Steve McNutt

Recent Abstracts

Volcanic Lightning: New Global Observations and Constraints on Source Mechanism

E Venzke, SR McNutt, ER Williams (Smithsonian Inst. (EV), MIT (ERW), and Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775 (SRM)); ph. (907) 474 7131; email: steve@giseis.alaska. edu

New data on volcanic lightning from the Smithsonian Volcano Reference File are added to an existing database and greatly expand the number of cases available for study. Lightning has now been documented at 154 volcanoes in ass ociation with 394 eruptions, a significant increase from the earlier numbers of 89 volcanoes and 240 eruptions. Lightning and electrification at volcanoes are important because they represent a hazard in their own right, they are a component of the global electrical circuit, and because they contribute to ash particle aggregation and mo dification within ash plumes. The role of water substance (water in all forms) in particular has not been well st udied. The Volcanic Explosivity Index (VEI) was determined for 177 eruptions. Eight percent of VEI=3-5 eruptions have reported lightning, and 10 percent of VEI=6, but less than 2 percent of those with VEI=1-2, suggesting consi stent reporting for larger eruptions but either less lightning or under-reporting for small eruptions. Ash plume heights (142 observations) show a bimodal distribution with peaks at 7-12 km and 1-4 km. The former are similar t o heights of typical thunderstorms and suggest involvement of water substance, whereas the latter suggest other f actors contributing to electrical behavior near the vent. The distributions of the latitudes of volcanoes with li ghtning and eruptions with lightning roughly mimic the distribution of all volcanoes; flat with latitude. Meteor ological lightning, on the other hand, is common in the tropics and decreases markedly with increasing latitude a s the ability of the atmosphere to hold water decreases poleward. This finding supports the idea that if lightnin g in large eruptions depends on water substance, then the origin of the water is primarily magma and not entrainm ent from the surrounding atmosphere.

Long swarms and short swarms

SR McNutt (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775); ph. (907) 474 7131; email: steve@giseis.alaska. edu

Many earthquake swarms at volcanoes last several months, then have a sharp uptick in rate in the hours before eru ption. Examples include 2006 Augustine, 8.5 months then 10 hours; 1992 Spurr, 10 months then 4 hours; 1994 Rabaul , ~1 year then 27 hours; 2008 Kasatochi, 6 weeks then 2 days; and 2011 Puyuehue Cordon Caulle, 5 weeks then 2 day s. For the well studied Augustine case, broadband data showed that very long period (VLP) energy accompanied 221 of 722 located earthquakes in the 10 hours before the first explosive eruption on 11 January 2006. This was revea led by low-pass filtering and the period of the VLP signal was 50 sec. The Augustine broadband stations were camp aign instruments at distances of 2-3 km from the vent. No similar VLP energy has been found in events during the 8.5 month long swarm. Okmok volcano had a short swarm only lasting 5 hours prior to its 12 July 2008 eruption. Lo w-pass filtering of data from broadband station OKSO, 10 km from the vent, showed that 23 of 42 located events ha d VLP energy with a period of 30-40 sec. Events from Kasatochi volcano were scanned on station ATKA. Here the bro adband station is much farther away at 88 km but the earthquakes in the short swarm 7 August 2008 were much large r with many M>3 events. The station suffered data gaps so only a few hours of data were scanned but numerous even ts were observed with VLP energy starting just after the P phase. Low-pass filtering showed VLP energy with a per iod of 10-12 sec. No VLP energy has been found in events of the preceding 6 week long swarm. These observations a t three different volcanoes suggest that the short swarms represent a different process than the long swarms. The long swarms likely reflect pressure increases in the surrounding country rock caused by increasing magma pressur e. The short swarms in contrast, appear to represent discrete pulses of magma injection at shallow depths. For al l three volcanoes the earthquakes looked like typical volcano-tectonic (VT) earthquakes on short-period stations. This demonstrates that broadband stations are needed at close distances to be able to make the needed observatio ns. The short swarms are very short - a few hours to 2 days - and have important implications for hazards a ssessments. It is not known how commonly the long swarm-short swarm pairs occur and the false alarm rate is also not known.

PERIODICALLY-TRIGGERED SEISMIC EVENTS AT MT. WRANGELL VOLCANO FOLLOWING THE SUMATRA-ANDAMAN ISLANDS EARTHQUAKE

M. West, J.J. Sanchez, and S.R. McNutt, Alaska Volcano Observatory-UAF-GI, Fairbanks, Ak, 99775; steve@giseis.alaska.edu.

Following the Sumatra-Andaman Islands earthquake (Mw=9.0) on December 26, 2004 Mt. Wrangell experienced a swarm of seismic events as the surface waves from the Sumatra-Andaman earthquake crossed the region suggesting that the local events were triggered by the transient stress changes. This is remarkable because Mt. Wrangell is nearly 11,000 km from the epicenter. Mt. Wrangell is one of the largest andesite shield volcanoes in the world. It is located in south-central Alaska and anchors the eastern end of the Aleutian-Alaska chain of arc volcanoes. Fumarolic activity and occasional steam plumes indicate that Wrangell continues to be an active hydrothermal system. Because of its volcanic and seismic activity, a network of seismometers is operated at Wrangell by the Alaska Volcano Observatory and the Alaska Earthquake Information Center. Though the waveforms from the teleseismic and local events are superimposed, the dominant periods and amplitudes of the teleseismic event are 2 orders of magnitude larger than those of the local events. The two are easily separable with frequency filtering. Events are located near Wrangell's summit with magnitudes ranging from -0.3 to 1.9. While swarms are not uncommon at Wrangell, the events in this swarm are significant because they are larger in amplitude and occur in-phase with the teleseismic surface wave ground motion. We show that the statistical likelihood of such a pattern occurring naturally is exceedingly low. The most striking observation is that the local events occur almost exclusively during the compressional phase of the teleseismic surface wave-that is, coincident with negative excursions of the vertical component displacement record. The fact that the ground motion frequency is the same as the occurrence frequency for local events can be exploited to better define the mechanism responsible for triggering.

Quasi-periodic Episodes of Volcanic Tremor at Okmok Volcano, Alaska

CG Reyes, SR McNutt (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775); ph. (907) 474 7131; email: steve@giseis.alaska.edu

A new seismic network, consisting of nine short-period stations, was installed at Okmok volcano, Alaska, in summer 2002. Continuous data began to be recorded in April 2003. Eventually, this array will be expanded to include four additional 3-component broadband stations and 4 co-located continuous GPS stations. The most recent eruption occurred in February - April 1997 and, like most historical eruptions, issued from Cone A - which is still steaming as of this writing. INSAR measurements of post eruptional deformation show that the volcano is inflating at a rate of ~10 cm/yr (Mann and Freymueller, 2002). Seismically, there are very few local earthquakes; however, there has been nearly continuous low level tremor with stronger amplitude bursts occurring at a rate of about 2 per hour (2.2 ( 0.8). These bursts have durations generally in the range of 5-8 minutes. Most bursts begin abruptly, reach a maximum amplitude within 3 minutes, then decline gradually over the next several minutes. Many bursts resemble each other quite closely, suggesting that there is a single main process or source location. No systematic trends have yet been observed between duration and either amplitude or the length of the preceding interval. The tremor is composed of irregular waves with a broad range of frequencies between ~1 and ~20 Hz, with main peaks at ~2.2 and ~6.2 Hz, and with a lesser peak at ~ 5.5 Hz. Okmok is the only volcano out of 25 monitored Aleutian volcanoes to display this type of tremor. This activity may be linked to magma transport and the observed high rate of deformation.

Seismic and Geodetic Unrest at Uturuncu Volcano, Bolivia

SR McNutt (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775) ME Pritchard (Caltech); ph. (907) 474 7131; email: steve@giseis.alaska.edu

A large-scale concentric pattern of deformation has been observed between 1994 and 2000 centered on Uturuncu Volcano, Bolivia from satellite geodetic surveys (Pritchard and Simons, 2002). A reconnaissance investigation by a team composed of scientists from Bolivia (M. Sunagua and R. Muranca), Chile (J. Clavero) , the USA (S. McNutt and M. Pritchard) and the UK (C Annen, M Humphreys, A le Friant, and R.S.J. Sparks) took place from 1 - 6 April 2003 to see if there were any other signs of volcanic unrest at Uturuncu. A single component vertical, one-second seismometer, was placed at five locations for periods up to 14 hours. Data were recorded at 100 samples per second on a laptop PC. Persistent low level seismicity was observed at all locations. The events were very similar to each other and most had a distinct P wave with a period of 0.1 sec and a clear S wave with longer period and higher amplitude. Nearly all the events at each station had similar S-P times, suggesting that the events came from mainly one source. Using bootstrapping methods we determine this to be from a source location on the north-west flank close to the center of deformation observed by satellite surveys. Several events with different S-P times and different waveforms suggest that two other sources exist within the volcanic edifice, but these cannot be located with the available data. The rate of volcanic earthquakes was up to 15 per hour; this is a surprisingly high rate for a dormant stratovolcano. The magnitudes were in the range 0.5 to 1.5 based on coda length. The sources were considered to be shallow within 3 - 4 km of the surface, although information on the velocity structure is not known. The summit region of Uturuncu (6,008 m) has an active fumarole field with substantial sulphur production and areas of silification. Temperatures in four fumaroles were measured at 79 - 80°C. A hot water spring on the NW flanks had a temperature of 20°C. The recent unrest manifested by substantial ground deformation, thermal activity and seismicity indicate that a magmatic system may still be present, and therefore further monitoring of this dormant volcano is warranted.

Reference: M Pritchard and M Simons, "A satellite geodetic survey of large-scale deformation of volcanic centres in the Central Andes" Nature, 418, p167-170, 2002.

Pressure Sensor Data Reveal New Details of the 1999 Eruption of Shishaldin Volcano, Alaska

J. Caplan-Auerbach and SR McNutt (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775; ph. (907) 474 7131; email: steve@giseis.alaska.edu

New data from a pressure sensor, along with seismic and satellite data enable us to perform a detailed analysis of the 1999 eruption of Shishaldin volcano, Alaska. The eruption was well monitored by a 6-station seismic network and frequent satellite passes, but visual observations were minimal. Consequently, we relied on satellite imagery and seismic tremor to interpret the chronology and behavior of the eruption. To refine our interpretation of the 1999 eruption we investigate acoustic data recorded on a pressure sensor 6.5 km north of Shishaldin. Four types of acoustic signals were identified, representing different types of eruptive behavior. A low-frequency (1-5 Hz), pulsating signal correlates with strong thermal anomalies in GOES and AVHRR satellite imagery and is interpreted as signifying lava fountaining. On April 19, 1999, the fountaining signal is replaced by a monotonic hum that builds in amplitude for $>$13 hours. At 1936 UTC on April 19, the humming frequency declines and the 2.3 Hz signal abruptly ends. This cessation of the humming signal coincides with a dramatic increase in seismic tremor amplitude, previously believed to represent the onset of a Subplinian eruption. Pressure sensor data, however, suggest that the main Subplinian phase begins six minutes later. At this time, the pressure sensor recorded a 20-minute broadband signal, over which several low-frequency bursts are superimposed. The final acoustic phase detected by the pressure sensor is a series of discrete pulses, interpreted as gas explosions. The strongest explosions, recorded on April 23rd, were associated with a small, ash-poor plume and strong seismic tremor. In time series, these events are similar to gas explosions observed at a number of other volcanoes. However, the Shishaldin events are of lower frequency (1-2 Hz) and are 1-2 orders of magnitude ($\sim$100 Pa at 7 km) larger than explosions observed at Stromboli, Arenal, and Karymsky volcanoes. The frequency of the gas explosions allows us to put constraints on the size of Shishaldin bubble bursts and the amount of gas released during the eruption. The 1999 eruption of Shishaldin shows that pressure sensors can serve as an excellent complement to traditional means of monitoring remote volcanoes.

Volcanic tremor and its use in Estimating Eruption Parameters

S.R. McNutt (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775; ph. (907) 474 7131; email: steve@giseis.alaska.edu)

Volcanic tremor, a continuous seismic signal, accompanies virtually all eruptions. Several published studies have examined the relation between tremor reduced displacement (DR, a normalized amplitude measure) and the Volcanic Explosivity Index (VEI) or ash plume height. The goal of these studies is to determine the physical relationship between tremor and eruptions and to use DR time histories or peak values to provide real-time estimates of eruption parameters. This study examines tremor for 50 eruptions from 31 volcanoes. Several systematic trends are observed: 1) large eruptions produce stronger tremor than small ones; 2) fissure eruptions produce stronger tremor than circular vents for the same fountain height; 3) eruptions with higher gas content produce stronger tremor than those with low gas content at the same volcano; and 4) phreatic eruptions produce stronger tremor than magmatic eruptions for the same VEI.

The task of using tremor DR to estimate eruption parameters is fundamentally a statistical problem with several factors contributing to uncertainties. First, tremor occurs when volcanoes do not erupt as well as when they do. Based on a worldwide sample, 60 to 80 percent of tremor episodes accompany eruptions, while 20-40 percent do not. Thus there is a significant chance that no eruption is occurring. Second, for each VEI, there is a range of DR, so it is possible to overestimate or underestimate the VEI from the DR. Hence there will always be a false alarm rate of about 10 percent. Improvements can be made in the estimates if the types of eruptions, shapes of vents, and gas contents are known. These can be estimated in advance from previous eruptions or measured near-real-time from independent data. However, adding additional information takes time, thus delaying forecasts. A primary benefit of seismic data is that it is real-time, it is not affected by darkness, and is usable during poor weather, although the signal-to-noise ratio can be worsened. Monitoring tremor DR is thus an effective way to characterize eruptions in progress.

Simultaneous Earthquake Swarms and Eruptions in Alaska, Fall 1996

SR McNutt, W. Marzocchi, and J. Power (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775; ph. (907) 474 7131; email: steve@giseis.alaska.edu

In 1996 the Alaska Volcano Observatory recorded an unprecedented level of seismic and volcanic activity. Pavlof Volcano erupted from Sept. 11 to Dec. 29; the strongest earthquake swarm to date occurred at Iliamna Volcano, peaking Aug. 15; a strong swarm began Sept. 25 at Strandline Lake, NE of Mount Spurr (Strandline may not be related to volcanism), peaking Oct. 7-28; and a vigorous swarm occurred beneath Martin and Mageik Volcanoes from Oct. 16-21, 1996. These volcanoes are located in an 870 km long section of the Alaska-Aleutian arc. No additional swarms or eruptions occurred at these volcanoes in the 5 years since 1996.

We conducted two formal statistical tests to determine the likliehood of these events occurring randomly in the same time interval. The first test considered only the volcanoes at which swarms or eruptions occurred (7 of 13). We produced 10,000 synthetic catalogs under the assumption that the sites are independent. Time intervals of 30 to 360 days were tested. The monitoring began at different times, and the interval chosen for analysis ended on December 31, 2000. The second method is a hierarchical Bayesian model in which the probability of a swarm at each volcano is different but the parent population is the same. We evaluated all 13 volcanoes that were monitored in 1996.

In both cases we found that the likliehood of the swarms and eruption occurring together by chance alone is small: less than 1 percent for the first test and about 1 percent for the second. Therefore we conclude that the events may have been triggered. We speculate that an arc-wide deformation pulse occurred, and also note that a M=8 earthquake took place in the central portion of the arc in May, 1996.The two may be related. No arc-wide continuous deformation data existfor 1996, so we cannot determine physical mechanisms. However, the high level of earthquake and volcanic activity in the Aleutian arc suggests that such triggering or interaction may occur again. It wouldbe prudent to install continuous deformation instruments at several places in the arc in order to better understand such simultaneous seismic and volcanic activity.

Source Processes of Volcanic Eruption Tremor: A Kinetic Energy Source Model

IASPEI Meeting, Athens Greece August 1997

T. Nishimura, Graduate School of Science, Tohoku Univ., Sendai 980-77, Japan; S. R. McNutt, Geophysical Institute, Univ. of Alaska, Fairbanks, AK, 99708, USA

Based on published data of 24 cases of eruption tremor of 18 volcanoes around the world, we find the characteristic features of their reduced displacement (DR, normalized amplitude by hypocentral distance) to be: (a) maximum DR is approximately proportional to the square root of the cross sectional area of the vent; (b) about one half of the cases show exponential increases in DR at the beginnings of eruptions; (c) one half of the cases show a sustained maximum level of tremor; (d) more than 90 percent of the cases show exponential decay at the ends of eruptions; and (e) exponential increases, if they occur, are commonly associated with the first large stageof eruptions. To explain these features, we propose a kinetic energy source model of eruption tremor, which is based on simple energy transfer assumptions: seismic wave (tremor) energy is proportional to strain energy in wall rocks, which is proportional to kinetic energy of magma flow. This kinetic energy model predicts that, when the average velocity and density of volcanic ejecta are constant, DR of tremor is proportional to the square root of the cross sectional area of the vent. We derive an expression for the exponential increase in DR by increasing the vent radius due to the mechanical energy of magma flow. Calculations suggest that the strength of the country rock and the material in the conduit may play a role in such exponential increases. We explain the exponential decay of tremor by approximating the mixture of volcanic ejecta and gas as a perfect gas migrating iscentropically upward through a vent. Magma properties and vent geometries both also affect seismic efficiency of tremor, with larger DR correlating with both basaltic compositions and planar fissures.

Seismic Monitoring of the September-December 1996 Eruptions of Pavlof Volcano by the Alaska Volcano Observatory

SR McNutt (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775; ph. (907) 474 7131; email: steve@giseis.alaska.edu

Pavlof, the most active volcano in North America, began to erupt September 15, 1996 after an 8-year repose. The eruption began just two months after the Alaska Volcano Observatory (AVO) installed a new seismic newtwork consisting of six local and six regional stations to monitor the volcano. New funding was provided by the Federal Aviation Administration in 1996 to monitor Pavlof and three other volcanic areas to improve airline safety. About 200 heavy jet aircraft fly over the Alaska/Aleutian volcanic arc each day, carrying 10,000 passengers and freight. Volcanic ash is present in the North Pacific airways an average of four days per year. AVO deployed seismic equipment including custom designed fiberglass enclosures and power systems. Data from all stations are telemetered to AVO offices in Fairbanks and Anchorage in real time, 24 hours per day. AVO and UAFGI are currently developing a new generation of automated monitoring and analysis tools using the world wide web. Selected volcano seismic data are fed into a near-real-time spectral analysis program which plots spectrograms automatically every 5 minutes for 15-minute windows. Peak amplitudes are then determined from each spectrum and converted to reduced displacement (DR), which is a normalized amplitude measurement used for volcanic tremor. Multi-station averages of DR are computed and plotted automatically, and selected threshold values are used to trigger email messages and alert a duty person by pager. A relation between DR and the Volcanic Explosivity Index (VEI), a measure of the size of eruptioins, has been previously developed by the author. We thus use the DR values to estimate the VEI of the eruptions in near-real-time. This sytem was used several times in fall 1996 to determine changes in the AVO color code alert levels. AVO also intends to use ER to estimate starting values for ash column heights to model the evolution of plumes in near-real-time. The new seismic network and analysis tasks, coupled with the world wide web, make it possible to quickly and effectively determine eruptive status and disseminate information.

An evaluation of b-value spatial mapping techniques based on an analysis of seismicity at Mt. Spurr, Alaska and synthetic data

A.D. Jolly, S.R. McNutt, and S. Wiemer, Alaska Volcano Observatory-UAF-GI, Fairbanks, Ak, 99775; art@giseis.alaska.edu. J.C. Lahr, USGS-UAF-GI, Fairbanks, AK, 99775.

Shallow seismicity located beneath Mount spurr summit between 1981 and 1992 is characterized by a persistent ellipsoidal seismic volume with a lateral diameter ~2.5 km, a ~5.0 km range in depth, and highly uniform focal mechanism solutions. The 170 events in the volume have magnitudes ranging from 0.0-2.0, and have b-value=0.78 +/- 0.05, where b is the coefficient of the frequency-magnitude relation (log N =a-bM). Within this seismic volume, spatial variationis of b range between 0.6 and 1.1, with b=1.1 at the center of the ellipse (spatial variations are determined by gridding the volume and calculating b for 80 nearest neighbor events per grid node). We tested the hypothesis that these observations might arise from a single point source and that variations in b result from artifacts by building a synthetic distribution of earthquakes having a perfect location a b-value=0.75. We used a synthetic seismic station geometry having the generalized form of the existing array. Travel-time reading errors were modeled by application of random normal errors to P and S wave arrival times (standard deviatioin of P = 0.04 s and S =0.08 s). Non-uniformity in P and S phase use was modeled through a station usage probablility distribution incorporating distance, magnitude and arrival time quality. Relocation of the synthetic data with these model parameters yields and elliptical volume having lateral diameter ~2.0 km, depth range ~4.0 km and b ranging from 0.6 to 1.0: substantially the same as the observed data. The randomizatioin process yields unstable spatial variations of b within the volume. The spatial instabilities become stable b-value artifacts after application of travel-time or magnitude perturbations to selected seismic stations. Introduction of asymmetric station geometries result in similar artifacts. We find stable b-value artifacts as large as ~1.3 without introducing anamalous averaged travel-time residuals. The synthetic data suggest taht seismicity at Mount Spurr summit probably results from a singe point source and that variations in b within the volume might occur by chance or from inadequately modeled travel-time or attenuation anomalies. These observations need not result from multiple sources having different b-values.

Variations in the frequency-magnitude distribution with depth at Mount St. Helens and Mt. Spurr

S.R. McNutt; S. Wiemer (Geophysical Institute, UA Fairbanks; ph. (907) 474 7131; email:steve or stefan@giseis.alaska.edu

The frequency-magnitude distribution (FMD) of earthquakes, cahracterized using the b-value, is examined as a function of space beneath Mt. St. Helens (1988-1996), and Mt. Spurr (1991-1995). At Mt. St. Helens, two volumes of anomalously high b (b>1.3) can be observed at depths of 2.6-3.6 km below the crater floor and below 6.4 km. These anomalies coincide with (1) the depth of vesiculation of ascending magma, and (2) the suggested location of a magma chamber at Mt. St. Helens. Study of Mt. Spurr reveals an area of high b-value (b>1.3) at a depth of about 2.3-4.5 km below the crater floor of the active vent Crater Peak. We propose that the higher material heterogeneity in the vicinity of the magma is the main cause of the increased b-value at shallow depths. Alternatively, interaction of magma with groundwater may have increased pore pressure and lowered the effective stress. The deeper anomaly at Mount St. Helens is likely caused by high thermal stress gradients in the vicinity of the magma chamber. Our results show that the b-value varies significantly as a function of depth and location under the two volcanoes investigated. The detailed analysis of the FMD is capable of resolving two distinctly different areas: (1) volumes containing average to low b-values (b=0.8 +/- 0.2), and (2) volumes containing high b-value anomalies (b=1.5 +/- 0.3). Regions of high b-value are characterized by an increased number of events, and typically have a largest event one magnitude unit smaller than the surrounding regions. Based on these observations we believe it is incorrect to assume that volcanic areas can simply be characterized by an overall high b-value. Instead we propose that anomalously high b-value pockets exist in a crust otherwise described by an average or normal b-value. Anomalies in the FMD similar to those described here have recently also been observed at several volcanoes in Japan. We suggest that detailed mapping of the FMD can be used as a tool to trace vesiculation and to locate active magma chambers.

Amplitude Scaling of Volcanic Tremor at Mt. Spurr, Pavlof, Redoubt, Karkar, Arenal, and Kilauea Volcanoes

J.P. Benoit; S.R. McNutt (Geophysical Institute, UA Fairbanks, Fairbanks, AK, 99775; ph. (907) 474 7131; email: benoit or steve@giseis.alaska.edu); V. Barboza (Univ. Nac., Heredia, Costa Rica; email: vbarboza@irazu.una.ac.cr)

The amplitude scaling or frequency-size distribution of volcanic tremor was examined at 6 volcanoes. The hypothesis that the frequency-size distribution may be approximated by an exponential function was tested. The exponential model, implying a characteristic source-length scale such as a conduit or crack dimension, is found to be a better fit to data then a power-law (self-similar) model. The exponential model gives a satisfactory description of the scaling of both eruptive and non-eruptive tremor. The frequency-size distribution of tremor is determined by measuring the duration of tremor at each particular amplitude. Real-time Seismic Amplitude Measurement (RSAM) data were used at Spurr and Redoubt to quickly measure average amplitudes over long periods of time. At the other volcanoes direct measurements were made on original seismograms or plots of tremor amplitude versus time. Both RSAM and direct observations give equivalent results within measurements errors. We used the exponential model described by: d(Dr) = d(total)e^(-lambda*Dr), where d is the duration of tremor greater than or equal to a particular amplitude Dr, d(total) is the total duration of tremor, and lambda is the characteristic amplitude or scaling constant. These scaling constants can then be related to a source length scale. Results show that Pavlof eruptive tremor and Kilauea's deep tremor have similar scaling constants. The scaling constant at Mt. spurr is greater than those from both Pavlof and Kilauea implying a longer characteristic source length at Mt. Spurr. At Pavlof the amplitude distribution departs from the exponential model at small amplitudes. We suggest that the break in scaling is due to a second source process generating tremor. Karkar, Redoubt, and Arenal all have tremor which shows exponential scaling with minor variations in the scaling constants. The exponential scaling of tremor demonstrates that tremor source processes are fundamentally different from those for earthquakes.