Skip Navigation

Online Journals GSA Book Series Online Maps Field Guides e-alerts
Multi-Media Data Repository Search Subscribe Feedback Help Submit a Manuscript

ABSTRACT

Thick, slow-moving block-lava flows are associated with extrusive activity in dacitic systems, where lava-core depressurization during flow-front collapse generates devastating block-and-ash flows. Dimensional and rare thermal data collected during January 2000 for an active dacitic block flow at Santiaguito (Guatemala) provide insight into cooling and emplacement mechanisms. Flow velocity was low (12.5 m·d⁻¹), in spite of steep (10°) slopes, a result of the high viscosity (>4 × 10⁹ Pa·s) that we calculate for this flow. The flow surface consisted of a thick (1.9–3.4 m), cool (40–111 °C) crust of meter-sized, subangular blocks. Extremely effective insulation by the thick crust results in model-derived core cooling of 0.08 °C·h⁻¹. These low cooling rates make block flows the most thermally efficient of all styles of lava-flow emplacement, allowing cooling-limited flow lengths of several kilometers, in spite of low eruption rates (<0.5 m³·s⁻¹). Flow-front observations along with a plug-flow model showed that collapse from the faster-moving flow top contributed to a caterpillar-track–type advance similar to that observed at basaltic ‘a‘a flows. Forward motion also caused toothpaste-like extrusions of the flow core through the frontal crust at basal and marginal shear zones. The axial part of the flow front was thicker than the marginal zones and was oversteepened. This geometry can be explained by a higher vertical velocity gradient in the axial zone, causing more frequent and larger-volume flow-front collapses. Axial-zone collapses also penetrate farther up flow, but not sufficiently to depressurize the flow core and generate a block-and-ash flow. For such a block-and- ash flow to occur, we calculate that an increase in velocity and/or thickness (due to increased slope or topographic confinement) must occur. Whereas low surface temperatures make block flows invisible to short-wave infrared sensors, the low velocity also contributes to the stealthy behavior of these flows. Their stealthy nature, however, masks the fact that they can extend many kilometers, moving block-and-ash flow sources closer to vulnerable communities.

Keywords: block-and-ash flow, block-lava flow, cooling, emplacement, Enhanced Thematic Mapper, Santiaguito, volcanology.

INTRODUCTION Return to TOC

Block lava was defined by Macdonald (1953) as having a surface characterized by individual fragments that are “relatively smooth polyhedral blocks bounded by dihedral angles, lacking the exceedingly rough and spinose character of typical ‘a‘a.” Such flows are typically associated with basaltic andesite to dacite lava and display lower eruption temperatures with higher crystal contents, viscosities, and yield strengths than their basaltic ‘a‘a counterparts. This combination of factors is typically cited as resulting in short, thick flows with high aspect ratios and low velocities (e.g., Cas and Wright, 1987 ).

Although these features mean that block- flow inundation of down-flow communities is unlikely, block-and-ash flows derived from flow-front collapse pose an extreme hazard to the same communities (e.g., Yamamoto et al., 1993 ; Fujii and Nakada, 1999 ; Ui et al., 1999 ). The high velocity and temperature of such pyroclastic flows are such that they pose a threat to human life, being the most common cause of death due to volcanic activity in the past 400 yr and accounting for 70% of the 52250 volcano-related deaths during 1900– 1982 (Blong, 1984 ). Their destructive nature also causes extensive building and vegetation damage due to impact, collapse, racking, burial, and ignition (Blong, 1984 ). In this regard, block-and-ash flows due to gravitational collapse of the Unzen lava dome (Japan) during 1990–1995 extended >4 km from their source, causing 43 deaths and widespread destruction (Yamamoto et al., 1993 ; Fujii and Nakada, 1999 ; Ui et al., 1999 ).

Many workers have observed active ‘a‘a lava flows to understand their emplacement mechanisms (e.g., Pinkerton and Sparks, 1976 ; Peterson and Tilling, 1980 ; Rowland and Walker, 1987 ; Kilburn and Guest, 1993 ). For block flows, however, workers have described features at inactive flows and drawn inferences regarding block-flow properties and mechanics from these observations (summaries in, e.g., Macdonald, 1972 ; Cas and Wright, 1987 ; Francis, 1993 ). In addition, observations of active flows at Arenal (Costa Rica) (Borgia et al., 1983 ; Cigolini et al., 1984 ; Borgia and Linneman, 1990 ) provide insights into the emplacement mechanisms and rheology of transitional ‘a‘a-block flows. However, although thermal measurements have been made on active dacitic lava domes by using airplane- and satellite-borne thermal sensors (e.g., Friedman et al., 1981 ; Oppenheimer et al., 1993 ; Wooster et al., 2000 ), detailed ground-based thermal and observational data for an active block flow during emplacement are extremely rare. Chief among the factors that contribute to the lack of observations of active block flows are unsafe access and the hazard posed by flow-front collapse. These preclude close approach and direct measurement.

During January 2000 we used a number of noncontact field- and satellite-based remote- sensing approaches to obtain data for an active dacitic block flow at Santiaguito (Guatemala). During the course of our measurements, frequent and spatially random collapses of the flow front and marginal levees generated localized ash clouds. Although these extended just a few meters from the flow, their thermal infrared thermometer–derived temperatures of up to 185 °C showed that they would have been extremely hazardous to workers making contact measurements. Here we describe our observations of the morphology and thermal structure of the flow, where our ground-based measurements provide a rare insight into block-flow emplacement and cooling. We use our observations to (1) gain an insight into the thermal and rheological characteristics of an active block flow, (2) consider issues that determine flow-front stability and hence the probability of block-and-ash flow generation, (3) infer the manner in which such well-insulated flows cool and are emplaced, and (4) consider the problems encountered if interpreting remotely sensed thermal data for such cold, but active flows.

ACTIVITY AND THERMAL MEASUREMENTS Return to TOC

Rose (1987b) defined a three-phase history for the evolution of Santa Maria and Santiaguito. The first phase began at ca. 30000 yr B.P. and involved the construction of the Santa Maria composite cone. This phase was characterized by an eruption rate of 1.1 m³·s⁻¹, with initial eruptions of basaltic lavas followed by a gradual transition to more silicic lavas. The second phase involved the Plinian eruption of 1902 during which 8.5 km³ of dense dacite and a minor proportion of basaltic andesite were erupted (Rose, 1987b ; Williams and Self, 1983 ). The third phase has included the ongoing construction of the Santiaguito dome complex.

Activity at Santiaguito began in 1922 and has been characterized by continuous, slow (0.4 m³·s⁻¹) extrusion of dacite lava (Rose, 1972, 1987a ). Persistent effusive activity from four distinct vents (El Caliente, La Mitad, El Monje, and El Brujo) has built an 1 km³ complex of overlapping domes and block flows centered within the 1902 crater. Since 1958, activity has been characterized by an increasing prominence of block flow, as opposed to dome emplacement (Rose, 1987a ). Harris et al. (2001) have noted that the length of individual block-flow units has increased through time, a possible result of a 2 wt% decrease in the SiO₂ content of erupted products since 1970 and hence a reduction in the lava viscosity. The block flows at Santiaguito do not display well-developed polyhedral blocks, nor do the flows carry the spinose clinker typical of ‘a‘a. Instead, Santiaguito block-flow surfaces are composed of angular to rounded dacite blocks, similar in form to the rubble- like surfaces that cover the dome units within the complex.

The activity that we observed during January 2000 was fed by effusion from El Caliente, the most eastern vent. Activity at the vent consisted of small, nonexplosive exhalations of gas and ash at a rate of approximately two events per hour, a style of activity that has persisted at this vent since 1975 (Rose, 1987a ). A low active dome rose 10 m above the vent rim, from which an active block-lava flow extended southward down the flank of the dome complex. At the base of the dome pile, the flow was deflected southwest by a low ridge, to attain a total length of 2370 ± 90 m (Fig. 1 ). Ground-based observations indicate that extrusion of this flow began during July 1999; an increase in volcanic radiance recorded by the Geostationary Operational Environmental Satellite (GOES) satellite indicates that extrusion may have begun around July 14, giving a time-averaged advance rate of 12.5 ± 1.5 m/day.

To make our field-based thermal measurements we used a Raytek Raynger 3i. This unit is a hand-held, noncontact thermal infrared (8–14 μm) thermometer, with a response time of 0.7 s. At the typical measurement distances of this study (20, 1000, and 1500 m), the Raytek will measure the temperature of 0.3-, 13-, and 20-m-wide surface areas. Measurements of the vent and proximal-medial flow sections were made vertically and obliquely from helicopter passes 1000 m above and to the south of the vent on January 21. On January 23, ground-based measurements were made of the vent and proximal-medial flow sections from a ridge 1500 m to the south and of the flow front from a distance of 20 m (Fig. 1 ).

Field measurements commenced 6 h before, and continued through, the time of an overpass of the Landsat 7 satellite. Landsat 7 carries the Enhanced Thematic Mapper (ETM+), a multispectral instrument capable of providing 15–60 m spatial resolution data in eight wave bands (Fig. 2 ). Multicomponent mixture modeling of thermal data provided by ETM+ bands 4 (0.76–0.90 μm), 5 (1.55–1.75 μm), 6 (10.42–11.42 μm), and 7 (2.08–2.35 μm) allows the thermal structure of active lava surfaces to be determined (Flynn et al., 2000) . All data were corrected for atmospheric effects by using a MODTRAN tropical atmospheric model. Emissivity corrections for ETM+ bands 4, 5, and 7 were chosen by using reflectance spectra from a dacite lava sample. This approach gave an emissivity of 0.82 in all three bands. For the thermal infrared (Raytek and ETM+ band 6), we used the emissivity for andesite and rhyolite of 0.90– 0.91 given by Salisbury and D'Aria (1992) .

MORPHOLOGIC, DIMENSIONAL, AND THERMAL OBSERVATIONS Return to TOC

On the basis of our down-flow thermal observations, we divided the flow into three thermally distinguishable sections: proximal, medial-distal, and toe (Figs. 2 and 3 ). The proximal section is defined by relatively high surface temperatures (T_surf) that show a consistent decline down section. The medial-distal section is characterized by relatively low T_surf. The toe section (which includes the flow front) is defined by an increase in T_surf, which rises to values similar to those obtained in the proximal section. In addition, by exposing hotter core material, rockfall events generated by flow-margin collapse can cause localized thermal anomalies at any point along the flow.

Proximal Flow Section

The proximal section extends 700 m from the rim of the vent area southward down the steep (30°) slopes of the dome flank. Over this section the flow was 60–100 m wide and covered with a continuous cooled crust of rubble (10–100-cm-wide clasts) and larger (1–3-m-wide) blocks. At the dome base the flow was contained within an 40-m-high embankment of rubble, resting at an angle of 30°, and the surface displayed well-developed surface folds. These flow dimensions and our estimate of the flow's time-averaged velocity (1.45 ± 0.17 × 10⁻⁴ m·s⁻¹) yield an eruption rate of 0.475 ± 0.165 m³·s⁻¹. This rate is in excellent agreement with eruption rates of 0.48 ± 0.09 m³·s⁻¹ calculated following the methodology of Harris et al. (1998) .

Measured surface temperatures (T_surf) were highest across this section of the flow, where T_surf values of 76–111, 40–85, and 55–86 °C were obtained from the helicopter-, ground-, and ETM+-based measurements, respectively (Fig. 3 ). Down this section, surface temperatures showed a consistent decline at a rate of 0.04–0.09 °C·m⁻¹, which, for a velocity of 1.3–1.6 × 10⁻⁴ m·s⁻¹, converts to a surface cooling rate of 0.02–0.05 °C·h⁻¹. Following Oppenheimer (1991) we use T_surf to calculate radiative and convective heat losses (Q_rad and Q_conv) and hence the thickness of the conductive boundary layer (δ). Assuming that all heat conducted across the flow's crust (Q_cond) is lost from the surface by radiation and convection (i.e., Q_cond = Q_rad + Q_conv), δ can be calculated by rearranging Fourier's law of heat conduction (Holman, 1992 ) so that δ = k (T_core − T_surf)/Q_cond, (1) where k is thermal conductivity and T_core is the flow's core temperature (Table 1 ). Over the entire length of the flow we calculate δ in the range of 1.9–3.3 m (Fig. 4 ). This result compares with observed crust thickness at the flow front of 2–10 m. Our calculated and observed δ values for this dacitic block flow are much higher than those obtained for active ‘a‘a flows and pahoehoe flows (Table 2 ).

Following Keszthelyi (1995) we calculated down-flow cooling of the core by using the flow's heat budget. In this case, by assuming that heat gained per unit length from latent heat of crystallization is negligible and that all heat lost is accounted by Q_cond, the cooling rate in the down-flow direction (T/x) can be calculated from where A_pixel is pixel area, d and V are flow depth and velocity, respectively, ρ is lava density and C_P is lava specific heat capacity (Table 1 ). Over the proximal part of the flow we calculate core cooling rates of 57–158 °C·km⁻¹, which converts to 0.028–0.091 °C·h⁻¹ (Fig. 5 ). These cooling rates are much lower that those calculated for ‘a‘a and pahoehoe flows (Table 2 ).

Rockfall Events

The proximal and medial flow sections are separated by a distinct peak in the temperature profile derived from ETM+ data (Fig. 3 ). We assume that this peak is due to a rockfall from the flow that occurred nearly simultaneously with the ETM+ acquisition. We infer this for three reasons: (1) the high frequency of rockfall events, (2) the size and orientation of the anomaly, and (3) association with a small ash cloud.

(1) Rockfall Frequency

(2) Rockfall Size and Orientation

(3) Rockfall-Generated Ash Clouds

Single-component thermal modeling (Rothery et al., 1988 ) for the ETM+ rockfall anomaly yields temperatures of 452–805 °C. The upper range of this estimate compares with a calculated core temperature for this section of the flow of 710–780 °C (Fig. 5 ). This finding indicates that even small collapses involve crust and core material, but such localized exposure of the core seems insufficient to trigger a block-and-ash flow.

Medial-Distal Flow Section

The medial-distal section extends from the point at which the flow reaches the base of the dome and is deflected southwest by topography. This section was 1000 m long with a width of 180–210 m derived from ETM+ data. Shadows apparent on the ETM+ image indicated that a 30–90-m-wide channel zone was flanked by 30–60-m-wide levees. Given an angle of repose of 30° for the levees, we obtain a levee height of 17–35 m, consistent with an observed flow-front thickness of 18– 30 m.

Surface temperatures across this section derived from ETM+ data were low (53–64 °C) and showed a weak down-flow decline of 0.01 °C·m⁻¹ (Fig. 3 ). An identical trend of 0.01 ± 0.02 °C·m⁻¹ was obtained down four inactive flows adjacent to this section. This trend appears to be forced by an external cause rather than by lava cooling. Therefore we do not trust this trend to be volcanically significant. The cause may be increased shadowing with down-flow distance due to extension of the flows from the open slopes of the dome into shaded valleys. Most surface temperatures for the active flow, however, are elevated above the maximum obtained at any of the inactive flows by 25.7–0.5 °C. This elevation is due to heat conducted from the core of the active flow, where we calculate δ of 2.7–3.6 m (Fig. 4 ) and core cooling rates of 57–115 °C·km⁻¹ or 0.027–0.066 °C·h⁻¹ (Fig. 5 ) for this flow section.

Flow Toe

Across this 300-m-long flow section, the flow had a width of 200 m and a flow-front perimeter and height of 260 m and 18–30 m, respectively. The flow surface consisted of a blocky crust with temperatures of 34–80 °C and 56–86 °C derived from Raytek and ETM+ data, respectively. The down-flow profiles show an increase in T_surf across this flow section (Fig. 3 ) and a decrease in calculated δ to 1.9–3.2 m (Fig. 4 ). Thinner, hotter surface crusts are maintained by two processes: (1) avalanching of surface crust at the flow-front perimeter to expose hotter surfaces and (2) spreading and thinning the flow across the toe section. Elevated T_surf across this section results in an increase in Q_cond and hence in the calculated rate of core cooling to 60–158 °C·km⁻¹ or 0.028–0.091 °C·h⁻¹ over this section (Fig. 5 ).

Flow-Front Morphology and Processes

Laterally, the flow front could be split into three segments on the basis of its morphologic and thermal characteristics: (1) an eastern marginal segment, (2) a medial-axial segment, and (3) a western marginal segment (Fig. 6A ). Because access to the western marginal segment was impossible, here we describe our observations and interpretations of the eastern and medial segments.

Marginal Segment of the Flow Front

This structure was maintained by flow-front collapse (Fig. 7B ). Collapses occurred once or twice per minute, affecting 2–5-m-wide sections of the flow front and generating small ash clouds at Raytek-derived temperatures of up to 185 °C. The source of all collapses was the crust zone and upper parts of the midzone. Higher temperatures at the midzone were maintained by repeated removal of surface material by the passage of collapse-fed avalanches; a temperature of 154 °C was obtained from a freshly exposed midzone surface immediately following a collapse. The similarity between the temperatures of the crust and talus zones results from the fact that the crust zone supplies material to the talus zone by collapse. Inclusion in the talus deposit of hotter material picked up from the midzone explains the slightly higher temperatures of the talus zone.

Clusters of massive, pink (oxidized) outcrops were common at the eastern and western edges of the marginal segment. These were associated with the most-forward-advanced parts of the flow front (Fig. 6 ). Each outcrop was 3–5 m wide, roughly circular to oval in cross section, showed crude radial jointing, and protruded 1–2 m outward from the flow front. During our observation period, one outcrop was truncated by a rockfall to reveal rooting of the extrusion back into the flow. All outcrops had higher temperatures (elevated by 6–96 °C) than the surrounding blocky crust (Fig. 6C ). We interpret these features to be toothpaste-like extrusions of the flow core around axes of advance.

Medial-Axial Segment of the Flow Front

MECHANICS OF BLOCK-FLOW ADVANCE Return to TOC

Our observations support a model whereby forward motion is related to two simultaneous mechanisms: (1) flow-front oversteepening and collapse to cause caterpillar-track–type advance and pushing aside of frontal crust and (2) forward extrusion of the flow core through the frontal crust. To assess the validity of these mechanisms we have applied a model for a Bingham fluid exhibiting laminar plug flow (see Appendix). Horizontally, this model gives an 59-m-wide axial zone of plug flow (Fig. 9, A–C ), consistent with a measured axial- zone width of 50 m. For this 30-m-thick axial zone we obtain plug and shear-zone thicknesses of 18 and 12 m, respectively, from the same model (Fig. 9, D and E ).

Vertically, the plug dominates the top half of the flow, creating an upper high-velocity collapsing zone and a lower-velocity basal shear zone (Fig. 9, D and E ). In this case, higher velocities at the flow top will result in oversteepening and subsequent instability and collapse of the upper zone. In the resulting caterpillar-track–type forward motion, material falling from the flow front is overridden by the advancing flow to contribute to a basal crust. This feature is typical of basaltic ‘a‘a flow emplacement (Wentworth and Macdonald, 1953 ; Macdonald, 1953, 1972 ; Cas and Wright, 1987 ; Kilburn and Guest, 1993 ) and also forms at transitional andesitic ‘a‘a-block flows, where identical flow-front structures develop (Cigolini et al., 1984 ; Borgia et al., 1983 ).

The extrusion features observed at the marginal and axial segments also support a flow- emplacement model in which forward motion results in multiple toothpaste-like extrusions of the core through the frontal crust. Such toothpaste-like extrusions of viscous lava have been observed to pass through the crust at the toe of basaltic ‘a‘a flows in Hawai'i (Macdonald, 1953 ). Flow-front core extrusion has also been noted at basaltic andesite ‘a‘a flows fronts on Etna (Kilburn and Guest, 1993 ) and andesitic ‘a‘a-block flows at Arenal (Linneman and Borgia, 1993 ).

In the case considered here, extrusions are located at the primary and secondary axes of advance. Both of these axes occur where the flow is extending within or toward ravines (Fig. 6A ). The resulting increase in slope, as well as flow confinement, at these locations will cause localized velocity increases, thus promoting shearing (Fig. 9A ). Extrusions are also confined to the lower half of the flow (Fig. 6B ), again a zone of high shear stress (Fig. 9E ). Extrusion therefore appears to be due to squeeze out between shear planes. An extrusion cluster at the edge of, and extending obliquely from, the primary advance axis indicates that these structures may have their roots in a zone of divergent flow at the margin of the axial stream (Fig. 6 ). Such divergent flows at the margins of a central zone of plug flow have previously been observed at channelized ‘a‘a flows at Etna (Booth and Self, 1973 ).

FLOW-FRONT STABILITY Return to TOC

Block-and-ash flows can be generated by flow-front collapse followed by pore depressurization, vesiculation, and pulverization (Rose et al., 1976 ; Mellors et al., 1988 ; Fink and Manley, 1989 ; Fink, 1993 ; Fink and Kieffer, 1993 ; Fujii and Nakada, 1999 ; Ui et al., 1999 ). Once under way, a dense gravitational current avalanches along topographic lows, slowing with decreasing slope. Convection and fluidization, due to air heating and expansion and release of pore-stored gas during rock collisions, generates a fluidized, turbulent ash component that overrides and outruns the gravitational current (Mellors et al., 1988 ; Fink and Kieffer, 1993 ; Fujii and Nakada, 1999 ). We consider two interrelated trigger mechanisms for such flow-front collapse and block-and-ash flow generation: (1) gravitational flow-front collapse due to high underlying slopes and flow-front oversteepening and (2) exposure of overpressurized and/or volatile-rich core zones causing explosive decompression.

As shown by Voight and Elsworth (2000) , increasing the angle of the failure plane (α) increases the forces acting to destabilize the flow, hence promoting collapse. Voight and Elsworth (2000) described stability in terms of the ratio of resisting to disturbing forces (F_s, where F_s 1 implies failure) and showed that such conditions are typically associated with α > 20°. For the Santiaguito case, we calculate F_s of 0.6 and 1.7 for flow fronts advancing over the 30° and 10° slopes characteristic of the proximal and distal flow sections, respectively. This result is consistent with our observation of seven small (300-m-long) incandescent avalanches from the vent and proximal flow region during a 1-h-long period, but none from the distal section.

Flow-front oversteepening will also promote collapse. At Unzen (Japan), lower velocities at the flow base than in the upper zone caused oversteepening, with gravitational failure occurring where the overhanging angle reached 12° (Yamamoto et al., 1993 ). This condition can be simulated by using the plug- flow model (see Appendix). For Unzen, the measured flow thickness (70 m) and plug velocity (4.2–9.2 × 10⁻⁴ m·s⁻¹), with a plug thickness of 20 m (from Yamamoto et al., 1993 ; Fig. 5 ), gives a simulated basal zone velocity () of 0.67–1.48 × 10⁻⁴ m·s⁻¹. This value compares with measured basal of 0.47–1.47 × 10⁻⁴ m·s⁻¹ (Yamamoto et al., 1993 ). Given the critical oversteepening angle (σ_crit) at which collapse occurs and flow depth (d), the up-flow distance that collapse will effect (L_crit) can be calculated from [d tan(σ_crit)], and the time taken for σ_crit to be reached (t_crit) from [d/]. For Unzen, this gives L_crit of 15 m and t_crit of 28–62 h, which compares with a block-and-ash flow temporal frequency of one every 51–76 h (from Yamamoto et al., 1993 ).

If we apply the same model-based approach to the Santiaguito flow, we gain an insight into why the axial flow front is oversteepened. For the 30-m-thick axial plug flow zone, a 12-m- high shear zone (d_shear) has a velocity gradient (Δ/Δd_shear) of 1.2 × 10⁻⁵ s⁻¹ (Fig. 9F ). If collapse occurs at σ_crit = 12°, then we obtain L_crit and t_crit values of 6.3 m and 12 h, respectively. In the marginal zone, the shear zone effectively comprises the entire 18 m flow thickness, and plug flow is absent so that d_shear = d (Fig. 9F ). Across this zone, lower velocities (typically _max/2, Fig. 9A ) and higher d_shear values result in lower Δ/Δd_shear and L_crit values and higher t_crit values, these being 0.4 × 10⁻⁵ s⁻¹, 3.8 m, and 15 h, respectively. Therefore in the axial zone, higher-velocity gradients cause more frequent and more voluminous collapses than in the marginal zone, favoring a steeper flow front.

The difference in between the axis and margin suggests that the axis should overshoot the margins. That a greater volume of the axial flow front is being destroyed by collapse, however, serves to dampen the velocity difference. For example, we calculate that over a 60 h period, five collapses will occur at the axial zone and four will occur at the marginal zone. The resulting difference between advance (due to forward motion) and retreat (due to collapse) will be 0.5 and 0.46 m at the axis and margin, respectively.

In the two cases considered (Unzen and Santiaguito), flow-front collapse due to oversteepening generated block-and-ash flows only in the Unzen case. To trigger block-and- ash flow, the collapse must have a sufficient depressurization effect on the flow core. Ui et al. (1999) suggested that block-and-ash flows may be generated where flow-base pore pressure exceeds the lava tensile strength. Following their model, decreased tensile strength due to higher strain rate and shear stress at the flow base (Fig. 9E ) will favor a situation where pore pressure can exceed the lava tensile strength during crack propagation or depressurization to cause explosive fragmentation. Fink and Manley (1989) also proposed a model where volatile concentration due to the migration and concentration of water vapor released during crystal growth and microfracturing increases the explosive hazard. During both cases, depressurization by flow-front collapse will trigger explosive decompression. We note that L_crit calculated for Unzen and Santiaguito are 14.9 m and 3.8–6.4 m, respectively. Lower L_crit at Santiaguito coupled with the lack of block-and-ash flow activity indicates that collapse involving low L_crit values removes an insufficient proportion of the flow front to intersect overpressurized zones and is therefore less likely to generate block-and-ash flow.

Following these considerations we propose an integrated model for flow-front stability at Santiaguito. In this model, block-and-ash flows will most likely occur during advance down the steep slopes of the dome complex and/or during flow-front oversteepening at thick flow fronts exhibiting well-developed plug flow. Increases in the velocity gradient across the shear zone with increased flow thickness and velocity not only contribute to oversteepening, but also increase collapse frequency and the distance back from the flow front affected by collapse. The presence of these factors increases the likelihood that collapse will penetrate overpressurized and/or volatile-rich core zones behind the flow front, hence triggering block-and-ash flow.

Features of the September 16, 1973, block- and-ash flow at Santiaguito are consistent with this scenario. This event was triggered by collapse of a thick block-lava flow front that caused flow-core vesiculation such that vesicular particles constituted up to 50% of the block-and-ash flow deposit (Rose et al., 1976 ). Collapse occurred at the break of slope between the constructional topography of the dome apron and surrounding erosional topography (Sam Bonis, 2000, personal commun.). The increase in slope at this point would have caused a decrease in flow-front stability, as well as an increase in flow velocity. An aerial photograph (Fig. 8 in Rose et al., 1976 ) also shows that the flow was entering a narrow box canyon, as was the case for the axial zone considered here. Increased velocity caused by extension into a confining canyon and onto steeper slopes would have resulted in increased Δ/Δd_shear, decreased t_crit, and, more important, increased L_crit to levels sufficient to intersect the overpressurized, volatile-rich core zones.

RHEOLOGICAL PROPERTIES Return to TOC

We have used the plug-flow model detailed in the Appendix to simulate the rheological properties of this flow. Owing to the extremely low velocity of this flow, calculated strain rates are low, giving 1.2 ± 0.5 × 10⁻⁵ s⁻¹ and 2.5 ± 2.2 × 10⁻⁶ s⁻¹ in the z and y flow directions, respectively (Fig. 9 ). Calculated shear stress and viscosities (Appendix) are, however, much higher at 6.7–8.5 × 10⁴ N·m⁻² and 0.4–2.6 × 10¹⁰ Pa·s, respectively. These viscosities are consistent with the 10⁸–10⁹ Pa·s value given for rhyolite and andesite at 800– 1000 °C by Murase and McBirney (1973) and with 5 × 10¹¹ Pa·s calculated following Shaw (1969) using cooling of 100 °C and a liquidus viscosity of 1000 Pa·s (Murase and McBirney, 1973 ). These high model-derived values explain the extremely low velocities encountered at this block flow, in spite of the 10°–30° slopes that the flow is advancing over. To place these values in context, we can compare this dacitic case with a basaltic case. Typical basaltic shear stresses and viscosities are much lower: 220–370 N·m⁻² and 9400–10⁵ Pa·s, respectively (Pinkerton and Sparks, 1978 ; Moore, 1987 ). Given these lower values, the same flow would advance at speeds of 5–59 m·s⁻¹ if it were basaltic (calculated following Dragoni, 1993 ).

DISCUSSION AND CONCLUSIONS Return to TOC

Flow Insulation and Hazard Implications

Insulation by a thick crust means that block flows can extend surprising distances in spite of low eruption rates and forward velocities. Insulation serves to reduce surface temperature and heat loss so that core cooling rates are extremely low (Table 2 ). Given these cooling rates and a difference between liquidus and solidus of 150–200 °C (Archambault and Tanguy, 1976 ), we calculate that the block- flow core of this Santiaguito flow will take 89–278 days to cool to solidus. Assuming that flow movement will continue as long as the core temperature is above the solidus gives a cooling-limited flow length of 1.3–2.9 km, in spite of an eruption rate and velocity of 0.475 m³·s⁻¹ and 0.000145 m·s⁻¹, respectively. Increasing flow depth and velocity retards the cooling rate (equation 2) such that, for a 70-m-thick flow with of 4.2–9.2 × 10⁻⁴ m·s⁻¹, the cooling rate is 0.004–0.020 °C·m⁻¹ or 0.015–0.030 °C·h⁻¹. This range of cooling rates yields a cooling-limited length of 7–50 km.

These low cooling rates show that block flows are one of the most insulated forms of lava-flow emplacement, where the thick, cool, crusted carapace is an extremely efficient insulator. In Table 2 we have compared the heat- loss characteristics of a variety of flow types. Channelized basaltic lava flows are thermally the most inefficient flow type, where thin, hot, discontinuous crusts result in high heat-loss rates and hence rapid core cooling. As a result, channelized flows are only able to extend tens of kilometers by virtue of rapid emplacement, i.e., high velocities and eruption rates (Keszthelyi and Self, 1998 ). As crusts become thicker, cooler, and more coherent, heat loss and hence flow cooling decreases (Table 2 ). Flow confined to a lava tube is thus a thermally efficient flow type, allowing even, low- effusion-rate flow to extend tens to hundreds of kilometers before cooling forces the flow to halt (Keszthelyi, 1995 ; Keszthelyi and Self, 1998 ). Table 2 shows that block flows have many thermal similarities with tube-confined flow, where the thick, cool crust insulates the flow, causing low heat losses and the lowest cooling rates for all flow types considered. As a result, a block flow has the capability of extending many kilometers at extremely low effusion rates and velocities before cooling forces stagnation. This conclusion is in agreement with Manley (1992, p. 27) who stated that “a common misconception about rhyolite lava flows is that they cannot advance far from their vents and are constrained to be small due to their high viscosities.” Manley's (1992) model-based results and field studies of Obsidian Dome (California) show that slow cooling of 100–300-m-thick units mean that such flows could remain active for decades, even though advancing at extremely low velocities (0.59–2.5 km·yr⁻¹).

The potential of a block flow to extend many kilometers has serious implications for the hazard posed to down-flow communities. As the block-flow front moves away from the vent, so the source of block-and-ash flows due to flow-front collapse extends closer to vulnerable communities. In the case of Santiaguito, the September 1973 block-and-ash flow issued from the lobe of a thick block flow 2 km from the El Brujo vent (Rose et al., 1976 ). The block-and-ash flow extended a further 3 km down valley, where the high temperature and “hurricane force” of the flow completely destroyed all vegetation and left a deposit that was up to 5 m thick and contained 4-m-diameter blocks (Rose et al., 1976 ). A block flow advancing 6 km from the El Caliente vent would place the flow front in the vicinity of the towns of El Palmar and San Felipe and within a region of extensive banana and coffee plantations. Such a flow would also move the block-and-ash flow source to within 10–15 km of the major population centers of San Sebastian and Retalhuleu, as well as the the Pan- American Highway. Given block-and-ash flow velocities of 15–25 m·s⁻¹ (Yamamoto et al., 1993 ), these communities would be affected within 7–17 min of a flow-front collapse. Such an event would have devastating effects. In this regard, the 1929 block-and-ash flow at Santiaguito traveled 10 km from the El Caliente vent, devastated several villages and plantations (including the region around El Palmar), and resulted in hundreds, possibly thousands, of fatalities (Rose, 1987a ; Simkin and Siebert, 1994 ).

Satellite-Based Thermal Detection

Insulated emplacement is strikingly evident from the extremely low surface temperatures obtained for this active block flow. Along the entire flow length, our data based on helicopter, ground, and ETM+ measurements gave T_surf of 76–111 °C, 34–85 °C, and 53–86 °C, respectively. Nighttime observations showed that there were no incandescent cracks anywhere on the flow. Extremely low velocities and a lack of thermal renewal of the flow surface means that the surface crust can exist for tens to hundreds of days, thus allowing time for significant cooling. Figures 3 and 4 show that at 750 m from the vent, the surface had cooled to a relatively stable 65 °C, and the crust had thickened to 3.3 m. Given the velocity of this flow, this distance translates to a flow-surface age of 50–70 days.

With the exception of the vent region, low surface temperatures mean that most of the flow is invisible in the ETM+ short-wave infrared (SWIR) bands (Fig. 2 ). To emit detectable thermal radiation in the SWIR bands of ETM+, surfaces of >900 m² in area must attain temperatures of >120 °C (Flynn et al., 2001) . The lack of incandescent cracks and the low crust temperatures mean that the flow surface has failed to attain this level and was thus invisible in the SWIR bands (Fig. 2 ). Without ground validation, the rockfall half way down the flow could easily have been mistaken for the thermal signature from a collapsing flow front. A further 1.3 km down flow from the rockfall, the flow front itself is in fact an extremely subtle thermal anomaly in the SWIR data (Fig. 2 ). Such an anomaly would have been extremely difficult to identify with confidence in the SWIR data without the insights gained from our field work.

In contrast, surface temperatures derived from the thermal infrared (TIR) are elevated above those encountered at surrounding inactive flows. This is especially true across the distal section, as well as at the rockfall location and flow toe (Fig. 10 ). However, the magnitude of the thermal anomaly over other sections of the flow is not large. This is especially true of the medial-distal flow sections in the January 2000 image. In this case, across the section between 1100 and 2100 m from the vent, band 6–derived T_surf values are in the range 53–65 °C. This range compares with 40–58 °C at nearby inactive flows (Fig. 10 ) and makes this section of the flow difficult to distinguish from the solar-heated background (Fig. 2B ).

Other Thematic Mapper (TM) images of block flows at Santiaguito acquired on October 25, 1987, and February 14, 1988, show identical thermal structures. In both cases, the active, 3.6-km-long, block flow, although evident as a low-magnitude TIR anomaly, is invisible in the SWIR data except for a localized thermal anomaly at the vent and a subtle thermal anomaly at the flow toe (Fig. 2, C and D ). The recurrence of these features leads us to suggest that low-magnitude TIR anomalies with subtle SWIR anomalies at their distal ends are diagnostic of block-flow activity in satellite thermal data. This correlation also argues for the need to examine TIR as well as SWIR data when locating and analyzing active block flows. Such data are currently available from Landsat ETM+ and the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) flown on the Terra satellite (Yamaguchi et al., 1998 ; Ramsey et al., 2001 ).

Thermal Stealth: A Feature of Block-flow Emplacement

During May and June 1987, block flows at Santiaguito extended 2.5 and 2.9 km, respectively (SEAN, 1988a, 1988b ). A TM image acquired during October 1987 reveals extension to 3.6 km (Fig. 1 ), giving a time-averaged velocity and TM-derived eruption rate of 5– 13 m/d and 0.72 ± 0.08 m³·s⁻¹ (Harris et al., 2001) . These low rates of advance and eruption rate are comparable with those obtained for the January 2000 block flow, as are the image-derived T_surf of 33–79 °C and calculated δ, δT/δx, and δT/δt of 2.2–4.9 m, 0.04–0.14 °C·m⁻¹ and 0.012–0.082 °C·h⁻¹, respectively (compare with Table 2 ). We therefore suggest that such block flows represent a form of insulated emplacement, where thick crusts result in low core cooling rates and surface temperatures.

Block flows thus exhibit stealthy behavior, their activity and advance disguised by imperceptibly slow forward motion and extremely low surface temperatures. Johnson et al. (1995) used the term “stealth plume” to describe gaseous SO₂ plumes on Io that would have been invisible to the remote-sensing instruments on the Voyager spacecraft. In the same way, Santiaguito's block flows are mostly invisible to the SWIR bands of terrestrial thermal remote-sensing instruments. In this case, the absence of strong SWIR emission typically associated with active lava flows, lakes, and domes (e.g., Rothery et al., 1988 ; Oppenheimer et al., 1993 ; Wooster and Rothery, 1997 ) would have caused this thermally stealthy flow to be overlooked.

APPENDIX Return to TOC

We calculate rheological parameters assuming that the lava behaves as a Bingham fluid and exhibits laminar flow. In this case, plug flow will occur where “the central part of the flow behaves as a relatively rigid plug, carried along by the flow as if it were a solid body” (Peterson and Tilling, 1980 ). Across the plug, velocity is constant and thus strain rates are zero (Gauthier, 1973 ; Williams and McBirney, 1979 ; Moore, 1987 ). Following Cigolini et al. (1984) , flow velocity at depth z (_xz) is calculated (within the limits _xz = 0 at z = d and _xz = _max at z = h_p) from and, for z = h_p, where , ρ, g, α, d, h_p, z, and _max are lava viscosity, density, acceleration due to gravity, slope, flow depth, plug-flow height, depth on the z-axis (downward and perpendicular to the flow direction), and maximum forward velocity, respectively. Strain rate at depth z (_z) is then calculated (according to Cigolini et al., 1984 ) by using the velocity gradient (d_xz/dz), where _z = d_xz/dz, (A3) and shear stress (τ_z) is calculated from τ_z = ρgsin(α)z. (A4) Following, for example, Holman (1992) , we calculate the flow-velocity profile in the cross-flow (y) direction (_xy) within the limits _xy = 0 at r_y = R and _xy = _max at r_y = r_p (r_p being the plug-flow radius) from where _f, R, and r_y are dynamic viscosity, channel radius, and radius at cross-flow position y, respectively. Pressure decrease in the down-flow direction in a noncircular duct (Pitts and Sissom, 1997 ) (dp/ dx) is obtained from where ƒ, D_h, V, and g_c are the friction factor, flow hydraulic diameter, mean velocity (_max/2), and proportionality constant (1.0), respectively. The hydraulic diameter is defined by D_h 4A_cross/P, where A_cross and P are flow cross-sectional area and wetted perimeter, respectively (Heslop et al., 1989 ; Pitts and Sissom, 1997 ). Friction factor is obtained from ƒ = K/Re (Heslop et al., 1989 ), where K is a constant dependent on channel shape (where we use K = 30 because we assume laminar flow in a rough- bottomed channel) and Re is the Reynolds number (Pitts and Sissom, 1997 ; Heslop et al., 1989 ), In the cross-flow direction, strain rate (_y) = d_xy/ dy, and we calculate shear stress as a function of flow depth and width (τ_zy) from (Cas and Wright, 1987 ) τ_zy = τ₀ + _zy(_y)ⁿ, (A8) where τ₀ is the stress required to initiate flow (and n = 1 for a Bingham substance (Cas and Wright, 1987 ). Viscosity at depth z and cross-flow distance y (_zy) is calculated from (Cigolini et al., 1984 ) where β has a value of 3 for broad flows (Cas and Wright, 1987 ; Chester et al., 1985 ).

ACKNOWLEDGMENTS

Field work was funded by a National Science Foundation grant and subsequent work was supported by National Aeronautics and Space Administration Landsat 7 grant NAG5-3951. The final form of this manuscript benefited from reviews by Mike Ramsey, Barry Cameron and Eugene Smith.

REFERENCES CITED Return to TOC