Cold Regions Science and Technology 69 (2011) 156164

Contents lists available at SciVerse ScienceDirect

Cold Regions Science and Technology

An explicit numerical model for the study of snow's response to explosive air blast
D.A. Miller , R.G. Tichota, E.E. Adams
Department of Civil Engineering, Montana State University, Bozeman, MT, United States

a r t i c l e

i n f o

a b s t r a c t
In this paper, an analytic tool is used to examine the internal dynamic snow response during explosive events. An explicit nonlinear dynamic model (using ANSYS/AUTODYN) is presented where the explosion, shock propagation through air and snow response is simulated in a single analysis. This versatile approach handles the complex interactions from explosive events and solids, gases and liquids. Nonlinear interactions and responses are modeled during the detonation and subsequent propagation. The model predicts internal structural response during explosive events, including important parameters such as stress, strain, density changes, velocity and acceleration. Snow shock Hugoniots are used for volumetric constitutive relationships with the deviatoric relationships modeled as linear elastic. The analysis shows stress waves in the snow resulting from the explosive shock wave traveling over the surface. While the normal load transits the weak layer, a shear stress wave concentrated above the weak layer develops. The intensity at depth and lateral extent of the stress wave may be an important consideration for initiating avalanches with explosives. Examples with charges on and above the snow surface support the well known air burst advantage, but also show that the dynamic enhancement is not due to peak air pressure alone. Results for two explosive congurations support enhanced dynamic response with increased air pressure impulse, providing further insight into the suspended charge advantage. Charge size is briey examined with larger explosives providing an advantage in stress wave intensity and range. Various snowpack congurations and explosive charges with variable locations can be examined with this approach. The analytic approach provides a tool for future detailed examination of critical avalanche control parameters. 2011 Elsevier B.V. All rights reserved.

Article history: Received 19 November 2010 Accepted 6 August 2011 Keywords: Avalanche control Snow explosives Explicit model

1. Introduction Explosives are routinely used to initiate avalanches to stabilize snow on slopes. Many ski resorts and highway departments use this method as part of efforts to maintain public safety by reducing avalanche hazards. Understanding the dynamic response of a snowpack during shock wave interaction is important for effective use of avalanche control explosives. Currently, the probability of success in inducing an avalanche is based largely on the personnel experience in charge type, size, placement and timing coupled with historical performance of a particular slope. Unintended avalanche release, after explosive control efforts, has recently had a devastating toll. During the 2008/09 winter, 4 people were killed and another 18 buried within US ski resort boundaries by avalanches (Abromeit, 2010). Since 2009, avalanches within US resort boundaries have killed three additional people. Most of these incidents were classied post control release where the avalanche unexpectedly initiated after explosive control efforts failed to stabilize the slope. While we cannot yet identify all the causal mechanisms of these events, they have
Corresponding author at: Department of Civil Engineering, Montana State University, PO Box, 173900, 205 Cobleigh Hall, Bozeman, MT 59717, United States. Tel.: + 1 406 994 6118. E-mail address: [email protected] (D.A. Miller). 0165-232X/$ see front matter 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.coldregions.2011.08.004

motivated new efforts looking at responses of snowpacks to explosive events. Historically, there has been signicant experimental investigation of explosives detonated on or above the snow surface (eg: Gubler, 1977; Ingram, 1962; Joachim, 1967; Wisotski and Snyer, 1966). These studies revealed the enhanced effectiveness of suspended explosives in inducing avalanches. As a result, suspended techniques are commonly practiced in the avalanche control industry. Buried explosives dedicate signicant energy to crater formation with local dissipation yielding little momentum transfer away from the crater (Johnson et al., 1994). Detonations on and above the surface are examined here, but the numerical techniques could accommodate analysis of buried explosives in snow covers. Detonations in the air produce a spherically expanding shock wave that decreases in intensity with distance from the blast due to geometric wave expansion and medium attenuation. As the wave travels over the snow surface, energy from the air shock is transmitted to the snowpack inducing shock in the pores and stress waves within the ice network. Within the snow, this produces a spherical shock that dissipates not only due to geometric expansion but also due to snow compaction. As detonation height above the snow is increased, the air pressure immediately beneath the blast decreases, but the air pressure at distances away from the blast will increase to a maximum value and then decrease as the explosive is raised further. Johnson et al. (1994) analyzed experimental air blast data from Ingram (1962) and Wisotski

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164

157

and Snyer (1966) (summarized in O'Keeffe (1965) and Mellor (1985)) and presented a scaled curve of maximum air pressure as a function of burst height and radius from the blast vertical axis. This curve identies the scaled burst height for maximum pressure at particular scaled distances. The data supports ~40% increase in overpressure as the detonation height is increased from the surface to the optimal height (for a particular range and net explosive weight). The maximum overpressure was relatively insensitive to scaled detonation heights above ~2 m/kg1/3. Gubler (1977) conducted eld experiments where charge mass, snowpack stratigraphy, explosive type, charge placement relative to the surface and ground type were considered. One signicant conclusion of Gubler's work is that 1 kg charges detonated from 1 to 2 m above the surface have enhanced results for releasing dry slab avalanches. Ueland (1992) used mining seismographs to measure the vertical response of snowpacks during explosive events. His work conrmed the air blast advantage over surface or buried charges, but also examined shock attenuation through the snowpack depth. He found snow hardness, more than density, was a signicant factor in shock attenuation. With the relationships of range, explosive weight and blast height established; what peak pressure at what range is required for avalanche release? Few data exist to answer this question, but Mellor (1973) suggests loading the anticipated avalanche release zone with at least 3.5 kPa, but detailed analysis or empirical data justifying this limit is lacking. A numerical approach that can predict internal snowpack responses would be valuable for investigating load distributions through the depth of the snowpack at various ranges from the blast center. Brown (1981) developed jump equations to describe the change in snow physical parameters across a shock wave. Of particular interest, he predicted the change in snow density across the shock. He found it difcult to validate his approach due to a shortage of good shock wave measurements in snow. Johnson (1991) developed a momentum model to predict shock wave attenuation in snow. He showed that snow attenuation was largely dependent upon the snow pressure density relationship. To predict the snow response to explosives, volumetric constitutive relationships must be used to describe snowpack compression during loading. Haehnel and Shoop (2004) present a capped DruckerPrager model simulating the loaddeformation characteristics of snow. Their model provides upper and lower bounds for the response of low density snow loaded at high strain rates, but was applied to tire movement through snow, not to explosive loading. Some experimental research has focused on snow response to explosives to better develop constitutive relationships. Furnish and Boslough (1996) conducted impact tests on snow samples and material simulators to experimentally derive shock Hugoniot states, reshock characteristics and release properties. They report reliable snow Hugoniot states, with snow density of 500 kg/m 3, at stresses up to ~4 GPa. Johnson and Solie (1993) conducted gas gun impact tests (with analysis) to determine the pressuredensity relationships for several initial snow densities with stresses up to 40 MPa. In all of the tests, snow had signicant shock attenuation. Johnson and Solie (1993, 1994) discuss snow's large load hysteresis due to its small volume recovery during unloading. This compaction characteristic is very important for understanding snow's response to explosives. In close proximity to the blast, the snow will compact to a critical density value before accepting signicant stress. The pressure required to compact snow to a nal density value increased with decreasing initial density. To initiate slab avalanche release, Heierli et al. (2008) present a two stage process for fracture and subsequent avalanche. In the rst stage, normal and shear stress combine in loading a preexisting aw. When the mechanical energy reaches a critical value for a particular crack nucleus, fracture progresses with a mixed mode anticrack. This failure mode is driven by a volumetric collapse of a layer containing the crack nucleus. After fracture, the nature of contact forces between the crack faces dictate if the slab will slide down the hill. This theory predicts the frequently observed phenomenon of crack propagation

with no accompanying avalanche (i.e. whumpfs). While Heierli et al. (2008) focuses on gravitational snow loading for the shear and normal comments of mechanical energy, they do briey mention other triggers such as explosives. While these studies have provided critical data and insights, modern analytical tools integrating the explosive, atmosphere, terrain and snowpack have not seen application. Historically there has been experimental investigation of explosives on and over snow, but there exists virtually no numerical or analytic research into the snow's dynamic response to explosive detonations, particularly at the avalanche slope scale. A comprehensive approach capable of examining the internal snow response during explosive loading is needed. Such an approach or analytical technique would open the door for studying the critical parameters and sensitivities eventually leading to enhanced effectiveness of explosive charges on avalanche prone slopes. This study is a rst step in such an approach applying modern analytic and numerical techniques to the avalanche control situation. The goal of this study is to develop an appropriate analytical tool capable of predicting snow response to an explosive event. The approach should allow for multiple snow and explosive types and congurations. Internal snow dynamic response must be predicted and available for analysis. 2. Approach 2.1. Explosive air blast Following an explosion in air, a compressive shock wave is generated and propagates radially through the air at supersonic speeds. Across a shock wave, the air experiences an increase in temperature, pressure and velocity from the front passage. The strength of a normal shock front is commonly characterized by the size of the pressure increase across the front, the wave propagation velocity and the increase in particle velocity of the medium. Other thermodynamic parameters such as stagnation pressure and stagnation temperature are also used. For avalanche control with air blast, this pressure jump is the primary loading mechanism into the snowpack as the high pressure air behind the shock loads the surface. For explosive shock waves in air, the shock strength diminishes as it propagates radially from the blast center. Eventually, the shock wave degenerates into a sound wave as the wave speed becomes sonic. Explosive blast wave characteristics depend upon the energy release and the medium through which the wave moves. The performance of a common explosive, TNT, has been studied extensively, consequently, its performance is well understood and documented. Scaling law can be used to estimate the blast wave characteristics of other explosives. Kinney and Graham (1985) present a comprehensive TNT scaling law approach. Their method is presented here to introduce basic air blast characteristics of a common avalanche control explosive. The method utilizes geometric similitude (spherical for this case) and conservation of momentum. A scaled distance allows other explosives at actual distances to be scaled up or down based on equivalent TNT distance. The scaled TNT equivalent distance z (m/kg 1/3) is given by fd z o W3
1

z=

where zo (m) is the actual distance of interest, W (kg) is the TNT equivalent explosive mass and fd is an atmospheric transmission factor. Considering air as an ideal gas, fd is given by fd =  1  1 P 3 To 3; Po T 2

158

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164

where Po (Pa) and To (K) are reference pressure and temperature and P (Pa) and T (K) are the actual atmospheric pressure and temperature. Once the TNT equivalent distance is found, the explosive parameters in question may be scaled to the well known TNT equivalent. In Eq. (1), the scaled distance is inversely proportional to the cube root of the TNT equivalent yield (due to spherical expansion), implying that to create a particular blast at double the distance requires eight times the explosive energy. As an example, Fig. 1 shows the air blast overpressure as a function of distance from the blast for 0.9 and 1.8 kg pentolite charges. The dramatic decrease of blast pressure with distance is shown as is the inuence of doubling the explosive weight for pentolite cast boosters, a common avalanche control charge. While doubling the charge does enhance over pressure, it is not a linear increase. Increasing explosive weight alone may not always be the most viable or efcient way to increase effectiveness for avalanche release. Factors such as terrain and snowpack conditions play a role in control efcacy. While scaling law helps quantify the rst order inuence of charge size and distances, it does not adequately address the more complicated scenario of detonating explosives on or over snowpacks and is not operationally utilized. A numeric approach for examining air blast and snow interaction is now presented. 2.2. Explicit nonlinear numerical modeling The mining community has made signicant research investment in the use of explosives for the fracture and movement of rock. The current state of the art approach utilizes an explicit nonlinear method for evaluating rock movement during explosive events. AUTODYN is a commercial explicit modeling program produced by ANSYS, Inc. AUTODYN has particular strengths in modeling high energy, short duration explosions and high velocity impacts by predicting the nonlinear behavior of solids, gases and liquids along with their interactions. One distinct advantage AUTODYN has for the current application is the ability to handle nonlinear solids and gases simultaneously. Preece and Lownds (2008) present a 3D method for analyzing rock blasting using explicit computational models. Their work motivated the current study in applying AUTODYN to snowpacks and explosives. While the response of snow is very different from rock, the powerful constitutive models and complex solid/gas interactions available in AUTODYN are encouraging. Stress and strain distributions, snow compaction, position, velocity and acceleration elds through the snow slope based on multiply

congured detonations with actual terrain features can be predicted. In the current study, a modeling approach was developed with suspended and surface based explosives. The model uses a coupled air/snow interaction and simulates the complex nonlinear interactions and responses. With this model, several variables can be studied and varied to determine parameter sensitivities and optimized solutions. The effects of thin snow cover, rebound of shock waves from rock walls, height of the charge above the snowpack, size of the charge, multiple charges, snow type/prole and shock wave coupling to snow are some of the topics in need of investigation. For this initial study, we demonstrate feasibility of the approach and limit the examples to a few select parameters such as charge placement and charge size on a single snowpack conguration. 2.3. AUTODYN model To demonstrate the approach and make preliminary investigations into explosive responses, an example conguration of a high density hard slab resting on a thin low density layer supported by a high density base was developed. A snowpack comprised of a 1 m base, 1 cm weak layer and 0.5 m top slab was developed. Snow densities (shown in Fig. 2) were selected due to the availability of explosive response data (presented later) and they are representative of a hard slab supported by a weak layer. A 2D axisymmetric square mesh (10 mm) was utilized for both the air and snow with the axis of symmetry about the vertical axis where the explosive is placed. This axis is referred to as the blast axis. The resulting simulation is a disk of snow and air with the explosive centered on or above the snow in the air. For this initial example, 1.8 kg of cast pentolite is suspended 1 m above the snow surface. A wedge visualization of the model is shown in Fig. 2 with air, snow and explosive identied. In Fig. 2, the air is modeled with Eulerian elements where the explosive gases and shock wave travel though the mesh. The snow is treated as a Lagrangian solid where the mesh deforms with the material. At the snow and air interface, Lagrange-Euler interactions are dened allowing the air to interact with the snow. The interaction of the solid and gas phases with the LagrangeEuler interface sets this approach apart and allows for direct loading of the snowpack from shockwaves traveling through the air. While each approach taken individually is common, their simultaneous implementation makes this a particularly powerful approach. 2.4. Constitutive relationships To adequately describe the snow response to loading, volumetric and deviatoric constitutive relationships must be dened. A shock wave is a discontinuous jump in states across an abrupt front. Prior to

100000

10000

AUTODYN Model Configuration

Explosive Air Snow

100

10

1
1.8 kg Pentolite

=400 kg/m3

=111 kg/m3

0.1 0.1

10

100

Distance from blast (m)

Fig. 2. AUTODYN axisymmetric model with 90 slice shown. Blue area is air, green is high density snow, black is a low density weak layer and red is the explosive. The snow/ air disk is 10 m in diameter, the upper snow layer is 0.5 m thick, the bottom layer is 1 m thick and the weak layer is 1 cm thick. The explosive is 1.8 kg of pentolite and is suspended 1 m above the snow surface. A 1 cm square mesh is used throughout the model.

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164

159

shock arrival, the original material state can be described by: the particle velocity (uo), material density (o), specic internal energy (eo) and pressure (Po). As the shock wave passes through the material, each of these change to new values u1, 1, e1 and P1 behind the front. Cooper (1996) derives relationships for mass conservation, momentum balance and energy conservation across the shock front, yielding (in order) the following relationships (with uo = 0): 1 U = ; Uu1 o P1 Po = o u1 U; e1 eo =   1 1 1 P1 + Po ; 2 o 1 3 4 5

And by eliminating U using Eqs. (3) and (4), the particle velocity after the shock is given by:   1 1 2 u1 = P1 Po : o 1

where U is the shock wave velocity. In Eqs. (3), (4) and (5) with the initial state known, there are ve unknown variables. To solve these relationships, more information is needed. For air, the ideal gas assumption is used. For a solid, a volumetric equation of state describing equilibrium states in terms of energy, pressure and density would provide the necessary relationships. If an equation of state existed such that e = f1(P,), then it could be used in the energy equation (Eq. (5)) eliminating e leaving P = f2(). Unfortunately, such an equation of state is unknown for most solids and in particular for snow. An expression relating any two of the four unknowns in the mass (Eq. (3)) and momentum (Eq. (4)) equations, that included equilibrium states, would provide an alternative to the equation of state. With an equation of state and using the shock velocity, U, as a boundary condition, the equations can be solved. Shock Hugoniots are experimentally derived empirical relationships for particular materials and are normally expressed as P = f3(), P = f4(u) or U = f5(u). In AUTODYN, the shock equation of state describes the volumetric material response as U = f5(u). For most materials, this is a linear relationship (Cooper, 1996) given by U = Co + su1 ; 6

where Co is the intercept and s is the slope of the experimental shockparticle velocity data. Limited experimental data exists for snow and suitable substitutes are not known. Fortunately, Johnson and Solie (1993) snow gas gun experiments and analysis yielded P = f3() data on a wide range of initial snow densities. Their data suggests two modes of snow loading; large deformation compaction followed by rapid load increase with little change in density. Large deformations occur as the snow is initially compacted. As the density increases, there is a sudden increase in deformation resistance requiring higher pressures for further strain. The critical density where signicant strain hardening begins was dependent upon the initial snow density. They found a power law relationship between pressure and density with derived coefcients (a and b) for pressures 220 MPa as P = ao :
b

Using Eqs. (8) and (9) with Johnson and Solie (1993) data, linear U = f(u1) curves were developed. A linear regression yielded the coefcients in Eq. (6) for implementation in AUTODYN. This translation was completed for each initial snow density with excellent t (R 2 N 0.99). The shock equation of state parameters for Eq. (6) are given in Table 1, completing the volumetric constitutive relationships. For deviatoric deformations, the snow was modeled as linear elastic followed by brittle failure using the maximum principal stress failure criterion for each snow density. While the equation of state describes the element volumetric response, the deviatoric deformations result in changes to element shape. In this case, the deviatoric stress/strain relationship is linear up to a critical maximum where failure is assumed to be instantaneous and complete. The assumption of deviatoric brittle failure is common for snow and implicit in fracture analysis (macroscopically). While deviatoric plastic deformations are not included, the equation of state does consider signicant volumetric compaction. Elastic moduli, shear strengths, maximum compressive normal stresses and maximum octahedral stresses were taken from material summarized in Mellor (1975) and are based on snow density. For the current example, the slab and base layers ( = 400 kg/m3): Young's modulus = 60 MPa, maximum compressive normal stress = 2 MPa, maximum shear stress = 0.5 MPa and maximum octahedral stress= 0.2 MPa. For the weak layer ( = 111 kg/m3): Young's modulus = 0.2 MPa, maximum compressive normal stress =0.03 MPa, maximum shear stress = 0.04 MPa and maximum octahedral stress= 0.015 MPa. AUTODYN uses superposition for the volumetric (from the equation of state) and the deviatoric contributions to get the overall material state. While the example represented here is limited to distances 5 m from the blast axis, it is believed that this approach would be appropriate for greater distances and is being considered in future efforts. The explosive is modeled with a JonesWilkinsLee (JWL) equation of state using pentolite data based on manufacturer provided JWL empirical parameters, detonation velocity and chemical energy. This model is used in conjunction with the modeled cast booster physical conguration to predict the high explosive detonation and rapid expansion of gas products. The resulting high pressure gases are allowed to operate in the Eulerian air elements. 3. Results 3.1. Air blast pressure The shock wave in the air, for a suspended 1.8 kg pentolite charge, moves through the air, impinges on and travels across the snow surface. Fig. 3 shows snap shots of the pressure wave in the air and across the snow surface at three different times after detonation. The shock front produces enhanced pressure at the snow surface as the wave reects obliquely and acts on air that has already been impacted by the primary shock. As the shock wave propagates, the pressure and wave velocity decease due to geometric expansion.
Table 1 Shock equation of state parameters derived from Johnson and Solie (1993). Initial snow density Co 53.9 m/s 7.63 m/s s 1.86 1.43

For pressures up to 40 MPa, additional exponential terms were added to Eq. (7). All of the snow densities considered in this study were included in Johnson and Solie (1993), but are of the form of Eq. (7). The AUTODYN shock equation of state requires the data in the form of Eq. (6). To translate for use in the numerical model, Johnson and Solie (1993)P relationships were calculated and then converted to Uu. By rearranging Eqs. (3) and (4) and eliminating u1, the shock velocity is given by   2 U = P1 Po o o = 1 :
2

400 kg/m3 111 kg/m3

160

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164

Shock wave 0.5 m

1.0 m 3.7 m

B
X Y

Shock wave

Fig. 4. Snow cross sectional vertical normal stress contours at times after detonation coincident with Fig. 3 (B) and (C) respectively. The planar view is of the snowpack. The current position of the shock wave is noted. The detonation originated along the right hand side of the plots and the shock is traveling to the left. The dashed line is the weak layer.

Fig. 3. Air pressure contours at times after detonation of 0.25 ms (A), 2.4 ms (B) and 4.55 ms (C) respectively. Snow is shown in green (no data) at the bottom for reference. In A, the shock wave is entirely contained in the air and has not yet reached the snow surface.

3.2. Snow stresses As the blast transits the snow surface, the EulerLagrange interaction feature in AUTODYN allows the air to load the snow and for the calculation of several physical parameters. In particular, the stresses that develop in the snow are of interest for examining the effects the explosive has internally on the snowpack. As the shock wave moves across the snow, a stress wave, consisting of normal and shear stresses, develops at the surface and moves into and through the snowpack. The vertical normal stresses along a radial slice of the model are shown in Fig. 4 at two different times, coincident with Fig. 3 (B) and (C). The view is a 3.7 m wide cross section of the snowpack with the shock wave originating on the blast axis (right) and moving left (+Y direction). A normal compressive stress wave moves through the snow as the shock wave transits the surface. The compressive stress wave decreases in intensity through the snow depth and radially with shock movement. In the area directly beneath the

explosive, the snow density increased from 400 kg/m 3 to ~ 480 kg/m 3 (according to the equation of state described in Section 2.4). When the wave reached the weak layer, the compressive normal loads transmitted through the weak layer from the slab to the underlying base layer. While the weak layer did not signicantly impede the compressive normal stress, it did have an interesting impact on the shear stress. The analysis points to a rolling shear stress wave that moves through the snow concentrated in the slab near the weak layer. Fig. 5 shows the shear stresses in the snow cross section at two different times, also coincident with Fig. 3(B) and (C). In Fig. 5, a shear stress wave moves through the snow and concentrates above the weak layer; very little shear load is transmitted to the snow below the weak layer. In this case, the stress was sufcient to cause weak layer failure, eliminating the layer's ability to transmit shear load. It is believed that the depth, intensity and extent of these stress waves may be a primary factor in determining the effectiveness of explosive avalanche control. 3.3. Surface blast In the next analysis, the scenario described in Fig. 2 is repeated, but with the 1.8 kg charge placed on the snow surface. As expected, a crater developed along with high snow stresses in the vicinity of the crater. The compaction of the snow (according to the constitutive relationships described in Section 2.4) was monitored here through material density. Fig. 6 shows the snow density contours at two different times after the surface detonation. In Fig. 2, the hard slab and base layers had an initial density of 400 kg/m 3 and the weak layer density was 111 kg/m 3. In Fig. 6(A), a compaction zone has moved through the slab below the crater, due to compressive normal stresses, resulting in density increases. The

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164

161

Shock wave

Shock wave

Shock wave

Fig. 5. Snow cross sectional shear stress contours at times after detonation coincident with Fig. 3 (B) and (C) respectively. The planar view is of the snowpack. The current position of the shock wave is noted and the geometry is identical to Fig. 4. The detonation originated along the right hand side of the plots and the shock is traveling to the left. The dashed line is the weak layer.

Fig. 6. Snow density for a surface blast at t = 3 ms (A) and t = 7 ms (B) after detonation. In (A), the shock wave location is shown, but in (B), the shock wave has exited the view through the left boundary.

density behind the stress wave rebounded somewhat after the stress wave passage. Johnson and Solie (1993) use analysis and post test density measurements to estimate release moduli and report a large range of values that vary widely with pressure. It is also noted that the peak density decreased from (A) to (B) as the stress wave attenuated and the snow rebounded somewhat, consistent with the hysteresis observed in the experiments of Johnson and Solie (1993). The weak layer is evident and has been compressed, but to a lesser extent based on its lower initial density. The crater in Fig. 6 was formed by releasing material after analytical material failure when the snow experienced very high strain. In this case, the crater did not extend below the weak layer, but the compaction zone did extend to a depth ~0.8 m below the original surface. The stresses in close proximity to the crater zone were very large when compared to stresses in the same snowpack location from the suspended charge. As the shock wave propagated away from the surface blast location, the stresses deceased quickly. The air blast shear and normal stresses were generally higher than for the surface blast at distances beyond ~1 m from the blast axis. For example, when the shock wave is located at 3.65 m from the blast axis (ref Figs. 3 and 5 (C) for air blast), the peak shear stress experienced by the upper slab layer was ~50% greater for the air blast when compared to surface detonation, supporting Gubler's (1977) experimental snowpack observations for suspended charges. While snow stresses were not generally reported, Johnson et al's (1994) summaries suggest ~40% increase in peak air pressure from raising the explosive above the surface to the optimal height. This increase is based on maximizing air pressure for a particular range while the current simulation is based on a typical avalanche control placement, making direct comparison difcult. As discussed later, peak air pressure alone does not explain the increases in shear stress due to explosive elevation.

Stresses are one good measure of explosive effectiveness in snow layers, but there are other dynamic parameters that can be examined. Particle velocity (u1) is the velocity of a particle as the wave transmits through the medium. In this case, particle velocity is found as the stress wave moves through snow and is a measure of snow disturbance. Additionally, particle acceleration (the time derivative of particle velocity) can be used to estimate forces in the snow from Newton's Second Law. Table 2 summarizes the peak vertical (downward) velocity and acceleration at the middle of the top layer slab at various radial locations for the two scenarios. At very close range to the blast, the peak vertical velocity and acceleration in the middle level of the snow slab are much greater for the surface blast. In particular, the velocity and acceleration at 0.25 m are signicantly higher for the surface blast. Between 0.25 m and 0.5 m, the suspended charge becomes the dominant method for inducing vertical velocity and acceleration. As distance from the blast axis increases, the advantage for the suspended explosive is evident in higher peak velocity and accelerations. While this analysis supports the well known avalanche control advantage of suspended explosives, it has not yet provided insight into why this is the case. The only difference between the two scenarios is

Table 2 Peak vertical velocity and acceleration in the middle of the snow top layer slab at various ranges from the blast axis for a 1.8 kg pentolite charge detonated from 1 m above the surface (1 m) and from the surface (Surf). Distance from blast axis (m) 0.25 0.5 1 3 Velocity (m/s) 1m 18.7 16.2 9.80 4.60 Surf 46.6 12.2 4.80 3.70 Acceleration (m/s2) 1m 139 116 69.9 67.1 Surf 453 53.8 34.1 26.6

162

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164

the location of the explosive. This location difference results in a varied load input to the snowpack. To investigate the differences, air pressure is now examined. Fig. 7 shows the peak air pressure and percent differences (from surface blast) at various distances from the blast axis located 3 cm above the snow surface. Since the evaluation points are close to the snow surface, the surface detonation has signicantly higher peak air pressure in the vicinity of the blast due to close proximity with slant range affecting the suspended results. As the range from the blast increases, the suspended blast has higher peak pressures with the decay of the two pressure peaks appear to reasonably track each other. The suspended blast has slightly larger peak pressures at ranges beyond ~2.2 m. While there are differences in the two scenarios, the suspended charge advantage (at large ranges from the blast axis) doesn't appear to be from enhanced peak pressure alone. The air pressures in Fig. 7 are higher than the open air blast predicted by scaling law alone (Fig. 1), again pointing to the limited application of scaling approaches. The air pressure contours in Fig. 3 (B) and (C) show pressure intensication near the snow surface, when compared to the open air scaled estimates. In part, this pressure enhancement is from previously discussed (experimentally veried summary in Mellor (1985)) shock reection off the snow surface, but modeling differences cannot be ruled out as a factor. The shock wave is an N-shaped pressure wave with a nearly instantaneous jump in pressure followed by decay as the wave moves past a xed position. The pressure impulse, the area under the overpressuretime curve, is found by integrating the overpressure over the time it acts at particular points. The pressure impulse accounts not only for the peak pressure, but is also inuenced by pressure wave action time and shape which can vary with shock engagement geometry, detonation velocity and gas volume.

As shown in Fig. 8, the pressure impulse from the surface blast is much greater than the suspended pressure impulse in close proximity to the blast axis, but the suspended impulse quickly exceeds the surface values as range increases. At distances N3 m from the blast axis, the impulse is ~50% greater for the suspended charge. Much of the surface charge energy is dedicated to crater generation, leaving less momentum spreading to the far eld (Johnson et al. (1994)). The pressure impulse represents the integrated pressure over time at points above the snow surface and is the primary loading input to the snow. It is likely that this increased impulse (at far eld) created by suspending the charge is a contributor in increasing snowpack dynamic responses. In addition, the shape of the N wave may also be important since many wave shapes can have identical impulses but may produce different results. 3.4. Charge size Next, the scenario described in Fig. 2 is repeated but with a 0.9 kg pentolite charge in the 1 m suspended position. This scenario has half of the net explosive weight than the previous suspended explosive example. To compare the two different charge sizes, the shear stresses in the middle of the top slab layer (0.25 m below the snow surface) are captured when the shock wave is in the position represented in Fig. 3 (B). Since the net explosive weights are different, the shock velocities and shock arrival times are also different, so the comparison is based on equivalent geometric location of the shock wave. A prole plot of shear stress across the middle of the top slab layer is presented in Fig. 9. In Fig. 9, it is noted while the peak shear stresses from the larger explosive are higher, they are not so in proportion to the net explosive

Fig. 7. Peak air pressure at 3 cm above the snow surface for suspended and surface detonations at various ranges from the blast axis. The peak air pressure percent difference, as compared to the surface charge, is shown.

Fig. 8. Pressure impulse at 3 cm above the snow surface for suspended and surface detonations at various ranges from the blast axis. The pressure impulse percent difference, as compared to the surface charge, is shown.

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164

163

weights. In this particular location, the doubling of the net explosive weight resulted in only a 21% percent increase in peak shear stress. When the shock wave later reached the location shown in Fig. 3 (C), the shear stress wave for the smaller charge was beginning to dissipate and lose denition while the shear wave from the 1.8 kg charge maintained a distinct concentration above the weak layer (ref Fig. 5 (C)). While the stresses may not be proportionally higher when the net explosive is doubled, the effective range is increased, but to what extent is currently not known and is a topic for further research. 4. Summary Using AUTODYN as an evaluation tool for snow's response to explosive events has the advantage of handing nonlinear interactions of gases and solids with complex material responses. The application to air and layered snowpacks provides an avenue for snow responses to explosives. The model allows for internal examination of vital snowpack parameters during all phases of explosive events. The examples selected for this study demonstrate the utility and power of the approach, but are far from comprehensive from an avalanche control perspective. Various parameters such as snow stress, density, velocity, acceleration, air pressure and pressure impulse were selected for examination and comparison, demonstrating the modeling versatility and strengths. There are many other potential parameters that could be examined in future analyses, again, a testament to the approach. The snow response for a particular air blast scenario revealed a traveling stress wave created by the shock front as it propagated over the snow surface. Both normal and shear stress were present supporting current thought in layer fracture and avalanche release. It is believed that the intensity, depth and lateral extent of this stress wave may be a fundamental measure of explosive effectiveness. In the specic case presented, the shear stress was intense enough to fail the weak layer throughout the model, which prevented transmission of signicant shear forces to the base layer. Other snowpack conditions can be envisioned where an explosive would not sufciently fail a layer or provide enough force to initiate an avalanche after fracture. The development of a crater and signicant snow compaction was also demonstrated on a surface blast example. In our example, the stresses, velocity and acceleration in the vicinity of the blast were much greater than for a suspended equivalent charge. As the shock wave propagated away from the blast center, the suspended charge produced enhanced dynamic response in the top slab layer. These results suggest that there may be an optimal conguration above the snowpack to maximize dynamic snow response that may not be based on peak air pressure alone. If the charge is suspended too high, the decay in air pressure due to spherical expansion will eventually outweigh the height of burst advantage. Conversely, if the charge is

too close to the surface, excessive explosive energy may be apportioned near the blast and reduce shock geometric advantage thereby limiting widespread effect. The increased effectiveness for explosives suspended 12 m above the snowpack has been experimentally developed and routinely employed, but the current approach provides the opportunity to study cause and effect for a variety of parameters in search of that optimal placement for particular snowpack conditions. The balance between charge size and placement may be used to optimize pressure impulse to maximize the snowpack loading over a large area. If a particular avalanche control situation presents a well dened localized trigger point with a high probability of propagation, a surface placement may be sufcient or even preferred. Reducing the explosive size in air blast revealed a slight reduction in stresses in the vicinity of the blast as well as a reduction in effective range of the rolling shear stress wave. While the methods developed here are sound and appropriate for the problem, this study is just scratching the surface for evaluating avalanche control effectiveness. The approach is fundamentally a tool that can focus future studies. The numerical examination of optimal charge size and placement, inuence of various snowpack conditions, use of available terrain features, evaluation of buried charges or artillery are examples of future studies that can now be analytically examined. One important aspect is to align the approach with measured eld data. In 2010, we developed an explosive research site with instrumentation to measure the blast pressure and snowpack dynamic response to explosives (Tichota et al., 2010). We plan to closely couple the measured responses and the analysis to further establish the techniques and to rene critical parameters for increasing avalanche control effectiveness. The analytical model merged with the experimental campaign comprises a new explosive avalanche control research program. Both projects are in their infancy, but have established new insights and techniques for supporting research. Acknowledgments The authors would like to thank the National Avalanche Center and snow safety personnel from Bridger Bowl, Yellowstone Club and Big Sky ski areas for their insights and support of the project. We would also like to acknowledge the two reviewers and the technical editor who provided excellent recommendations for improving the paper. References
Abromeit, D., 2010. Inbounds incidents & fatalities 2008/9. Avalanche Review 28 (3), 26. Brown, R.L., 1981. A method for evaluating shockwave propagation in snow. Cold Regions Science and Technology 5, 151156. Cooper, P.W., 1996. Explosives Engineering. Wiley-VCH, New York Chichester Weinheim Brisbane Singapore Toronto0-471-18636-8. Furnish, M.D., Boslough, M.B., 1996. Measuring Hugoniot, reshock and release properties of natural snow and stimulants. Sandia National Laboratory Report, SAND92-0985, UC-703. Gubler, H., 1977. Articial release of avalanches by explosives. Journal of Glaciology 19 (81), 419429. Haehnel, R.B., Shoop, S.A., 2004. A macroscale model for low density snow subjected to rapid loading. Cold Regions Science and Technology 40, 193211. Heierli, J., Gumbsch, P., Zaiser, M., 2008. Anticrack nucleation as triggering mechanism for snow slab avalanches. Science 321, 240243. Ingram, L.F., 1962. Air blast in an Arctic Environment. US Army Waterways Experiment Station, Vicksburg ,MS, Technical Report, pp. 2399. Joachim, C.E., 1967. Shock transmission through ice and snow. US Army Waterways Experiment Station, Vicksburg, MS, Technical Report, pp. 1794. Johnson, J.B., 1991. Simple model of shock-wave attenuation in snow. Journal of Glaciology 37 (127), 303312. Johnson, J.B., Solie, D.J., 1993. Shock Response of Snow. Journal of Applied Physics 73 (10), 48524861. Johnson, J.B., Solie, D.J., Barrett, S.A., 1994. The response of a seasonal snow cover to explosive loading. Annals of Glaciology 19, 4954. Kinney, G.F., Graham, K.J., 1985. Explosive shocks in air. Springer-Verlag, Berlin Heidelberg New York Tokyo3-540-15147-8. Mellor, M., 1985. Blasting and blast effects in cold regions Part 1: air blast. US Army Cold Regions Research & Engineering Laboratory Special Report, 8525.

Fig. 9. Shear stresses vs distance from the blast axis for suspended 0.9 and 1.8 kg pentolite charges. The shock wave is located 2.45 m from the blast axis in each case.

164

D.A. Miller et al. / Cold Regions Science and Technology 69 (2011) 156164 Tichota, R.G., Miller, D.A., Larson, R., Richmond, D., 2010. An experimental investigation of explosives and snowpack dynamic response. 2010 International Snow Science Workshop, Lake Tahoe, CA. Ueland, J., 1992. Effects of explosives on the mountain snowpack. 1992 International Snow Science Workshop, Breckenridge, CO. Wisotski, J., Snyer, W.H., 1966. A study of the effects of snow cover on high explosive blast parameters. University of Denver, Denver Research Institute, CO: Final Report, 2303.

Mellor, M., 1975. A review of basic snow mechanics. Proceedings from the International Snow Mechanics Symposium, Grindelwald. International Association of Hydrological Sciences Pub No 114, 251291. Mellor, M., 1973. Controlled release of avalanches by explosives. Symposium on Advances in Avalanche Technology, Reno, NV. US Forest Service Technical Report RM-3, pp. 3749. O'Keeffe, D., 1965. Height of burst curves for snow. Naval Ordnance Laboratory, White Oak, MD, Report 6424. Preece, D., Lownds, C., 2008. 3D computer simulation of bench blasting with precise delay timing. 34th Annual Conference on Explosives and Blasting Technique, New Orleans, LA.