From 24 August 1989 until 18 June 1991, Mount St. Helens produced at least 28 shallow, explosion-like seismic events with signatures similar to those produced by gas explosions on the dome during the mid 1980s. At least six were accompanied by violent emission of non-juvenile tephra, ejection of blocks of rock nearly 1 km from the vent, and avalanching of debris off the north side of the dome. Four produced no emission of tephra, and the remainder occurred when the crater could not be observed (although later crater visits found no new tephra deposits). All six confirmed emissions and most (although not all) other seismic events took place hours to days after storms. Over time periods ranging from 2 to 15 days, statistical goodness of fit tests give a probability of 1x10^-2 to 1x10^-5 that the precipitation prior to emissions was fortuitously greater than normal. Similar tests for precipitation prior to all 28 events give probabilities of 3x10^-2 to 1x10^-3. The short delay between storms and emissions suggests that the events that follow storms originate at very shallow depth, probably within the dome itself. Although the exact causal mechanism is not known, it is speculated that slope instability or accelerated growth of cooling fractures following storms may have released gas trapped within or at the base of the dome. Of the events that did not follow storms, three were located seismically between 1.4 and 2.1 km depth. Thus the events as a group are not confined to shallow depth and probably record gas transport from a deeper, magmatic source. A deep source is also suggested by an unprecedented, 4-yr-long increase in background seismicity at 2- to 9-km depth that peaked in 1989 and 1990. Calculations indicate that the energy content of gas exsolved from the crystallizing conduit was probably sufficient to power the emissions.
In the spring of 1980, Mount St. Helens in southwestern Washington State (Fig. 1), began a decade-long period of eruptive activity that has been among the most extensively documented of any eruptive sequence (for example, Lipman and Mullineaux, 1981; Pallister and others, 1992). The activity included six major Plinian eruptions in 1980, followed by six years of dome growth from October, 1980 to October 1986. From October, 1986 until mid-1989, Mount St. Helens was relatively quiet. On 24 August 1989, however, Mount St. Helens was rocked again by a shallow seismic event strikingly similar to signals of the early- and mid-1980s that had accompanied explosive emissions of steam and ash (Jonientz-Trisler and others 1991). From August 1989 until June 1991 there followed at least 27 more shallow, explosion-like seismic events of this type. At least six produced violent emissions of gas (mostly steam) and tephra, showers of ballistically ejected blocks that covered the crater floor, and small debris flows that traveled down the Toutle River Valley.
[Graphic,19K,GIF]
Figure 1:
Map of the Mount St. Helens area.
Small gas explosions of this type have in the past not been especially well documented or their triggering mechanisms well understood, but recent deaths at Galeras Volcano, Colombia, and Guagua Pichincha, Ecuador from eruptions of this type (Kerr, 1993) demonstrate the need to understand and predict them. The emissions at Mount St. Helens show important similarities to the eruptions at Galeras and Guagua Pichincha in two respects: as at Galeras (Kerr, 1993), volatile exsolution from magma is a plausible source of gas for these events; as at Guagua Pichincha (Smithsonian Institution, 1993), the emissions at Mount St. Helens are associated with periods of wet weather. Although associations between gas or steam explosions and rainfall have been mentioned in past studies (e.g. McBirney, 1955; Murray, 1980), this is the first case known to this author where a such correlation has been statistically documented. This paper describes these events and interprets their cause.
In contrast to the magmatic eruptions of the early and mid-1980s (Chadwick and others 1983; Dzurisin and others 1983; Swanson and others 1983), the shallow, explosion-like seismic events of August 1989, and later had no apparent short-term precursors. Seismicity at 2- to 9-km depth below the crater had been increasing gradually since late 1987 (Moran, in press), but did not deviate significantly from this trend in the days or weeks preceding any of the events. Tiltmeters and repeated geodetic surveys also recorded no uplift of the crater floor or other deformation that could be intrusion-related.
The onsets to the seismic events ranged from emergent to abrupt. Within the crater, each was characterized by continuous, high-amplitude shaking of medium frequency (for example 5-7 Hz for the 24 August 1989 event) that lasted from less than a minute to several hours (Fig. 2; Table 1). -- (Web note: Figure 2 not yet available) -- The signatures are quite different from discrete, seconds-long monochromatic ~0.5-Hz signals inferred to be caused by gas transport at Redoubt Volcano (Chouet and others, in press). The shaking was followed for minutes to hours by many small high-frequency earthquakes, at the rate of a few per minute to one every few minutes. One emission event, on 20 December 1990, was followed for more than two weeks by several hundred small-magnitude earthquakes, located below or just north of the dome. Those occurring in the first few days were all less than 3 km below the base of the dome, but later ones ranged to more than 5 km (S. C. Moran, 1992, written commun.).
[Table,14K,GIF]
Table 1:
Summary of shallow, explosion-like seismic events and gas-and-tephra emissions,
August 1989-June 1991.
The long duration and high amplitude of the signal, and the sequence of small earthquakes in the latter part of the record distinguished shallow, explosion-like seismicity from the rumbling of rock fall, the buffeting of the stations by wind, and the sharp arrivals and rapid exponential decay (within seconds) of earthquakes that typify the Mount St. Helens magmatic system (Jonientz-Trisler and others 1991). Strong shaking was generally confined to the crater, suggesting that the source was shallow.
Blasts from five of the six confirmed emissions issued in a northward direction from a vent on the north shoulder of the lava dome (NV, Fig. 3). This area overlies a normal fault scarp on the crater floor that channeled gas to the surface (as shown, for example, in Fig. 2 of Chadwick and Swanson, 1989) (Web note: not available) before it was covered by lava and rock avalanche debris. One small event (on 20 December 1990) issued from a point on the northeast shoulder of the dome (EV, Fig. 3), and one (on 5 February 1991; Fig. 4) appeared to have issued from both vents simultaneously. The largest blasts ballistically ejected dacite blocks 0.5-1.0 m in diameter (b, Fig. 4) up to 1 km north of the vent, suggesting ejection velocities of several tens of meters per second to more than 100 m/s based on ballistic relations described in Mastin (1991). Average velocities were probably somewhat less (perhaps 40-80 m/s). Accompanying or following these blasts were small ash-cloud surges (producing deposit ac, Fig. 4) and poorly inflated lithic gravity-driven flows that poured off the dome. Traces of ash were occasionally blown by high winds as far as 140 km from the mountain, and white steam plumes were reported by airline pilots as high as 5.5 km above the dome (perhaps an overestimate). Mixtures of ejecta and melted snow frequently mobilized into small debris flows (df, Fig. 4) that propagated northward out of the crater.
[Graphic,26K,GIF]
Figure 3:
Map of the Mount St. Helens crater (from 1988) showing the locations
of vents, seismic stations, geodetic stations, and the Rampart fault scarp.
NV, north vent;
EV, east vent; A, geodetic station Ansel; Am, Amos 3; H, Hoser;
Ha, Harold; L, Luther;
S, Squab; T, Tiki; Tk, Tinker. Contour elevations in meters.
The deposits consist of lithic dacite from the dome, with a small amount (<1%) of dark, glassy, high-silica andesite or low-silica dacite that may represent the early (1981) interior of the dome or the conduit (Pallister and others 1992). No debris from below the base of the dome has been found, although a significant fraction of debris (>10%) is unidentified and may include sub-domal fragments. There is no evidence of juvenile glass in any debris.
In 1990, some workers (J. Costa, R. LaHusen) at the Cascades Volcano Observatory observed a coincidence in timing between emissions and major storms. To see if this coincidence was significant, data from three tipping-bucket-type rain gages near Mount St. Helens were analyzed for the period 16 August 1989 through 30 June 1991. These gages register rainfall in 0.01in. (0.025cm) increments when water fills and tips a counterbalanced cylinder connected to an electrical switch. Their primary advantage is their good time resolution because data are transmitted every 15 min. Their primary disadvantage is that, if precipitation is received as snow, they tend to record low values (since snow does not flow readily into the measuring cylinder) or delayed values (if water flows into the cylinder only after the snow melts).
The rain gage at Spirit Lake (Fig. 1) was used as the standard for this analysis and was checked against gages at Castle Lake (Fig. 1) and Muddy River, 20 km southeast of the crater. Nearly all storms recorded at Muddy River (365 m elevation) or Castle Lake (795 m) were also recorded at Spirit Lake (1,058 m) in spite of the elevation difference, suggesting either that most precipitation at Spirit Lake was rain or that the falling snow melted quickly enough to be recorded.
The timing of storms at Spirit Lake, and of shallow, explosion-like seismic events, is shown in Fig. 5. -- (Web note: Figure 5 not yet available) -- As suspected, all six confirmed emission events closely followed major storms. Most occurred only a few hours after the end of precipitation. The longest delay between the end of a storm and the onset of an emission is 68 hr, prior to the 7 December 1989 event. The correlation between seismic events that were not confirmably associated with emission of gas or tephra is less clear. Those that occurred in May, June, October, and November of 1990, generally followed storms, but those in August 1989 and September 1990 did not.
To estimate the likelihood that the correlation between precipitation and emissions is fortuitous, I use two types of probability tests. The first involves calculating the simple binomial probability that a given number of emission events fortuitously followed a given period of precipitation. As an example, consider the fact that five of six confirmed emissions were preceded by at least 4 cm of precipitation during a 96-hr period. What is the probability that this would happen, assuming that emission events occur randomly with respect to storms? Within the time period from 16 August 1989 until 30 June 1991 (a total of 683 days), 100 days (y in Table 2) were preceded by four days with at least 4 cm of total precipitation. Given the occurrence of an emission event, the probability (p) that it would take place on one of these days is 100/683 = .1464. The probability that five of the six events would take place on such days is given by the equation:
P(r;n;p) = \F(p^r(1-p)^(n-r)n!,r! (n-r)!) = \F((.1464)^5(.8536)^1 (6!),(5!)(1!)) = .0003
where r is the number of events that took place after a 4-day period of >4 cm of precipitation, and n is the total number of events.
[Table,13K,GIF]
Table 2:
Probability values for the fortuitous occurrence of shallow, explosion-like seismicity
following a given amount of precipitation in a given time period.
Table 2 gives the results of a similar calculations for time periods ranging from 4 to 15 days and for precipitation ranging from 1 to 15 cm. Groups include (1) all 28 shallow, explosion-like seismic events; (2) confirmed emissions; (3) events that confirmably did not produce emissions; and (4) events where concurrent observations could not confirm the presence or absence of an emission, but where later visits found no new tephra ("remainder of events"). For the "all events" category and for the confirmed emissions, probabilities are frequently less than 0.01 and occasionally less than 0.001 that their timing relative to storms was fortuitous. Confirmed non-emissions give consistently high probabilities, whereas for the "remainder of events," probabilities are frequently less than 0.05 and occasionally less than 0.01.
Binomial probability calculations are intuitively understandable, but they suffer the disadvantage that one must arbitrarily choose a value of precipitation and calculate a probability just for that chosen value. It seems more appropriate to ask, without assigning arbitrary thresholds, whether the overall amount of precipitation preceding these events differs significantly from the amount that one would normally expect in a similar time period. The test employed to address such a question is known as the one- sided, one-sample Kolmogorov test for goodness of fit (Conover, 1971). The Kolmogorov test is similar to the chi squared test in that it involves examining a sample of some population of data (in this case, precipitation preceding emissions) in order to test the hypothesis that the distribution of this sample follows some specified distribution (in this case, the normal for precipitation at this location). Our "null hypothesis" in this case is that the precipitation preceding emissions was not significantly greater than normal. For this study, the Kolmogorov test is more appropriate than the chi squared test for two reasons. First, the Kolmogorov test does not assume that the data under examination follow a normal Gaussian distribution. This is especially important, because the distribution curve for precipitation over a given time period is not Gaussian. Second, the Kolmogorov test gives exact or conservative probabilities even for small sample populations, whereas the chi squared test assumes that the sample population is large enough to provide a good approximation of the test statistic.
The test is utilized as follows. First, an empirical distribution curve is constructed (the light gage solid line in Fig. 6) for precipitation in a given time period (in this case, 96 hr) preceding confirmed emissions. This curve gives the fraction of all emissions (on the y axis) that were preceded by precipitation of a given amount or less (on the x axis). It is compared with a similar curve representing 4-day precipitation preceding each of the 683 days from 16 August 1989 to 30 June 1991 (the heavy solid line in Fig. 6). If the timing of emissions were independent of the precipitation that preceded them, then the two curves should approximately follow each other. The curve for emissions in fact lies well below that for normal precipitation. To determine the probability (a) that the difference in these curves is due to random sampling, we find the largest vertical difference between the two curves where the normal precipitation curve is above the curve for precipitation preceding emissions (K-S, Fig. 6). Using this value (.8501, known as the Kolmogorov- Smirnov Statistic) and the size of the sample population (6 emissions), equations in Birnbaum and Tingey (1951) give the probability as about 1x10^-5.
[Graphic,12K,GIF]
Figure 6:
Cumulative distribution of 96-hr precipitation values recorded at Spirit
Lake: for all precipitation recorded, 8/16/89-6/30/91 (heavy solid line); for precipitation
preceding confirmed emission events (light solid line); for precipitation preceding non-
emission seismic events (long-dashed, short-dashed line); for precipitation preceding
seismic events which are neither confirmed emission events nor confirmed non-emission
events (long-dashed line); and for precipitation preceding all shallow, explosion-like
seismic events (dotted line).
[Table,14K,GIF]
Table 3:
Probabilities that precipitation values recorded in the days preceding seismic
and emission events was fortuitously greater than normal.
Table 3 presents results of similar analyses for time intervals from 2 to 15 days, for the same four populations of events considered in the binomial probability calculations. The results are not very different from those for the more intuitive but less statistically appropriate binomial probability calculations. If one arbitrarily considers results with a <0.05 to be significant, the confirmed emissions results are highly significant, as are the results for the 28 events as a whole. The non-emission events show no significant deviation at all, whereas the results for the "remainder of events" category are rather ambiguous. The deviations for all groups are most significant over time periods of 4 to 6 days but can still be significant for periods as short as 2 days (for confirmed emissions and "all events") or as long as 15 days (for all but non-emission events). This fact indicates that shallow, explosion-like seismic events not only tend to follow individual storms, but can also be brought on by the effect of storm sequences lasting weeks.
One could perhaps argue that the test for emission events gives a significant result because we are comparing precipitation before these events, which took place at the rainiest time of the year, with the entire two-year precipitation record, both rainy season and dry season. Table 3 also gives 4- and 6-day results comparing the precipitation prior to emissions with just the rain gage records for the months November through February. A similar comparison was made using other seismic events (to be consistent, only those that took place between November and February were included in the comparison). The results do not differ very much from previous ones.
The fact that most shallow, explosion-like seismic events follow within hours to, at most, days after storms indicates that the triggering mechanism must be shallow, probably within the dome itself. It would seem surprising that seismic data, described in the next section, suggests a source of gas that is well below the dome.
A deep gas source is indicated by two lines of evidence: the locations of shallow, explosion-like seismic events; and an increase in seismicity at greater depth.
More than 61 earthquakes were located by personnel at the Cascades Volcano Observatory during shallow, explosion-like seismic events. Most locatable earthquakes (the small black squares in Fig. 7) were small bursts of energy that occurred in the tails of the seismic signals. They cluster in an area that has been a locus of non-magmatic earthquakes since 1980 (Weaver and others, 1987) and may have been a conduit for magmatic gas transport preceding a violent 19 March 1982 eruption (Weaver and others, 1983). Three locations (the large, hollow squares in Fig. 7) record onsets to shallow, explosion- like seismic events, on 24 August 1989, 25 October 1990, and 20 November 1990. These are highly significant because they indicate the source depth of the seismicity. Solutions were located using 7 to 12 stations around the mountain (Fig. 1) including 1 to 3 in the crater. They give low RMS residuals (0.05-0.11 s), and yield depths of 1.44 km to 2.10 km with standard errors of 0.3 km or less. Using a technique developed by Endo and others (1990), the depth was checked independently of the velocity model used in the above solutions by taking the difference in times of arrival at stations GDN and REM, along a radial path from the conduit. For the 24 August 1989 event (the only one where both stations were in operation), the time difference (0.12 s with an uncertainty of +0.01 s in each pick) is less than half what one would expect for a source within the dome but is consistent with a source depth of 1-2 km. It is therefore very unlikely that these locations were near the surface.
[Graphic,19K,GIF]
Figure 7a:
Location of earthquakes that occurred during shallow, explosion-like seismic events and that
meet U.S.Geological Survey, Cascades Volcano Observatory requirements for "A" or "B" quality.
Large
hollow squares represent onsets of shallow, explosion-like seismic events.
Small black squares are
earthquakes that occurred after event onsets.
Standard errors in horizontal and vertical location
are 0.1-0.6 km and 0.1-0.3 km, respectively.
Only earthquakes that took place less than 24-hr after
the 20 December 1990 event are plotted to prevent those
events from dominating the pattern.
[Graphic,9K,GIF]
Figure 7b:
South/North and West/East Profile Lines.
The depth of these onsets appears inconsistent with the inference, derived from the association with storms, that the shallow, explosion-like seismic events are near- surface phenomena. It is important to note that none of the three located events followed storms, nor did they produce confirmable evidence of gas or tephra emission. Do the located events represent a different type of seismicity from those associated with storms? The facts that the seismic signatures for both events (the located, deep ones and the storm-related, presumably shallow ones) are indistinguishable from one another, and that both events took place in 1989-1991 and not before or after, lead one to believe that they do not. A more likely possibility is that they all have the same cause (that is, gas transport) but took place at depths ranging from near the surface to at least 1-2 km. Confirmed emissions may represent a subset at the shallow end of this range, but confirmed non-emissions may be disproportionately represented by deeper members. In any case, the fact that some originate well below the dome leads one to conclude that the source of gas is not within the dome or anywhere near the surface. It is deep enough to be derived from the magmatic conduit or from the magma chamber.
The deep seismicity at Mount St. Helens has been studied by Moran (in press). Moran found that background seismicity between 2- and 9-km depth began to increase from an average of ~2 locatable earthquakes per month in early 1987 to 90 per month in the first half of 1990 (Fig. 8), -- (Web note: Figure 8 not yet available) -- then declined to ~20 per month in the latter half of 1992. The peak in background level coincided remarkably well with the early part of the period of shallow, explosion-like seismicity. This increase is unprecedented in the post-1980 earthquake record and differs from typical precursory activity in two respects: (1) it took place over years, rather than weeks or months as was typical for precursors to eruptions of the early 1980s; and (2) most earthquakes between 1987 and 1990 were below 3 km depth, whereas shallow depths were associated with all precursory records except a March 1982 explosion (Weaver and others, 1983; Moran, in press).
The increase in deep seismicity since 1987 alone does not necessarily indicate that the gas source to the emissions was deep; however, when combined with the onset locations, the story for a deep source is difficult to dismiss. New, small, deep intrusions may have been a source of gas, although there is no evidence (other than the seismic increase) of new intrusions. An alternative possibility, for which there is more evidence, is that gas came directly from the crystallizing conduit.
The crystallization of magma in the Mount St. Helens eruptive conduit and the accompanying exsolution of volatile species has been the subject of at least two previous papers (Cashman, 1988, 1992). Based on examination of dome rock samples, Cashman (1988, 1992) estimated that crystallization of plagioclase microphenocrysts by cooling of magma in the eruptive conduit averaged ~0.1 vol%/month from early 1981 to 1986. The amount of gas exsolved during crystallization, Cashman (1988) argued, could cause episodes of second boiling regularly enough to trigger the dome-building eruptions. By 1986, crystals made up some 60% of the total rock volume. The corresponding increase in lava viscosity was probably a major factor preventing further eruptions (Cashman, 1988).
Since the last dome-building eruption, magma in the conduit has presumably continued to cool, but gas has not been allowed to escape during eruptive episodes. If crystallization continued at the 1981-1986 rate after dome-building eruptions ended, the crystal content would have increased by approximately another 4.5% of the total magma volume (that is, volume of melt plus crystals) by mid-1990. Assuming that the magma between 2- and 7-km depth was saturated at ~3-4.5 wt% water (~3% at shallow depth, 4.5% at greater depth), the crystallization of an additional 4.5 vol% plagioclase would have released ~3-5 kg water per cubic meter of melt, or 4x10^7 to 3x10^8 kg total water from a conduit of 30- to 60-m radius (Pallister and others, 1992).
Would this amount of gas have been sufficient to drive the emissions? Some simple energetics calculations suggest that it may have. The internal energy (above that of liquid water at 1 atm, 100 degrees C) of this water, released from the conduit at magmatic temperature (870-930 degrees C; Rutherford and others, 1985; Cashman, 1992) and subsurface pressures (~50-175 MPa) would have been approximately 1x10^14 to 1x10^15 joules (Haar and others, 1984). Even at temperatures and pressures appropriate to the interior or the dome (4-5 MPa, ~300-600 degrees C), the available energy would have been essentially the same: 1x10^14 to 1x10^15 joules.
The energy required to produce the emissions can be made only approximately, but data are available to make an order of magnitude estimate for one of the more powerful ones. The average ejection velocity (v = 40-80 m/s) from the 5 November 1990 event, taken from ballistic studies mentioned above, combined with estimates of the erupted mass (m) calculated from the volume of the crater (1.2-2.4 x 10^3 m3) and an assumed in situ rock density of ~2000 kg/m3, suggest a kinetic energy of erupted debris (=0.5 m v2) on the order 2x10^9 to 1.5x10^10 joules. Based on ratios of kinetic to total energy of steam eruptions elsewhere (White, 1955; Le Guern and others, 1980), the total energy of this event may have been two to three orders of magnitude greater, perhaps 10^11-10^13 joules. This is one to three orders of magnitude less than the energy available in the exsolved gas. The total energy release for all emissions may have ranged from about the same order of magnitude as the energy contained in the exsolved gas, to a few orders of magnitude less.
Given the current level of knowledge, it is impossible to confidently identify a causal mechanism for the association between emissions and precipitation. The fact that gas was released during explosive episodes rather than continuously suggests that it was contained beneath a shallow, impermeable cap, probably within the dome, that breached during storms. What was that cap and how was it breached? A bit of background on the local geology may shed some light on this question.
The dome's north flank consists of three rubbly exogenous lobes, below which lies a thick section of lava emplaced endogenously during at least three periods of dome growth (Fig. 9). At the base of the dome lie rubbly zones that may have either dissipated or stored gas. Magnetic field strength measurements indicate that the dome's interior, including probably the inner dome below the North and East vents, is well above the magnetization temperature (~375 to 600 C) and could still deform viscously. Above the viscous interior, the entire dome has been fracturing and disintegrating as it cools. Massive rock outcrops and spires many meters high on the dome's surface have crumbled since 1986 to piles of rubble. Aprons of debris of have grown as unstable talus rolls and avalanches to the dome's base. The North Vent itself formed within a previously unrecognized, arcuate pull-apart feature that resembled the head scarp of a landslide (D. Dzurisin, 1993, oral commun.), and a geodetic station (Luther, Fig. 10) subsequently placed below the North Vent showed greater downslope movement than nearby stations before it was destroyed in the next explosion. The final three emissions from the North Vent were accompanied by an unprecedented removal of 8x10^5 m3 of debris from the dome's north flank (Fig. 9), mostly by avalanches and rockslides. This is at least an order of magnitude more debris than had been removed from any part of the dome since 1986.
[Graphic,16K,GIF]
Figure 9:
Schematic north-south cross section through the Mount St. Helens lava
dome illustrating the mechanisms suggested for explosive release of magmatic gas
following storms. Internal structures in the central and southern portions of the dome are
not included in this figure. Location of exogenous lava lobes on the north side of the
lava dome are taken from air photos and from maps in Swanson and others (1987).
Thickness of exogenous lobes is taken from photogrammetric measurements and from
field observations. Thickness of convectively cooled upper carapace on the north side of
the dome is estimated by integrating Dzurisin and others' (1990) calculated average rate
of thickening of the dome's magnetic carapace (0.03 + 0.01 m/day) from late 1984 (the
time of last lava extrusion on the north side) until 1990. Profiles were done using
photogrammetry from air photos taken in September 1990 and September 1991.
[Graphic,10K,GIF]
Figure 10:
Changes in slope distance versus time, measured from points north of
the dome to geodetic stations on the north side of the dome. Locations of geodetic
stations are shown in Figure 3.
Before 5 November 1990, stations were surveyed from
the point Tinker, near the north base of the dome. After 5 November 1990, they were
surveyed from a station (Swani) ~1.8 km north of the dome.
Magnetic field strength measurements (Dzurisin and others, 1990) also show that the solidified carapace around the viscous interior has thickened at an average rate of 0.03 + .01 m/day since the end of dome growth. In accord with observations of Kilauea Iki lava lake in Hawaii (Hardee, 1980), Dzurisin and others (1990) inferred that thickening of the carapace is accompanied by downward propagation of cooling fractures. Both Hardee (1980) and Dzurisin (1990) found that atmospheric precipitation is dominant in determining the rate of cooling of these lava bodies. At Kilauea Iki, the heating to 100 degrees C and vaporization of all water that fell on the lava lake accounts for about two-thirds of the cooling rate. At Mount St. Helens, it accounts for one-third to one-half. The rate of cooling at Mount St. Helens has been found to be several tens of percent higher during wet seasons than during dry seasons and, most importantly, exhibits jumps during individual storms (D. Dzurisin, 1992, oral commun.).
One could imagine that accelerated fracture growth and disaggregation of the dome during storms may have opened passageways between gas-charged regions and the surface. In the North Vent area, slope instability caused by the added weight of water, increases in pore pressure, rapid expansion of steam, or reduction of strength by thermal fracturing could also have exposed regions of pressurized gas. In the East Vent area, no significant amount of mass wasting took place during the explosions, although dome disintegration and fracturing were clearly evident from air photos of this area and presumably were also taking place in the subsurface.
In the dome's interior, propagation of cooling fractures through the viscous center of the dome during storms could conceivably have released gas from the base of the dome to the surface. Calculations of average fracture growth rate suggest that the cooling fractures probably extended to somewhere near the center of the endogenous lobe below the North Vent by late 1989 (Fig. 9) and could have connected with fractures growing up from the base of the dome. If the fracture growth rate during storms was several times the average rate, cooling fractures would have propagated only a few centimeters to several tens of centimeters over the duration of a storm; perhaps up to meters over the course of several storms. If this growth occured in an area that was loaded by topographic stresses or pressure from underlying gases, however, the potential growth of fractures would have been unlimited. Stable or subcritical crack growth could have given way to unstable crack growth that propagated completely through the dome. Ductile flow could have annealed the fractures between emission events, though progressive cooling would ultimately have terminated the annealing process.
At least one other lava dome, at Merapi Volcano, Indonesia, produces large flank failures during storms, apparently also by fracture and disaggregation. This dome first appeared in January 1992 and extruded onto the volcano's steeply sloping (30-40 ) flank during the first half of that year (Smithsonian Institution, 1992). Since January 1992, more than 45 pyroclastic flows have been triggered by dome collapse events (as of April 1993); more than 90% of these have taken place on days of rainfall. Members of the Merapi Volcano Observatory (S. Bronto, 1993, written commun.) noted that, during rainstorms, blocks on the dome's flank are dislodged by pressure from expanding steam, and the dome eventually detaches along a surface that propagates all the way through the dome's massive interior.
The more moderate slope beneath the Mount St. Helens dome and its cooler temperature mean that fracture growth and disaggregation are probably less extensive at Mount St. Helens during each storm than at Merapi. Unlike Merapi, the primary force driving fracture growth at Mount St. Helens was probably internal gas pressure. As gas pressures increased in the dome, the internal seal between regions of high pressure and low pressure were subjected to greater stress until small perturbations, caused by water infiltration, triggered complete failure. These explosions probably would have taken place with or without precipitation, although the storms apparently forced the explosions to occur slightly sooner than they would otherwise have.
The emissions and shallow, explosion-like seismicity have discontinued, although magma in the conduit is presumably continuing to crystallize. One can imagine two possible explanations for the discontinuation: (1) gas pressures that built up during 1986-1989 subsided during 1989-1991 and are now building up again, in which case we might expect renewed emissions and seismicity in the near future. Alternatively, (2) pathways have been opened that allow continuous slow escape of magmatic gas, in which case renewed emissions would be unlikely. These alternatives can be tested by future observations.
Information in this report was collected by many workers at the U.S. Geological Survey in Vancouver and at the Geophysics Program, University of Washington. Field observations were provided by Richard Waitt, Edward Wolfe, Donald Swanson, Daniel Dzurisin, and Jack Kleinman. Terry Gerlach proposed that the recent events may be related to second boiling. Bobbie Myers and Chris Jonientz-Trisler (with help from others at the University of Washington) identified the shallow, explosion-like seismic events. Earthquake locations were determined by Gloria Smith, and helpful reviews of this paper were made by John Eichelberger, Steve Ingebritsen, Donald Swanson, and Elliot Endo.
Birnbaum, Z. W., and Tingey, F. H., 1951, One-sided confidence contours for probability distribution functions: Annals of Mathematical Statistics, v. 22, p. 592-596. Cashman, K. V., 1988, Crystallization of Mount St. Helens 1980-1986 dacite: A quantitative textural approach: Bulletin of Volcanology, v. 50, p. 194-209. Cashman, K. V., 1992, Groundmass crystallization of Mount St. Helens dacite, 1980- 1986: A tool for interpreting shallow magmatic processes: Contributions to Mineralogy and Petrology, v. 109, p. 431-449. Chadwick, W. W., and Swanson, D. A., 1989, Thrust faults and related structures in the crater floor of Mount St. Helens volcano, Washington: Geological Society of America Bulletin, v. 101, p. 1507-1519. Chadwick, W. W., Swanson, D. A., Iwatsubo, E. Y., Heliker, C. C., and Leighley, T. A., 1983, Deformation monitoring at Mount St. Helens in 1981 and 1982: Science, v. 221, p. 1378-1380. Chouet, B. A., Page, R. A., Stephens, C. D., and Lahr, J. C., in press, Precursory swarms of long-period events at Redoubt volcano (1989-1990), Alaska: Their origin and use as a forecasting tool: Journal of Volcanology and Geothermal Research. Conover, W. J., 1971, Practical nonparametric statistics: New York, John Wiley and Sons, 462 p . DeGraff, J. M., and Aydin, A., 1987, Surface morphology of columnar joints and its significance to mechanics and direction of joint growth: Geological Society of America Bulletin, v. 99, p. 605-617. Dzurisin, D., Westphal, J. A., and Johnson, D. J., 1983, Eruption prediction aided by electronic tiltmeter data at Mount St. Helens: Science, v. 221, p. 1381-1382. Dzurisin, D., Denlinger, R. P., and Rosenbaum, J., 1990, Cooling rate and thermal structure determined from progressive magnetization of the dacite dome at Mount St. Helens, Washington: Journal of Geophysical Research, v. 95, p. 2763-2780. Endo, E. T., Dzurisin, D., and Swanson, D. A., 1990, Geophysical and observational constraints for ascent rates of dacitic magma at Mount St. Helens: New York, John Wiley and Sons, p. 317-334. Haar, L., Gallagher, J. S., and Kell, G. S., 1984, NBS/NRC steam tables: New York, Hemisphere Publishing Corporation, 320 p. Hardee, H. C., 1980, Solidification in Kilauea Iki Lava Lake: Journal of Volcanology and Geothermal Research, v. 7, p. 211-233. Jonientz-Trisler, C., Myers, B., and Power, J., 1991, Seismic identification of gas-and- ash explosions at Mount St. Helens: Capabilities, limitations, and regional application, in First International Symposium on Volcanic Ash and Airline Safety, Seattle, WA, July 8-12, 1991, Proceedings: U.S. Geological Survey Bulletin 1065, Washington, D.C., U.S. Government Printing Office, p. 26-27. Kerr, R. A., 1993, Volcanologists ponder a spate of deaths in the line of duty. Science, v. 260, p. 289-290. Le Guern, F., Bernard, A., and Chevrier, R. M., 1980, Soufriere of Guadeloupe 1976- 1977 eruption--Mass and energy transfer and volcanic health hazards: Bulletin Volcanologique, v. 43, p. 578-592. Lipman, P. W., and Mullineaux, D. R., eds., 1981, The 1980 eruptions of Mount St. Helens, Washington: U.S. Geological Survey Professional Paper 1250, Washington, D.C., U.S. Government Printing Office, 844 p. Mastin, L. G., 1991, The roles of magma and groundwater in the phreatic eruptions at Inyo Craters, Long Valley Caldera, California: Bulletin of Volcanology, v. 53, p. 579-596. McBirney, A. R., 1955, Thoughts on the eruption of the Nicaraguan volcano Las Pilas: Bulletin Volcanologique, v. 17, p. 113-117. Moran, S. C., in press, Seismicity at Mount St. Helens, 1987-1992: Evidence for repressurization of an active magmatic system: Journal of Geophysical Research. Murray, J. B., 1980, The Bocca Nuova: its history and possible causes of the September 1979 explosion, in Huntington, A. T., Guest, J. E., and Francis, E. H., eds., United Kingdom Research on Mount Etna, 1977-1979: London, England, The Royal Society of London, p. 46-50. Pallister, J. S., Hoblitt, R. P., Crandell, D. R., and Mullineaux, D. R., 1992, Mount St. Helens a decade after the 1980 eruptions: Magmatic models, chemical cycles, and a revised hazards assessment: Bulletin of Volcanology, v. 54, p. 126-146. Rutherford, M. J., Sigurdsson, H., Carey, S., and Davis, A., 1985, The May 18, 1980, eruption of Mount St. Helens 1. Melt composition and experimental phase equilibria: Journal of Geophysical Research, v. 90, p. 2929-2947. Smithsonian Institution, 1992, Merapi (Java): Pyroclastic flows follow earthquakes and rainfall; gas data: Bulletin of the Global Volcanism Network, v. 17, no. 10, p. 3. Smithsonian Institution, 1993, Guagua Pichincha (Equador): Phreatic explosion kills two scientists: Bulletin of the Global Volcanism Network, v. 18, no. 2, p. 2. Swanson, D. A., Casadevall, T. J., Dzurisin, D., Malone, S. D., Newhall, C. G., and Weaver, C. S., 1983, Predicting eruptions at Mount St. Helens, June 1980 through December 1982: Science, v. 221, p. 1369-1376. Weaver, C. S., Zollweg, J. E., and Malone, S. D., 1983, Deep earthquakes beneath Mount St. Helens: Evidence for magmatic gas transport?: Science, v. 221, p. 1391-1394. Weaver, C. S., Grant, W. C., and Shemeta, J. E., 1987, Local crustal extension at Mount St. Helens, Washington: Journal of Geophysical Research, v. 92, p. 10170-10178. White, D. E., 1955, Violent mud-volcano eruptions of Lake City Hot Springs, northeastern California: Geological Society of America Bulletin, v. 66, p. 1109- 1130.
