the separation of solid pyroclasts 77 from the eruption cloud and treat the eruption cloud as a single gas whose 78 density is calculated from the mixing ratio of the ejected material and en- 79 trained air. The fluid dynamics model solves a set of partial differential 80 equations describing the conservation of mass, momentum, and energy, and 81 constitutive equations describing the thermodynamic state of the mixture of 82 pyroclasts, volcanic gas, and air. These equations are solved numerically by 83 a general scheme for compressible flow (Roe, 1981; van Leer, 1977). Details 84 of the numerical procedures used in this study are described in Suzuki et al. 85 (2005). 86 The simulations are designed to describe the injection of a mixture of 87 pyroclasts and volcanic gas from a circular vent above a flat surface of the 5 88 earth in a stationary atmosphere with a temperature gradient typical of the 89 mid-latitude atmosphere (Table 1). The vent is located in the center of the 90 ground surface. The physical domain extends horizontally and vertically 91 from a few kilometers to several tens of kilometers. At the ground boundary, 92 the free-slip condition is assumed for the velocities of the ejected material 93 and air. At the upper and other boundaries of the computational domain, 94 the fluxes of mass, momentum, and energy are assumed to be continuous, 95 and these boundary conditions correspond to the free outflow and inflow of 96 these quantities. 97 In order to correctly reproduce the general feature of turbulent mixing 98 that the efficiency of entrainment is independent of the Reynolds number 99 (Dimotakis, 2000), it is essential to apply 3-D coordinates with a sufficiently 100 high spatial resolution (Suzuki et al., 2005). In this study, the calculations 101 were performed on a 3-D domain with a non-uniform grid. The grid size 102 was set to be sufficiently smaller than L0 /8 near the vent, where L0 is the 103 vent radius. In order to effectively simulate the turbulent mixing with high 104 spatial resolutions both far from and close to the vent, the grid size increases 105 at a constant rate (by a factor of 1.01 for the vertical coordinate and 1.02 106 for the horizontal coordinate) up to L0 with the distance from the vent, such 107 that the grid size is small enough to resolve the turbulent flow far from the 108 vent (cf. Suzuki and Koyaguchi, 2010). 109 We assume steady conditions; for each run, magma temperature T0 , water 110 content ng0 , vent radius L0 , and exit velocity w0 are fixed. The mass discharge 111 rate ṁ0 is determined from the relationship of ṁ0 = πρ0 w0 L20 , where ρ0 is 112 a function of T0 and ng0 . In this paper, for simplicity we assumed that the 6 113 pressure at the vent is equal to the atmospheric pressure. In general, the 114 pressure of a gas-pyroclast mixture deviates from the atmospheric pressure 115 during explosive eruptions; the mixture is accelerated and/or decelerated 116 because of decompression and/or compression within a short distance from 117 the vent (5–20 vent radius) (Woods and Bower, 1995; Ogden et al., 2008b; 118 Koyaguchi et al., 2010). Another extensive parameter study focusing on this 119 effect is in progress and the results will be reported elsewhere. 120 In order to classify different flow patterns and determine the column col- 121 lapse condition in pressure-balanced jet/plumes, we performed a parameter 122 study involving 95 numerical simulations. The conditions for these simula- 123 tions were divided into four groups according to magma temperature (T0 ) and 124 water content (ng0 ); group H with a high magma temperature (T0 = 1000 125 K and ng0 = 2.84 wt %), group I with an intermediate magma temperature 126 (T0 = 800 K and ng0 = 2.84 wt %), group L with a low magma temperature 127 (T0 = 550 K and ng0 = 2.84 wt %), and group N with a small water content 128 (T0 = 1000 K and ng0 = 1.23 wt %). These magma properties are con- 129 sidered to represent those of typical explosive eruptions (see Carazzo et al., 130 2008; Koyaguchi et al., 2010). Low temperature mixture of group L can be 131 ejected during eruptions involving the interaction of ground water and/or 132 country rock with magma (e.g., Koyaguchi and Woods, 1996). In order to 133 cover the conditions of typical explosive eruptions, we set the mass discharge 134 rate to be 104 –109 kg s−1 and the exit velocity to be 0.5c0 –3c0 , where c0 135 136 137 is the sound velocity of the gas-pyroclast mixture at the vent, defined as p c0 = 0.01ng0 Rg T0 . Generally, exit velocity, magma properties (i.e., T0 and ng0 ), vent radius, and hence, magma discharge rate are coupled such that 7 138 they are consistent with the fluid dynamics in the conduit and crater (Wil- 139 son et al., 1980; Koyaguchi et al., 2010). In this study, we assessed whether 140 or not the above ranges are physically realistic on the basis of Koyaguchi 141 et al. (2010). The wide range of exit velocity allows us to investigate fea- 142 tures of subsonic, sonic, and supersonic flows. A summary of the simulation 143 parameters for the four groups is listed in Table 2. 144 3. Results 145 3.1. Classification of Flow Patterns 146 The flow patterns in the simulation results are classified into four flow 147 regimes: an eruption column with a jet structure (e.g., run 99), an eruption 148 column with a fountain structure (e.g., run 33), jet collapse (e.g., run 78), 149 and fountain collapse (e.g., run 49). The representative features of each flow 150 regime are described below. 151 3.1.1. Eruption Column Regimes 152 In run 99 (T0 = 1000 K, ṁ0 = 1.0 × 107 kg s−1 , L0 = 49 m), a stable 153 eruption column with a typical jet structure develops (Fig.1). In this run, 154 the mixture of volcanic gas and pyroclasts is ejected from the vent as a 155 dense, high speed jet. After traveling a short distance from the vent, the 156 shear flow at the boundary between the jet and the ambient air becomes 157 unstable. The jet entrains ambient air by this shear instability; it forms an 158 annular mixing layer which surrounds an unmixed core flow (Fig.1a). The 159 unmixed core is eroded by the annular mixing layer and disappears at a 160 certain level (at a height of ∼ 3 km (60 times L0 ) in run 99). As the eruption 8 161 column further ascends, the flow becomes highly unstable and undergoes a 162 meandering instability that induces more efficient mixing (e.g., Suzuki et al., 163 2005). The stream-wise evolution of the flow results in a complex density 164 distribution in the eruption cloud (Fig.1b). The unmixed core is denser than 165 the ambient air, whereas the annular mixing layer is lighter than the ambient 166 air and gains buoyancy owing to expansion of the entrained air. After the 167 meandering instability develops, the dense unmixed core disappears before 168 it exhausts its initial momentum. We refer to the eruption column that is 169 characterized by the disappearance of the unmixed core before it exhausts 170 its initial momentum as “the jet-type column”. 171 When the gas-pyroclast mixture is ejected from a larger vent (run 33; 172 L0 = 154 m), a stable eruption column with a fountain structure develops 173 (Fig.2). In this run, because the vent radius is relatively large, the erosion 174 by the annular mixing layer does not reach the central axis at the height 175 where the unmixed core exhausts its initial momentum (Fig.2a). The dense 176 unmixed core around the central axis spreads radially at this height (Fig.2b). 177 Such a radially suspended flow is commonly observed in the fountain which 178 results from the injection of a dense fluid upwards into a less dense fluid (e.g., 179 Neri and Dobran, 1994; Lin and Armfield, 2000). The large-scale eddy of this 180 radially suspended flow causes an intensive mixing between the ejected mate- 181 rial and ambient air. After this mixing, the resultant mixture becomes buoy- 182 ant and generates an upward flow from the large-scale eddy; this produces 183 another type of stable column. The eruption column that is characterized 184 by the radially suspended flow is referred to as “the fountain-type column”. 185 Fig.3 shows the time evolutions of the vertical profiles along the central 9 186 axis for the jet-type column (run 99) and fountain-type column (run 33). In 187 the jet-type column, the mass fraction of the ejected material, ξ, gradually 188 decreases with height (Fig.3a). The high-ξ region intermittently rises at a 189 time interval of ∼ 10 s; the flow of the unmixed core is highly unsteady. In the 190 fountain-type column, on the other hand, ξ remains at 1.0 below a height of 191 2.2 km (Fig.3b) and decreases rapidly at this level, which corresponds to that 192 of the large-scale eddy at the top of the fountain. The height of the rapid 193 change of ξ is constant with time and does not show oscillatory behavior, 194 which suggests that the flow of the unmixed core is almost steady. 195 In Fig.4, the time-averaged mass fraction of the ejected material, ξ, and 196 normalized upward velocity, w/w0, along the central axis are presented as 197 functions of the normalized vertical position, z ∗ (≡ z/(w02 /2g)) for the jet- 198 type column (run 99) and fountain-type column (run 33); normalized quan- 199 tities are used for quantitative comparison between results of different vent 200 conditions in these diagrams. In the jet-type column, ξ gradually decreases 201 with height from 1.0 to 0.8 below z ∗ = 1 and rapidly falls to 0.15 from z ∗ = 1 202 to 2 (solid curve in Fig.4a). The averaged upward velocity, w/w0, gradually 203 decreases from 1.0 to 0.5 below z ∗ = 1 and slightly increases above it (solid 204 curve in Fig.4b). The height of z ∗ = 1 coincides with the level where the 205 unmixed core disappears (z = 3 km). These profiles of ξ and w/w0 are ex- 206 plained by the gradual growth of the annular mixing layer and the erosion of 207 the unmixed core below z ∗ = 1. In the fountain-type column, on the other 208 hand, ξ remains 1.0 below z ∗ = 0.7, rapidly decreases to 0.1 at the fountain 209 top (z ∗ = 0.7), and remains 0.1 above it (dashed curve in Fig.4a). The aver- 210 aged upward velocity, w/w0, decreases from 1.0 to −0.1 in the fountain below 10 211 z ∗ = 0.7, and rapidly increases above it (dashed curve in Fig.4b). These pro- 212 files of ξ and w/w0 reflect the features of entrainment in the fountain-type 213 column: extensive entrainment occurs at the fountain top, whereas the en- 214 trainment is inefficient in the fountain. It is noted that w/w0 has a negative 215 value just above the fountain top (at z ∗ = 0.7–0.9). This negative value of 216 w/w0 is caused by vortical motion associated with the large-scale eddy at the 217 fountain top. The extensive entrainment by the large-scale eddy at the foun- 218 tain top provides new insight into the mixing process in eruption columns, 219 because the mixing at the vertical edge of column has been assumed in the 220 1-D models. 221 In the simulation results of the fountain-type column regimes, we ob- 222 served that the qualitative feature of pressure distribution inside the foun- 223 tain changes with the exit velocity (Fig.5). When the exit velocity is sonic 224 (the Mach number is 1.0 in run 34), the time-averaged pressure inside the 225 fountain is uniform and almost equal to the atmospheric pressure (Fig.5a). 226 When the exit velocity is supersonic (the Mach number is 1.5 in run 33), a 227 spatially periodic pattern of higher and lower pressure is observed in the foun- 228 tain (Fig.5b). This periodic pattern results from the generation of standing 229 shock waves in the fountain; like in an overexpanded jet (i.e., a jet with exit 230 pressure lower than the surrounding fluid), compression waves (i.e., oblique 231 shock waves) are produced in the supersonic jet, and these compression waves 232 reflect from the symmetry axis and generate expansion waves (Dobran et al., 233 1993). 11 234 3.1.2. Column Collapse Regimes 235 In the present simulations, column collapse occurs when the vent radius is 236 large. For given vent radius and exit velocity, column collapse tends to occur 237 as magma temperature decreases. The flow pattern of the column collapse 238 regimes depends on the magma temperature and vent radius. 239 When a low-temperature gas-pyroclast mixture is ejected from a narrow 240 vent (run 78; T0 = 550 K, L0 = 11m), the eruption column collapses from 241 the jet (Fig.6). In this run, the jet entrains ambient air by the shear instabil- 242 ity; the annular mixing layer develops and the unmixed core disappears at a 243 height of 0.5 km (Fig.6a). Because the initial temperature is low, the erup- 244 tion cloud remains heavier than air even though it entrains a large amount of 245 air. As a result, the low-concentration mixture generates a pyroclastic flow 246 (Fig.6b). The column collapse regime that is characterized by the disappear- 247 ance of the unmixed core before it exhausts its initial momentum is referred 248 to as “the jet-type collapse”. 249 In contrast to the jet-type collapse, when a high-temperature gas-pyroclast 250 mixture is ejected from an extremely large vent (run 49; T0 = 1000 K, 251 L0 = 403 m), the eruption column collapses from a fountain (Fig.7a). As was 252 observed in the case of the fountain-type column (Fig.2), when the vent ra- 253 dius is large, the unmixed core reaches a height where the initial momentum 254 is exhausted and generates a radially suspended flow which causes mixing by 255 a large-scale eddy. As the vent radius exceeds a certain value, the large-scale 256 eddy of the radially suspended flow cannot entrain sufficient air for the whole 257 mixture to become buoyant (Fig.7b). As a result, the unmixed part in the 258 radially suspended flow collapses and generates a pyroclastic flow. The col- 12 259 umn collapse regime that is characterized by the radially suspended flow is 260 referred to as “fountain-type collapse”. When the dense eruption cloud hits 261 the ground and spreads horizontally, it efficiently entrains ambient air. The 262 entrained air expands by heating from the pyroclasts so that the mixture 263 of the ejected material and ambient air forms a buoyant co-ignimbrite ash 264 cloud. 265 Fig.8 shows the time-averaged vertical profiles of physical quantities along 266 the central axis for the jet-type collapse (run 78) and fountain-type collapse 267 (run 49). In the jet-type collapse, ξ and w/w0 gradually decrease with height 268 and fall to zero at z ∗ ∼ 1 (solid curves in Fig.8). The gradual decreases in 269 ξ and w/w0 represent the dispersion of the unmixed core due to the gradual 270 growth of the annular mixing layer. In the fountain-type collapse (dashed 271 curve in Fig.8a), ξ remains 1.0 in the fountain (z ∗ < 1.2) and rapidly de- 272 creases to 0.1 at the fountain top (z ∗ = 1.2). The value for z ∗ > 1.2 (ξ = 0.1) 273 represents the mass fraction of the ejected material in the co-ignimbrite ash 274 cloud. The upward velocity decreases with height in the fountain and falls to 275 zero at the fountain top (dashed curve in Fig.8b). In contrast to the case of 276 the jet-type collapse, the unmixed core itself is the main source of pyroclastic 277 flows in the case of the fountain-type collapse. 278 3.1.3. Transitional Regimes 279 The boundaries of the above four flow regimes (the jet-type column, 280 fountain-type column, jet-type collapse, and fountain-type collapse regimes) 281 are determined by two types of transition. One is the transition between the 282 jet-type and fountain-type regimes; we call the condition for this transition 283 the jet-fountain (JF) condition. The other is the transition between the col13 284 umn convection and collapse regimes; we call the condition for this transition 285 the column collapse (CC) condition. 286 Around the JF condition, some intermediate features between jet-type 287 and fountain-type regimes were observed. Run 46 (T0 = 1000 K, L0 = 82 m) 288 shows an intermediate feature between the jet-type column and the fountain- 289 type column regimes (Fig.9). In this run, the unmixed core is eroded to a 290 considerable extent by the annular mixing layer, while relatively large-scale 291 eddies develop at the fountain top (z ∼ 4 km). In run 92 (T0 = 550 K, 292 L0 = 38 m), the type of column collapse changes from the jet-type to the 293 fountain-type with time (Fig.10). At the initial stage of eruption (t < 50 s), 294 the unmixed core disappears at z = 0.6 km (see the contour of ξ = 0.95), 295 whereas the total height of the negatively buoyant jet reaches 1.4 km high. 296 Subsequently, the height of the unmixed core increases with time (50 s < t < 297 80 s) and reaches the top of the jet so that the fountain structure (i.e., the 298 radially suspended flow) develops (t > 80 s). 299 Around the CC condition, some intermediate features between column 300 and collapse regimes are observed. In run 27 (T0 = 1000 K, L0 = 345 m), a 301 fountain-type column and pyroclastic flow develop simultaneously (Fig.11). 302 After mixing due to the large-scale eddy at the fountain top, most of the 303 cloud becomes a buoyant plume, whereas the side of the fountain partially 304 flows down to form a pyroclastic flow. This pyroclastic flow is less energetic 305 than the pyroclastic flow of the fountain-type collapse regime which falls 306 down directly from the fountain top (cf. Fig.7). Around the CC conditions 307 for groups L and I, a jet-type column becomes unstable with time and can 308 collapse to generate a dilute pyroclastic flow; for example, in run 120 (T0 = 14 309 800 K, L0 = 18 m), the transition from the jet-type column to the jet-type 310 collapse regimes occurs at t ∼ 150 s from the beginning of the eruption (the 311 result is not shown). 312 3.2. Flow-Regime Map 313 On the basis of extensive parameter studies, we made flow-regime maps 314 in the parameter space of the mass discharge rate (ṁ0 ) and exit velocity (w0 ) 315 for 550 K < T0 < 1000 K and 1.23 wt % < ng0 < 2.84 wt % (Fig.12). 316 When the temperature is low (group L) or intermediate (group I), the 317 flow changes from the jet-type column to the jet-type collapse regimes, and 318 from the jet-type collapse to the fountain-type collapse regimes as the mass 319 discharge rate increases for a given exit velocity (Figs.12a,b). The boundary 320 between the jet-type column and the jet-type collapse regimes represents the 321 CC condition (solid curves), and that between the jet-type collapse and the 322 fountain-type collapse regimes represents the JF condition (dashed curves). 323 Both the critical mass discharge rate for the CC condition (MDRCC ) and 324 for the JF condition (MDRJF ) increase as the exit velocity increases. For a 325 low temperature, MDRCC is substantially smaller than MDRJF ; as a result, 326 the jet-type collapse occurs in a wide range of parameters in this diagram 327 (see Fig.12a). As the temperature increases, the distance between these two 328 boundaries decreases so that the region of the jet-type collapse regime shrinks 329 (Fig.12b). 330 When the temperature is high (group H), MDRCC becomes greater than 331 MDRJF for a given exit velocity; as a result, the flow changes from the 332 jet-type column to the fountain-type column regimes, and from the fountain- 333 type column to the fountain-type collapse regimes as the mass discharge rate 15 334 increases (Figs.12c,d). In these cases, the CC condition is represented by the 335 boundary between the fountain-type column and the fountain-type collapse 336 regimes, and the JF condition is represented by the boundary between the 337 jet-type column and the fountain-type column regimes. The results of group 338 N (T0 = 1000 K and ng0 = 1.23 wt %) shows that the distance between 339 these two boundaries decreases as the water content decreases for a given 340 temperature (Fig.12d). 341 4. Physical Meanings of the Two Critical Conditions 342 The variation in flow-regime maps suggests that which flow regime can 343 occur is determined by the positional relationship between the JF and CC 344 conditions. This positional relationship depends on the magma properties; 345 for a given exit velocity, MDRCC is smaller than MDRJF when the tem- 346 perature is relatively low, whereas MDRCC is larger than MDRJF when the 347 temperature is high. In this section, we discuss how each critical condition 348 depends on the magma properties. 349 In order to understand the physical meanings of the two critical condi- 350 tions, we redraw Fig.12 in a non-dimensional space, namely in the space 351 of the Richardson number (Ri) and the Mach number (M) (Fig.13). The 352 Richardson number is defined by Ri = g0′ L0 , w02 (1) 353 where the source buoyancy, g0′ , is defined using the density of the ejected 354 material, ρ0 , and the atmospheric density, ρa , as g0′ = (ρ0 − ρa )g/ρ0 . The 355 Mach number is defined as the ratio of the exit velocity and the sound velocity 16 356 of the ejected material, c0 : M= w0 . c0 (2) 357 Fig.13 indicates that both the two critical conditions are determined 358 mainly by the Richardson number for given magma properties (cf. Valentine 359 and Wohletz, 1989). We define the Richardson number at the CC condi- 360 tion as RiCC and that at the JF condition as RiJF , respectively. For a given 361 Mach number, RiCC increases as T0 and/or ng0 increases (see gray stars in 362 Fig.13), whereas RiJF is almost independent of T0 and/or ng0 (see white 363 stars in Fig.13). Whether RiCC < RiJF or RiCC > RiJF determines the type 364 of the intermediate flow regime. In the low-temperature case (Figs.13a,b), 365 RiCC < RiJF so that the intermediate flow regime is the jet-type collapse. 366 In the high-temperature case (Figs.13c,d), RiCC > RiJF so that the inter- 367 mediate flow regime is the fountain-type column. The physical mechanisms 368 that cause the dependencies of RiJF and RiCC on magma properties will be 369 discussed below. 370 4.1. Physical mechanisms to control RiJF 371 The JF condition is determined by whether the unmixed core is eroded 372 by the annular mixing layer or not before the initial momentum is exhausted. 373 This problem has two length scales: the height where the unmixed core dis- 374 appears (Hcore ), and the height where the initial momentum is exhausted 375 (Hmmt ). When Hcore > Hmmt , the fountain is generated. According to ex- 376 perimental studies on high-speed jets (e.g., Nagamatsu et al., 1969), Hcore is 377 a function of the vent radius, L0 , and M: Hcore = (a1 M + a2 )L0 , 17 (3) 378 where a1 and a2 are positive constants. On the other hand, Hmmt is estimated 379 from the balance between the kinetic and potential energy: Hmmt = 380 w02 . 2g0′ These two equations yield 2(a1 M + a2 )g0′ L0 Hcore = 2(a1 M + a2 )Ri. = Hmmt w02 381 (4) (5) Consequently, we obtain the JF condition (i.e., Hcore = Hmmt ) as RiJF = 1 . 2(a1 M + a2 ) (6) 382 This equation successfully explains the simulation results in Fig.13 that RiJF 383 is independent of the magma properties for a fixed M, and that RiJF decreases 384 as M increases. 385 The above result that the transition between the jet and fountain occurs 386 at a certain critical Ri is supported by laboratory experiments of negatively 387 buoyant jet/fountains (e.g., Kaye and Hunt, 2006). In these experiments, 388 when Ri at the source is smaller than a certain value of O(1), the flow behaves 389 as a highly forced jet where the injected material efficiently mixes with the 390 ambient fluid (i.e., the jet-type). When Ri is larger than the critical value, 391 on the other hand, the flow becomes a fountain where there is little mixing 392 between the injected material and the ambient fluid (i.e., the fountain-type). 393 In order to extend the conclusion derived from these experimental results for 394 subsonic flows to the case of supersonic flows, we carried out supplementary 395 simulations (Appendix A). 18 396 4.2. Dependency of RiCC on the Magma Properties 397 The CC condition is determined by whether the eruption cloud becomes 398 buoyant or not at z = Hmmt . When the mass fraction of the ejected material, 399 ξ, is smaller than a certain value, ξcrt , the mixture of ejected material and en- 400 trained air is lighter than the ambient air. On the other hand, when ξ > ξcrt , 401 the density of the mixture is larger than the atmospheric density. Therefore, 402 the column collapse occurs when the mass fraction of the ejected material 403 at z = Hmmt , ξmmt, is larger than ξcrt . The value of ξmmt is approximately 404 estimated by ξmmt = ṁ0 , ṁ0 + ṁair (7) 405 where ṁ0 is the mass discharge rate at the vent, and ṁair is the mass of air 406 entrained below z = Hmmt per unit time. By definition, ṁ0 is represented by ṁ0 = πρ0 w0 L20 . (8) 407 Because ṁair is proportional to the surface area of the eruption column and 408 the exit velocity, w0 , it is expressed by ṁair = 2πkρa w0 L0 Hmmt (9) 409 under the assumption that the efficiency of turbulent mixing (i.e., the en- 410 trainment coefficient, k (Morton et al., 1956)) is constant. Substituting Eqs. 411 (8) and (9) into Eq.(7), we obtain ξmmt = πρ0 w0 L20 . πρ0 w0 L20 + wπkρa w0 L0 Hmmt (10) 412 From Eqs. (1), (4), and (10), we obtain the CC condition (i.e., ξmmt = ξcrt ) 413 as RiCC = k ρa ξcrt . ρ0 1 − ξcrt 19 (11) 414 This result (i.e., Eq.(11)) successfully explains the simulation result that 415 RiCC depends on magma properties even for a fixed M; because 1/ρ0 and ξcrt 416 increase with T0 and ng0 , RiCC increases with T0 and ng0 . This result is also 417 consistent with the predictions based on the 1-D model of eruption column 418 by Woods and Caulfield (1992) (Appendix B). 419 4.3. Dependency of RiCC on the Mach Number 420 The results of our 3-D simulations show that RiCC is independent of 421 the Mach number when RiCC is smaller than RiJF (Figs.13a,b). When 422 RiCC > RiJF , on the other hand, RiCC decreases as the Mach number in- 423 creases (Figs.13c,d). According to Eq.(11), RiCC is independent of the Mach 424 number under the assumption that the efficiency of turbulent mixing (i.e., 425 the entrainment coefficient) is constant. The variation of RiCC in Figs.13c,d, 426 therefore, means that the efficiency of turbulent mixing changes with the exit 427 velocity for RiCC > RiJF . 428 The efficiency of turbulent mixing for the results of 3-D simulations (the 429 effective entrainment coefficient) can be estimated by comparing the CC con- 430 dition based on the 3-D simulations with that predicted by the 1-D model (cf. 431 Suzuki et al., 2005). In Fig.12, the column collapse conditions predicted by 432 the 1-D model with different assumed entrainment coefficients are also illus- 433 trated by the dotted curves. For jet-type collapse (RiCC < RiJF ), the column 434 collapse condition based on the 3-D simulations coincides with that of the 1- 435 D model with a constant entrainment coefficient (∼ 0.03 in Figs.12a,b). For 436 fountain-type collapse (RiCC > RiJF ), the effective entrainment coefficient of 437 the 3-D simulations decreases from 0.10 to 0.03 with the increasing Mach 438 number (Figs.12c,d). 20 439 We attribute the systematic decrease in the efficiency of turbulent mixing 440 with the increasing exit velocity to the formation of shock wave structures 441 inside the fountain. When the gas-pyroclast mixture is ejected from the 442 vent at high speed, the standing shock waves are generated in the fountain 443 even if the pressure at the vent is balanced with the atmospheric pressure 444 (Fig.5). These shock wave structures suppress efficient mixing between the 445 ejected material and ambient air. Recent experimental results by Saffaraval 446 et al. (in review) also indicates that the efficiency of entrainment is reduced 447 when shock wave structures are formed in high-speed jets. As the Mach 448 number increases, the flow with a series of shock waves reaches a higher 449 level, which causes less efficient mixing. Thus, the region of the fountain- 450 type collapse regime expands with the increasing Mach number in the Ri–M 451 space (Figs.13c,d). 452 5. Geological Implications 453 The present 3-D numerical simulations of eruption clouds show two types 454 of column collapse (CC) conditions. The CC condition is characterized by 455 the jet-type collapse when the initial temperature of the ejected material is 456 relatively low, whereas the CC condition is characterized by the fountain-type 457 collapse when the initial temperature of the ejected material is high. 458 The flow patterns in the jet-type and fountain-type collapse regimes 459 are considered to affect depositional structures in pyroclastic flows. When 460 the jet-type collapse occurs, the eruption cloud contains a large amount of 461 air, generating a dilute pyroclastic flow. The column collapse during low- 462 temperature eruptions, therefore, tends to generate a dilute pyroclastic flow. 21 463 Generally, eruptions with low temperatures are attained by the interaction 464 between the magma and water (e.g., Koyaguchi and Woods, 1996). The 465 present numerical results are consistent with the fact that surge deposits 466 from dilute low-temperature pyroclastic flows are commonly observed in 467 the deposits of phreatomagmatic eruptions (e.g., Moore, 1967). When the 468 fountain-type collapse occurs, on the other hand, high-concentration and 469 high-temperature pyroclastic flows are generated. Such pyroclastic flows ac- 470 count for the occurrence of highly welded pyroclastic flow deposits (e.g., 471 Branney and Kokelaar, 1992). During the fountain-type collapse, a large 472 amount of pyroclasts fall to the ground directly from the fountain top, which 473 generates a chaotic flow around the vent as well as outward pyroclastic flows 474 (see Fig.6). The chaotic flow around the vent may explain complex features 475 of proximal facies in some pyroclastic flow deposits (e.g., Druitt and Sparks, 476 1982; Walker, 1985). 477 Another implication is concerned with the CC condition predicted by 478 the 1-D models. The efficiency of entrainment estimated in the present 3-D 479 numerical simulations (k ∼ 0.03 for the jet-type CC condition and k = 0.03– 480 0.10 for the fountain-type CC condition) is substantially smaller than the 481 value commonly assumed in the 1-D models (e.g., k = 0.09 in Woods (1988)). 482 This means that the eruption column collapse is likely to occur for smaller 483 mass discharge rates than the CC condition predicted by those 1-D mod- 484 els. Our results also suggest that the efficiency of entrainment decreases as 485 the exit velocity increases along the fountain-type CC condition (Figs.12c,d). 486 Small values of k have been proposed in some 1-D models (Carazzo et al., 487 2008, 2010; Koyaguchi et al., 2010) on the basis of recent laboratory exper- 22 488 iments (Kaminski et al., 2005) and direct measurements in numerical simu- 489 lations (Suzuki and Koyaguchi, 2010). The effect of the exit velocity on the 490 efficiency of entrainment (e.g., Saffaraval et al., in review), however, has not 491 been taken into account in any 1-D models. We suggest that this effect may 492 play an important role in column collapse, especially during high-temperature 493 eruptions. 494 On the basis of the 3-D numerical simulations, we have shown that the 3- 495 D structures of flow largely affect the column collapse condition. Our results 496 indicate that the differences between the jet-type and fountain-type collapses 497 are closely related with the unsteady vortical structures (e.g., Suzuki and 498 Koyaguchi, 2010). For better understanding of column collapse condition, 499 further analyses based on 3-D flow visualization are required. 500 Appendix A. Heights of Negatively Buoyant Jet and Fountain 501 In section 4.1, we presented a simple analysis for the JF condition on 502 the assumption that the fountain is generated when Hcore > Hmmt (Eq.(6)). 503 The analysis is based on previous experimental studies on the transition 504 between negatively buoyant jets and fountains for subsonic flows (e.g., Kaye 505 and Hunt, 2006). According to the experimental results, the JF condition of 506 negatively buoyant jet/fountains is closely related to the relationship between 507 the maximum height (Hmax ) and Ri. When Ri is sufficiently small so that 508 Hcore < Hmmt (see also Eq.(5)), Hmax of the negatively buoyant jet increases 509 with Ri for a fixed exit velocity. This is because the ratio of the mass of 510 entrained air to that of the ejected material decreases as Ri increases, and 511 hence, Hcore /Hmmt increases. When Ri is large so that Hcore > Hmmt , on the 23 512 other hand, Hmax of the negatively buoyant fountain remains Hmmt for a fixed 513 exit velocity and is independent of Ri. These results indicate that, at least 514 for subsonic flows under the experimental conditions, the JF condition can 515 be determined by the critical value of Ri at which the Hmax –Ri relationship 516 changes. 517 Here, we attempt to extend the above idea to the JF condition of super- 518 sonic flows; we carried out supplementary simulations for negatively buoyant 519 jet/fountains of a gas-pyroclast mixture for various Mach numbers. In order 520 to avoid unnecessary complications (e.g., rapid expansion of the cloud due 521 to heating of the entrained air) and to extract the effect of the Mach number 522 on the JF condition, the temperature of the gas-pyroclast mixture was set 523 to be equal to the ambient temperature (273 K). Under this condition, the 524 cloud density is always larger than the ambient density. 525 Fig.A.1 shows the Hmax –Ri relationship of the negatively buoyant jet/ 526 fountains for M= 1–3. For sonic flows (M= 1), negatively buoyant jets de- 527 velop and Hmax increases with Ri when Ri < 0.6. When Ri > 0.6, negatively 528 buoyant fountains with a constant Hmax develop. The results show that the 529 critical Richardson number for the JF condition (RiJF ) based on the change 530 in the Hmax –Ri relationship decreases with the increasing Mach number (ar- 531 rows in the figure). From this RiJF –M relationship, we obtain the values 532 of a1 and a2 in Eq.(6); the estimated values are a1 = 1.25 and a2 = 0.42, 533 which are consistent with those estimated from Fig.13 (a1 = 1.3–3.4 and 534 a2 = 0.16–4.7). 24 535 Appendix B. Derivation of a Formula for the Column Collapse 536 Conditions by 1-D Model on the Basis of Analogue 537 Experiments 538 Woods and Caulfield (1992) carried out a series of analogue experiments 539 simulating column collapse and proposed a simple formula for the CC condi- 540 tion. In their experiments, mixtures of methanol and ethylene glycol (MEG) 541 were injected as a downward propagating jet into fresh water. The MEG is 542 lighter than fresh water, whereas the MEG-water mixture can be denser than 543 fresh water. The density difference between the fresh water and MEG-water 544 mixture was approximated by a function of the volume fraction of MEG in 545 the mixture, W , as ∆ρ ≡ β−α W 2 =1− 1− , βm − α Wm (B.1) 546 where β is the MEG-water mixture density and α is the water density. The 547 MEG-water mixture has a maximum density of βm when W = Wm . If the 548 MEG entrains sufficient water before its initial momentum is exhausted, the 549 MEG-water mixture becomes denser than the ambient water and continues 550 to flow downwards as a negatively buoyant “plume”. If the MEG does not 551 mix with sufficient water, the mixture “collapses” and rises back toward the 552 top of the tank. Although the direction of the flow is upside-down, the 553 experiments reproduce fundamental features of eruption column collapse. 554 Woods and Caulfield (1992) showed that, when the ∆ρ–W relationship is 555 explicitly given as Eq.(B.1), a simple formula of the CC condition (i.e., the 556 region of collapse regime) for the MEG-water system can be derived on the 25 557 basis of the 1-D model with a constant entrainment coefficient (k) as: 5/2 2 4kαṀ0 W0 W0 , −1 ≥ Wm 2Wm 5(βm − α)g ṁ30 (B.2) 558 where W0 is the initial volume fraction of MEG, Ṁ0 (≡ w02 L20 ) is the initial 559 momentum flux, and ṁ0 (≡ w0 L20 ) is the initial mass flux. It should be noted 560 that this formula is applicable only to the MEG-water system, because it is 561 based on the fact that the density of the mixture is equal to the ambient 562 density when W = 2Wm (≡ Wcrt ) in Eq.(B.1). Here, we extend Eq.(B.2) to 563 the general case and show that it is consistent with our result (i.e., Eq.(11)). 564 In order to obtain the CC condition for the general case, Wm in Eq.(B.2) 565 must be given as a function of general quantities such as the critical vol- 566 ume fraction (Wcrt ) or the critical mass fraction (ξcrt ) of ejected materials. 567 Considering that Eq.(B.1) and Wcrt = 2Wm are the relationships for the 568 MEG-water system, Eq.(B.2) is rewritten using the general quantities as −Ri ≥ 8k Wcrt . 5 W0 − Wcrt (B.3) 569 Here, the definition of Ri is given by Eq.(1) (note that Ri is negative because 570 the direction of the flow is upside-down). Furthermore, because the mass 571 fraction is expressed by ξ= 572 W β0 , W β0 + (1 − W )α (B.4) where β0 is the initial density of MEG, Eq.(B.3) is rewritten as −Ri ≥ Ricrt ≡ 8k α ξcrt . 5 β0 1 − ξcrt (B.5) 573 This has the same form as our formula for the CC condition (i.e., Eq.(11)). 574 These two formulae have different proportionality constants (1.0 in Eq.(11) 26 575 and 8/5 in Eq.(B.5)). This difference is attributed to the fact that the mass 576 of entrained air is assumed to be constant at any heights in our analysis 577 (see Eq.(9)), whereas it is calculated on the basis of the 1-D model with a 578 constant k in Woods and Caulfield (1992). 579 Acknowledgments 580 The manuscript was improved by comments from L. Wilson, D. E. Ogden 581 and an anonymous reviewer. Computations were carried out in part on Earth 582 Simulator at the Japan Agency for Marine-Earth Science and Technology, 583 Altix 4700 at the Earthquake Research Institute, University of Tokyo, and 584 Primergy RX200S6 at Computer System for Research, Kyushu University. 585 Part of this study was supported by the ERI cooperative research program 586 and by KAKENHI (No. 21340123). 587 References 588 Branney, M.J., Kokelaar, P., 1992. A reappraisal of ignimbrite emplacement: 589 progressive aggradation and changes from particulate to non-particulate 590 flow during emplacement of high-grade ignimbrite. Bull. Volcanol. 54, 591 504–520. 592 593 Bursik, M.I., Woods, A.W., 1991. Buoyant, superbuoyant and collapsing eruption columns. J. Volcanol. Geotherm. Res. 45, 347–350. 594 Carazzo, G., Kaminski, E., Tait, S., 2008. On the dynamics of volcanic 595 columns: A comparison of field data with a new model of negatively buoy- 596 ant jets. J. Volcanol. Geotherm. Res. 178, 94–103. 27 597 Carazzo, G., Kaminski, E., Tait, S., 2010. The rise and fall of turbulent 598 fountains: a new model for improved quantitative predictions. J. Fluid. 599 Mech. 657, 265–284. 600 Carey, S., Sigurdsson, H., Gardner, J.E., Criswell, W., 1990. Variations in 601 column height and magma discharge during the May 18, 1980 eruption of 602 Mount St. Helens. J. Volcanol. Geotherm. Res. 43, 99–112. 603 604 605 606 607 608 Dimotakis, P.E., 2000. The mixing transition in turbulent flows. J. Fluid. Mech. 409, 69–98. Dobran, F., Neri, A., Macedonio, G., 1993. Numerical simulation of collapsing volcanic columns. J. Geophys. Res. 98, 4231–4259. Druitt, T.H., Sparks, R.S.J., 1982. A proximal ignimbrite breccia facies on Santorini, Greece. J. Volcanol. Geotherm. Res. 13, 147–171. 609 Kaminski, E., Jaupart, C., 2001. Marginal stability of atmospheric eruptoin 610 columns and pyroclastic flow generation. J. Geophys. Res. 106, 21785– 611 21798. 612 613 Kaminski, E., Tait, S., Carazzo, G., 2005. Turbulent entrainment in jets with arbitrary buoyancy. J. Fluid. Mech. 526, 361–376. 614 Kaye, N.B., Hunt, G.R., 2006. Weak fountains. J. Fluid. Mech. 558, 319–328. 615 Koyaguchi, T., Suzuki, Y.J., Kozono, T., 2010. Effects of the crater on 616 eruption column dynamics. J. Geophys. Res. 115, B07205. 28 617 Koyaguchi, T., Woods, A.W., 1996. On the formation of eruption columns 618 following explosive mixing of magma and surface-water. J. Geophys. Res. 619 101, 5561–5574. 620 van Leer, B., 1977. Towards the ultimate conservative difference scheme. III. 621 Upstream-centered finate-difference schemes for ideal compressible flow. J. 622 Comput. Phys. 23, 263–275. 623 624 625 626 Lin, W., Armfield, S.W., 2000. Direct simulation of weak axisymmetric fountains in a homogeneous fluid. J. Fluid. Mech. 403, 67–88. Moore, J.G., 1967. Base surge in recent volcanic eruptions. Bull. Volcanol. 30, 337–363. 627 Morton, B.R., Taylor, G.I., Turner, J.S., 1956. Turbulent gravitational con- 628 vection from maintained and instantaneous sources. Proc. R. Soc. Lond. 629 A 234, 1–23. 630 Nagamatsu, H.T., Sheer, R.E.J., , Horvay, G., 1969. Supersonic jet noise 631 theory and experiments. Basic Aerodynamics Noise Research, NASA SP- 632 207 , 17–51. 633 Neri, A., Dobran, F., 1994. Influence of eruption parameters on the ther- 634 mofluid dynamics of collapsing volcanic columns. J. Geophys. Res. 99, 635 11833–11857. 636 637 Oberhuber, J.M., Herzog, M., Graf, H.F., Schwanke, K., 1998. Volcanic plume simulation on large scales. J. Volcanol. Geotherm. Res. 87, 29–53. 29 638 Ogden, D., Glatzmaier, G.A., Wohletz, K.H., 2008a. Effects of vent over- 639 pressure on buoyant eruption columns: Implications for plume stability. 640 Earth Planet. Sci. Lett. 268, 283–292. 641 Ogden, D.E., Wohletz, K.H., Glatzmaier, G.A., Brodsky, E.E., 2008b. Nu- 642 merical simulations of volcanic jets: Importance of vent overpressure. J. 643 Geophys. Res. 113, B02204. 644 Rajaratnam, N., 1976. Turbulent jets. Elsevier. 645 Roe, P.L., 1981. Approximate Riemann solvers, parameter vectors, and dif- 646 ference schemes. J. Comput. Phys. 43, 357–372. 647 Saffaraval, F., Solovitz, S.A., Ogden, D.E., Mastin, L.G., in review. Impact 648 of reduced near-field entrainment of overpressure volcanic jets on plume 649 development. J. Geophys. Res. . 650 651 652 653 Sparks, R.S.J., Wilson, L., 1976. A model for the formation of ignimbrite by gravitational column collapse. J. Geol. Soc. Lond. 132, 441–451. Suzuki, Y.J., Koyaguchi, T., 2009. A three-dimensional numerical simulation of spreading umbrella clouds. J. Geophys. Res. 114, B03209. 654 Suzuki, Y.J., Koyaguchi, T., 2010. Numerical determination of the efficiency 655 of entrainment in volcanic eruption columns. Geophys. Res. Lett. 37, 656 L05302. 657 Suzuki, Y.J., Koyaguchi, T., Ogawa, M., Hachisu, I., 2005. A numerical 658 study of turbulent mixing in eruption clouds using a three-dimensional 659 fluid dynamics model. J. Geophys. Res. 110, B08201. 30 660 661 662 663 664 665 666 Taylor, G.I., 1945. Dynamics of a mass of hot gas rising in air. U. S. Atomic Energy Commission MDDC 919 , LADC276. Valentine, G.A., Wohletz, K.H., 1989. Numerical models of Plinian eruption columns and pyroclastic flows. J. Geophys. Res. 94, 1867–1887. Walker, G.P.L., 1985. Origin of coarse lithic breccias near ignimbrite source vents. J. Volcanol. Geotherm. Res. 25, 157–171. Wilson, L., Sparks, R.S.J., Walker, G.P.L., 1980. Explosive volcanic 667 eruptions-iV. The control of magma properties and conduit geometry on 668 eruption column behaviour. Geophys. J. R. astron. Soc. 63, 117–148. 669 Wohletz, K.H., McGetchin, T.R., Stanford II, M.T., Jones, E.M., 1984. Hy- 670 drodynamic aspects of caldera-forming eruptions: Numerical models. J. 671 Geophys. Res. 89, 8269–8285. 672 673 Woods, A.W., 1988. The fluid dynamics and thermodynamics of eruption columns. Bull. Volcanol. 50, 169–193. 674 Woods, A.W., Bower, S.M., 1995. The decompression of volcanic jets in a 675 crater during explosive volcanic eruptions. Earth Planet. Sci. Lett. 131, 676 189–205. 677 678 Woods, A.W., Caulfield, C.-C., P., 1992. A laboratory study of explosive volcanic eruption,. J. Geophys. Res. 97, 6699–6712. 31 679 Table 1: List of material properties and values of physical parameters 680 Table 2: Input parameters and flow regimes 681 Fig.1. Representative numerical results of the eruption column with a jet 682 structure (run 99). Parameters used and conditions at the vent are listed in 683 Tables 1 and 2, respectively. (a) Cross-sectional distribution of the mass frac- 684 tion of the ejected material (ξ) in the r–z space at 600 s. (b) Cross-sectional 685 distribution of the density difference relative to the stratified atmospheric 686 density at the same vertical position (∆ρ = ρ/ρa − 1) in the r–z space at 600 687 s. 688 Fig.2. Representative numerical results of the eruption column with a foun- 689 tain structure (run 33). Parameters used and conditions at the vent are 690 listed in Tables 1 and 2, respectively. (a) Cross-sectional distribution of the 691 mass fraction of the ejected material (ξ) in the r–z space at 190 s. (b) 692 Cross-sectional distribution of the density difference relative to the stratified 693 atmospheric density at the same vertical position (∆ρ = ρ/ρa − 1) in the r–z 694 space at 190 s. 695 Fig.3. Timewise distributions of the mass fraction of the ejected material 696 (ξ) along the central axis (spatial average within the range of |r| < L0 ) for 697 (a) run 99 and (b) run 33. 32 698 Fig.4. Time-averaged vertical profiles of physical quantities along the central 699 axis (spatial average within the range of |r| < L0 ) for run 99 (solid curves) 700 and run 33 (dashed curves). (a) The mass fraction of the ejected material, 701 ξ. (b) The upward velocity normalized by the exit velocity at the vent, 702 w/w0. The vertical axis represents the vertical position normalized by the 703 characteristic height (z ∗ = z/(w02 /2g)). 704 Fig.5. Time-averaged distribution of the pressure difference relative to the 705 stratified atmospheric pressure at the same vertical position, ∆p = p/pa − 1, 706 where pa is the atmospheric pressure. Cross-sectional distributions for (a) 707 run 34 and (b) run 33. Parameters used and conditions at the vent are listed 708 in Tables 1 and 2, respectively. 709 Fig.6. Representative numerical results of the column collapse from the jet 710 structure (run 78). Parameters used and conditions at the vent are listed in 711 Tables 1 and 2, respectively. (a) Cross-sectional distribution of the mass frac- 712 tion of the ejected material (ξ) in the r–z space at 120 s. (b) Cross-sectional 713 distribution of the density difference relative to the stratified atmospheric 714 density at the same vertical position (∆ρ = ρ/ρa − 1) in the r–z space at 120 715 s. 716 Fig.7. Representative numerical results of the column collapse from the 717 fountain structure (run 49). Parameters used and conditions at the vent 718 are listed in Tables 1 and 2, respectively. (a) Cross-sectional distribution of 719 the mass fraction of the ejected material (ξ) in the r–z space at 500 s. (b) 33 720 Cross-sectional distribution of the density difference relative to the stratified 721 atmospheric density at the same vertical position (∆ρ = ρ/ρa − 1) in the r–z 722 space at 500 s. 723 Fig.8. Time-averaged vertical profiles of physical quantities along the central 724 axis (spatial average within the range of |r| < L0 ) for run 78 (solid curves) 725 and run 49 (dashed curves). (a) The mass fraction of the ejected material, 726 ξ. (b) The upward velocity normalized by the exit velocity at the vent, 727 w/w0. The vertical axis represents the vertical position normalized by the 728 characteristic height (z ∗ = z/(w02 /2g)). 729 Fig.9. Cross-sectional distribution of the mass fraction of the ejected mate- 730 rial (ξ) in the r–z space for the result of run 46 at 190 s. Parameters used 731 and conditions at the vent are listed in Tables 1 and 2, respectively. The 732 contour levels are ξ = 0.8 and 0.9. 733 Fig.10. Timewise distributions of the mass fraction of the ejected material 734 (ξ) along the central axis (spatial average within the range of |r| < L0 ) for 735 run 92. Parameters used and conditions at the vent are listed in Tables 1 736 and 2, respectively. The contour levels are ξ = 0.9 and 0.95. 737 Fig.11. Cross-sectional distribution of the mass fraction of the ejected ma- 738 terial (ξ) in the r–z space for the result of run 27 at 590 s. Parameters used 739 and conditions at the vent are listed in Tables 1 and 2, respectively. 34 740 Fig.12. Flow-regime maps for the results of (a) group L, (b) group H, (c) 741 group I, and (d) group N. Parameters used are listed in Table 1. The condi- 742 tions at the vent for simulations are listed in Table 2. Pluses represent the 743 jet-type column regime. Circles indicate the fountain-type column regime. 744 Diamonds are the jet-type collapse regime. Triangles represent the fountain- 745 type collapse regime. Solid curves are the column collapse condition. Dashed 746 curves are the jet-fountain condition. Dotted thin curves are the column col- 747 lapse condition which are predicted by the 1-D model of Woods (1988) with 748 variable entrainment coefficients. The values at the edge of these curves are 749 the assumed values for the entrainment coefficient. 750 Fig.13. Flow-regime maps for the results of (a) group L, (b) group H, (c) 751 group I, and (d) group N in the non-dimensional space. Parameters used are 752 listed in Table 1. The conditions at the vent and non-dimensional parameters 753 for simulations are listed in Table 2. The horizontal axis is the Richardson 754 number (Ri) and the vertical axis is the Mach number (M). Pluses repre- 755 sent the jet-type column regime. Circles indicate the fountain-type column 756 regime. Diamonds are the jet-type collapse regime. Triangles represent the 757 fountain-type collapse regime. Solid curves are the column collapse condi- 758 tion. Dashed curves are the jet-fountain condition. Gray and white stars 759 are the critical points for the column collapse condition and those for the 760 jet-fountain condition for M= 1.2, respectively. 761 Fig.A.1. Maximum heights of the negatively buoyant jet/fountains (Hmax ) 762 as a function of the Richardson number (Ri) on the basis of the 3-D numerical 35 763 simulations. White and black points represent the jet-type and fountain-type 764 regimes, respectively. Grey plots are the intermediate regime between the jet- 765 type and fountain-type regimes. Circles, diamonds, and triangles indicate the 766 results for M= 1, 2, and 3, respectively. Arrows represent the JF condition 767 based on the Hmax –Ri relationship. 36 Table 1 Click here to download Table: Table1.pdf Variable Value (Unit) Meaning 9.81 (m s−2 ) gravitational body force per unit mass Rg 462 (J kg−1 K−1 ) gas constant of volcanic gas Ra 287 (J kg−1 K−1 ) gas constant of atmospheric air Cvs 1100 (J kg−1 K−1 ) specific heat of solid pyroclasts Cvg 1348 (J kg−1 K−1 ) specific heat of volcanic gas at constant volume Cva 713 (J kg−1 K−1 ) specific heat of air at constant volume Ta0 273 (K) atmospheric temperature at z = 0 km pa0 1.013 × 105 (Pa) atmospheric pressure at z = 0 km ρa0 1.29 (kg m−3 ) atmospheric density at z = 0 km μ1 -6.5 (K km−1 ) temperature gradient at z = 0.0–11.0 km μ2 0.0 (K km−1 ) temperature gradient at z = 11.0–20.0 km g 37 Table 2 Click here to download Table: Table2.pdf Simulation ṁ0 (kg s−1 ) w0 (m s−1 ) L0 (m) Ri M Regime1 group H (T0 = 1000 K, ng0 = 2.84 wt %) run 24 1.0 × 109 115 598 2.2 × 100 1.0 F-CC run 25 1.0 × 109 230 423 3.9 × 10−1 2.0 F-CC run 26 1.0 × 109 173 488 8.0 × 10−1 1.5 F-CC run 27 1.0 × 109 346 345 1.4 × 10−1 3.0 F-EC run 28 1.0 × 109 288 378 2.2 × 10−1 2.5 F-CC run 29 1.0 × 108 230 134 1.2 × 10−1 2.0 F-EC run 30 1.0 × 108 346 109 4.5 × 10−2 3.0 J-EC run 31 1.0 × 108 115 189 7.0 × 10−1 1.0 F-CC run 32 1.0 × 108 288 120 7.0 × 10−2 2.5 F-EC run 33 1.0 × 108 173 154 2.5 × 10−1 1.5 F-EC run 34 1.0 × 107 115 59.8 2.2 × 10−1 1.0 F-EC run 35 1.0 × 107 230 42.3 3.9 × 10−2 2.0 J-EC run 36 1.0 × 107 346 34.5 1.4 × 10−2 3.0 J-EC run 37 1.0 × 107 57.6 84.6 1.2 × 100 0.5 F-CC run 39 1.0 × 106 57.6 26.8 3.9 × 10−1 0.5 J-EC/F-EC run 40 1.0 × 106 115 18.9 7.0 × 10−2 1.0 J-EC run 41 3.2 × 106 57.6 47.6 7.0 × 10−1 0.5 F-CC run 42 3.2 × 106 115 33.6 1.2 × 10−1 1.0 J-EC run 43 3.2 × 106 173 27.5 4.5 × 10−2 1.5 J-EC run 44 1.0 × 106 80.7 22.6 1.7 × 10−1 0.7 J-EC 38 run 45 3.2 × 106 80.7 40.2 3.0 × 10−1 0.7 J-EC/F-EC run 46 3.2 × 107 196 81.6 1.0 × 10−1 1.7 J-EC/F-EC run 47 3.2 × 107 253 71.7 5.5 × 10−2 2.2 J-EC run 48 1.0 × 108 150 166 3.6 × 10−1 1.3 F-EC run 49 1.0 × 109 254 403 3.1 × 10−1 2.2 F-CC run 50 1.0 × 106 69.1 24.4 2.5 × 10−1 0.6 J-EC run 51 1.0 × 107 80.6 71.5 5.4 × 10−1 0.7 F-CC run 52 1.0 × 107 92.2 66.9 3.8 × 10−1 0.8 F-EC run 53 1.0 × 108 138 173 4.4 × 10−1 1.2 F-CC run 54 3.2 × 108 173 275 4.5 × 10−1 1.5 F-CC run 55 3.2 × 108 196 258 3.3 × 10−1 1.7 F-CC run 56 3.2 × 108 219 244 2.5 × 10−1 1.9 F-EC run 57 3.2 × 108 242 232 1.9 × 10−1 2.1 F-EC run 67 3.2 × 106 69.1 43.4 4.4 × 10−1 0.6 F-EC run 68 1.0 × 106 50.0 28.7 5.6 × 10−1 0.4 F-EC run 69 1.0 × 106 34.6 34.5 1.4 × 100 0.3 F-CC run 99 1.0 × 107 173 49.0 7.5 × 10−2 1.5 J-EC run 101 1.3 × 108 288 136 7.5 × 10−2 2.5 F-EC group I (T0 = 800 K, ng0 = 2.84 wt %) run 120 1.0 × 106 103 17.9 1.1 × 10−1 1.0 J-EC/J-CC run 121 4.0 × 106 103 35.7 2.1 × 10−1 1.0 F-EC/F-CC 39 run 122 1.0 × 107 103 56.6 3.4 × 10−1 1.0 F-CC run 123 2.5 × 107 103 89.7 5.4 × 10−1 1.0 F-CC run 124 4.0 × 106 206 25.2 3.8 × 10−2 2.0 J-EC run 125 1.0 × 107 206 40.0 6.0 × 10−2 2.0 J-EC run 126 2.5 × 107 206 63.4 9.5 × 10−2 2.0 J-CC/F-CC run 127 4.0 × 105 103 11.3 6.8 × 10−2 1.0 J-EC run 128 6.3 × 107 206 101 1.5 × 10−1 2.0 F-CC run 129 1.6 × 108 206 159 2.4 × 10−1 2.0 F-CC run 130 1.6 × 106 103 22.5 1.3 × 10−1 1.0 J-CC run 131 2.5 × 106 103 28.4 1.7 × 10−1 1.0 J-CC/F-CC group L (T0 = 550 K, ng0 = 2.84 wt %) run 58 1.0 × 106 128 13.3 7.8 × 10−2 1.5 F-EC run 59 1.0 × 106 145 12.5 5.7 × 10−2 1.7 J-CC run 60 1.0 × 106 171 11.5 3.8 × 10−2 2.0 J-CC run 61 1.0 × 107 214 32.6 6.9 × 10−2 2.5 J-CC run 62 1.0 × 107 231 31.4 5.7 × 10−2 2.7 J-CC run 64 1.0 × 107 171 36.4 1.2 × 10−1 2.0 J-CC run 65 1.0 × 107 128 42.1 2.5 × 10−1 1.5 F-CC run 66 1.0 × 107 85.5 51.5 6.8 × 10−1 1.0 F-CC run 70 1.0 × 106 214 10.3 2.2 × 10−2 2.5 J-EC run 71 1.0 × 107 248 30.2 4.7 × 10−2 2.9 J-CC 40 run 74 1.0 × 106 197 10.7 2.7 × 10−2 2.3 J-EC/J-CC run 75 1.0 × 106 205 10.5 2.4 × 10−2 2.4 J-EC run 77 1.0 × 106 188 11.0 3.0 × 10−2 2.2 J-CC run 78 1.0 × 106 179 11.2 3.4 × 10−2 2.1 J-CC run 80 1.0 × 105 111 4.52 3.5 × 10−2 1.3 J-CC run 81 1.0 × 105 120 4.35 2.9 × 10−2 1.4 J-CC run 85 1.0 × 106 85.5 16.3 2.2 × 10−1 1.0 J-CC/F-CC run 86 1.0 × 106 42.7 23.0 1.2 × 100 0.5 F-CC run 87 1.0 × 106 59.8 19.5 5.3 × 10−1 0.7 F-CC run 88 1.0 × 106 68.4 18.2 3.8 × 10−1 0.8 F-CC run 89 1.0 × 105 128 4.21 2.5 × 10−2 1.5 J-EC run 90 1.0 × 105 137 4.07 2.1 × 10−2 1.6 J-EC run 91 1.0 × 107 145 39.5 1.8 × 10−1 1.7 J-CC/F-CC run 92 1.0 × 107 154 38.4 1.6 × 10−1 1.8 J-CC/F-CC run 93 1.0 × 107 162 37.4 1.4 × 10−1 1.9 F-CC/F-CC run 94 1.0 × 107 137 40.7 2.1 × 10−1 1.6 F-CC run 95 1.0 × 107 94.0 49.1 5.4 × 10−1 1.1 F-CC run 96 1.0 × 107 103 47.0 4.3 × 10−1 1.2 F-CC run 97 1.0 × 107 111 45.2 3.5 × 10−1 1.3 F-CC run 98 1.0 × 107 120 43.5 2.9 × 10−1 1.4 F-CC 41 group N (T0 = 1000 K, ng0 = 1.23 wt %) run 103 1.0 × 107 75.6 48.6 1.1 × 100 1.0 F-CC run 104 1.0 × 108 75.6 154 3.4 × 100 1.0 F-CC run 105 1.0 × 109 75.6 486 1.1 × 101 1.0 F-CC run 106 1.0 × 107 151 34.4 1.9 × 10−1 2.0 F-EC run 107 1.0 × 108 151 109 6.0 × 10−1 2.0 F-CC run 108 1.0 × 109 151 344 1.9 × 101 2.0 F-CC run 109 4.0 × 106 75.6 30.7 6.8 × 10−1 1.0 F-CC run 110 4.0 × 106 151 21.7 1.2 × 10−1 2.0 J-EC run 111 4.0 × 105 75.6 9.70 2.1 × 10−1 1.0 J-EC run 112 4.0 × 105 151 6.86 3.8 × 10−2 2.0 J-EC run 115 1.0 × 106 75.6 15.4 3.4 × 10−1 1.0 J-EC run 116 1.6 × 106 75.6 19.4 4.3 × 10−1 1.0 F-EC/F-CC run 117 2.5 × 106 75.6 24.4 5.4 × 10−1 1.0 F-CC run 118 1.6 × 107 151 43.3 2.4 × 10−1 2.0 F-CC run 119 2.5 × 107 151 54.5 3.0 × 10−1 2.0 1 J-EC: jet-type eruption column, J-CC: jet-tyle column collapse, F-CC F-EC: fountain-type eruption column, F-CC: fountain-type column collapse 42 z [km] 3 z [km] 3 Figure 1 Click here to download Figure: figure1.pdf (b) a 3 0 0 0 (a) 0 r [km] 0.5 3 1 3 -0.5 0 r [km] 0 3 0.5 z [km] 3 z [km] 3 Figure 2 Click here to download Figure: figure2.pdf 3 0 0 (b) a 0 (a) 0 r [km] 0.5 3 1 3 -0.5 0 r [km] 0 3 0.5 1 10 Figure 3 Click here to download Figure: figure3.pdf 0 0 0.5 z [km] 5 (a) run 99 0 200 400 600 200 300 1 10 Time [sec] 0 0 0.5 z [km] 5 (b) run 33 0 100 Time [sec] Figure 4 Click here to download Figure: figure4.pdf 2 z* z* 2 1 1 (a) 0 0 (b) 0.2 0.4 0.6 0.8 1 0 -0.2 0 0.2 0.4 0.6 w/w0 0.8 1 1.2 z [km] 2 z [km] 2 Figure 5 Click here to download Figure: figure5b.pdf (b) run 33 2 -0.2 0 0 (a) run 34 0 r [km] 0 2 0.2 2 -0.2 0 r [km] 0 2 0.2 z [km] 0.5 z [km] 0.5 Figure 6 Click here to download Figure: figure6.pdf 0.5 0 0 (b) a 0 (a) 0 r [km] 0.5 0.5 1 0.5 0 r [km] -0.5 0 0.5 0.5 z [km] 10 z [km] 10 Figure 7 Click here to download Figure: figure7.pdf 10 0 0 (b) a 0 (a) 0 r [km] 0.5 10 1 10 0 r [km] -0.5 0 10 0.5 Figure 8 Click here to download Figure: figure8.pdf 2 z* z* 2 1 1 (a) 0 0 (b) 0.2 0.4 0.6 0.8 1 0 -0.2 0 0.2 0.4 w/w0 0.6 0.8 1 z [km] 4 Figure 9 Click here to download Figure: figure9.pdf 0.8 0 0.9 4 0 0 r [km] 0.5 4 1 1 2 Figure 10 Click here to download Figure: figure10.pdf 0.5 0 0.95 0 z [km] 1 0.9 0 50 100 Time [sec] 150 0 z [km] 10 Figure 11 Click here to download Figure: figure11.pdf 10 0 0 r [km] 0.5 10 1 Jet-type Collapse 100 5 6 7 10 10 10 6 10 200 0 5 10 Exit velocity [m/s] Exit velocity [m/s] 300 Jet-type Column Fountain-type Collapse 100 (c) group H (T0=1000 K n0 =0.0248) 6 10 7 10 8 10 9 10 10 10 Mass discharge rate [kg/s] 03 0. 0.1 0.0 05 7 0 8 10 9 10 Fountain-type Column Fountain-type Column 200 7 10 Mass discharge rate [kg/s] 0.0 3 0.0 5 0.0 7 0.1 0 1 0.0 01 (b) group I (T0=800 K Jet-type n0 =0.0248) Collapse Mass discharge rate [kg/s] 400 Fountain-type Collapse 100 0 5 10 8 10 Jet-type Column 3 0.0 5 0.0 7 0.1 0 0 (a) group L Fountain-type (T0 =550 K n0 =0.0248) Collapse 200 0.0 200 Jet-type Column 07 0. 0 300 1 0. Exit velocity [m/s] Exit velocity [m/s] 300 0. 03 04 0. 0. 0. Figure 12 Click here to download Figure: figure12.pdf 11 10 Jet-type Column Fountain-type Collapse 100 (d) group N (T0=1000 K n0=0.0123) 0 5 10 6 10 7 10 8 10 Mass discharge rate [kg/s] 9 10 Figure 13 Click here to download Figure: figure13.pdf 3 3 (b) group I (T0=800 K n0 =0.0248) Mach number Mach number (a) group L (T0 =550 K n0 =0.0248) 2 Fountain-type Collapse 1 Jet-type 2 Jet-type Column Fountain-type Collapse 1 Column Jet-type Collapse 0 0.01 0.1 Jet-type Collapse 1 0 0.01 10 Richardson number 1 10 Richardson number 3 3 (c) group H (T0=1000 K n0 =0.0248) Mach number Fountain-type Column Mach number 0.1 2 Fountain-type Collapse Jet-type Column 1 0 0.01 0.1 1 Richardson number 10 Fountain-type Column (d) group N (T0=1000 K n0=0.0123) Jet-type Column Fountain-type Collapse 2 1 0 0.01 0.1 1 Richardson number 10 Height of Negatively Buoyant Jet [m] Figure A1 Click here to download Figure: figureA1.pdf 104 M=3 M=2 103 0.01 M=1 0.1 1 Richardson Number 10
© Copyright 2026 Paperzz