Taryn Lopez Geophysical Institute, University of Alaska Fairbanks, 903 Koyukuk Drive, Fairbanks, Alaska 99775, USA

Transcription

1 Article Volume 14, Number December 213 doi: ISSN: Characterization and interpretation of volcanic activity at Karymsky Volcano, Kamchatka, Russia, using observations of infrasound, volcanic emissions, and thermal imagery Taryn Lopez Geophysical Institute, University of Alaska Fairbanks, 93 Koyukuk Drive, Fairbanks, Alaska 99775, USA David Fee Wilson Infrasound Observatories, Alaska Volcano Observatory, Geophysical Institute, University of Alaska Fairbanks, Fairbanks, Alaska, USA Fred Prata Norweigen Institute for Air Research, Kjeller, Norway Jonathan Dehn Geophysical Institute, University of Alaska Fairbanks, Fairbanks, Alaska, USA [2] A multiparameter data set including measurements of infrasound, volcanic emissions, and thermal imagery is used to characterize and interpret diverse volcanic activity observed during field campaigns in August 211 and July 212 at Karymsky Volcano, Kamchatka, Russia. Four activity types are visually identified and characterized according to: SO 2 emission rate, ash mass, event duration, peak temperature, thermal radiation energy, infrasound onset and frequency, reduced infrasonic pressure, and acoustic energy. These activity types include: (1) ash explosions, (2) pulsatory degassing, (3) gas jetting, and (4) explosive eruption. Unique infrasound signals are associated with all four activity types suggesting that infrasound can be used to help remotely and continuously detect and characterize volcanic activity at Karymsky and other similar volcanoes. Our observations suggest that SO 2 is emitted continuously, though in varying abundance, while ash is emitted discontinuously and is only associated with certain types of activity. Our data set supports previous models that attribute variations in surface activity to changes in conduit permeability at Karymsky Volcano. Evidence for a decrease in conduit permeability as a trigger for ash explosions and explosive eruption activity types is supported by weakened but still detectable SO 2 emission rates prior to eruption, along with the highly impulsive infrasonic onset and large reduced infrasound pressure indicating high pressure at the vent. We speculate that changes in conduit permeability at Karymsky Volcano result from changes in magma supply from the shallow-crustal storage region though additional measurements are required to validate this hypothesis. Components: 15,56 words, 9 figures, 2 tables. Keywords: Karymsky volcano; infrasound; volcanic emissions; thermal imagery; Multiparameter; conduit permeability. Index Terms: 8485 Remote sensing of volcanoes: Volcanology; 8428 Explosive volcanism: Volcanology; 843 Volcanic gases: Volcanology; 8494 Instruments and techniques: Volcanology; 8419 Volcano monitoring: Volcanology; 432 Geological: Natural Hazards; 4337 Remote sensing and disasters: Natural Hazards; 728 Volcano seismology: Seismology American Geophysical Union. All Rights Reserved. 516

2 Received 3 May 213; Revised 24 September 213; Accepted 28 October 213; Published 11 December 213. Lopez, T., D. Fee, F. Prata, and J. Dehn (213), Characterization and interpretation of volcanic activity at Karymsky Volcano, Kamchatka, Russia, using observations of infrasound, volcanic emissions, and thermal imagery, Geochem. Geophys. Geosyst., 14, , doi:. 1. Introduction [3] Recent technological advancements in the field of volcano remote sensing allow measurements of temperature, composition, and flux of volcanic emissions, including gas and tephra, at much higher spatial and temporal resolutions than was previously possible [e.g., Francis et al., 1995; Galle et al., 22; Edmonds et al., 23a; Mori and Burton, 26; Bluth et al., 27; Harris and Ripepe, 27a; Prata and Bernardo, 29], facilitating comparison with high-temporal resolution geophysical data sets and enabling volcanic processes to be inferred. Changes in volcanic sulfur dioxide (SO 2 ) emissions detected by ground and air-based remote sensing have frequently been observed prior to eruption and have been ascribed to (1) changes in relative magma degassing depth due to a decrease in volatile solubility within decompressing magma [Gerlach, 1986; Daag et al., 1996; Aiuppa et al., 27; Burton et al., 27; McGee et al., 21; Lopez et al., 213; Werner et al., 212], (2) changes in conduit permeability [Fischer et al., 1996, 22], and/or (3) scrubbing of acid gases by a shallow water system [Symonds et al., 21; Gerlach et al., 28; Werner et al., 212]. When evaluated in conjunction with complementary gas composition and/or geophysical data sets, the process producing these changes may be deduced. Remote thermal imagery of volcanic activity has been used successfully to estimate and/or evaluate the temperature, mass flux, particle size, ascent speeds, and trajectories of erupted material [Pieri and Baloga, 1986; Ripepe et al., 25a; Harris and Ripepe, 27b; Patrick et al., 27; Harris et al., 213]. The eruption of volcanic ash can indicate vent clearing of either previously erupted material, or the fragmentation and eruption of shallow juvenile magma [Taddeucci et al., 22; Rowe et al., 28]. Complementary high-temporal resolution remotesensing and geophysical data sets from active volcanoes have been used successfully to: identify changes in magma supply and/or depth [e.g., Ripepe et al., 25b; Poland et al., 28], constrain conduit geometry [e.g., Harris and Ripepe, 27a; Nadeau et al., 211], and characterize eruption dynamics [e.g., Johnson and Lees, 2; Harris and Ripepe, 27a; Johnson, 27; Sahetapy-Engel et al., 28; Bull and Buurman, 213; Fee et al., 213]. [4] Infrasound, or low frequency sound waves below 2 Hz, has particular promise for combining with remote-sensing data to infer eruptive behavior, as the source regions for both volcanic emissions and infrasound are typically within the shallow conduit or above the vent [Johnson and Ripepe, 211; Fee and Matoza, 213]. Infrasound is produced at volcanic vents primarily due to the rapid release and expansion of volcanic gases, and the eruption of ash and lava. As a stand-alone tool, high-amplitude infrasound signals can indicate high vent overpressures [Johnson and Ripepe, 211; Fee and Matoza, 213] and may provide insight into conduit permeability. Furthermore, infrasound energy has been shown to correlate with both ash cloud height [Fee et al., 21] and cumulative SO 2 emissions [Fee et al., 213], suggesting that infrasound energy may be used to infer relative eruption size and/or explosivity. Additionally, when used in combination with thermal radiation energy, infrasound energy has been used to determine relative fragmentation of eruptive products and to discriminate between different styles of volcanic activity [Johnson et al., 24a; Marchetti et al., 29]. [5] Here we use high-temporal resolution, coincident measurements of infrasound, SO 2, ash, and thermal radiation from Karymsky Volcano to: (1) characterize the observed activity, (2) identify unique data signals indicative of certain styles of volcanic activity, and (3) refine interpretations of surface processes and subsurface conditions. 2. Karymsky Volcano [6] Karymsky Volcano ( N, E, 1536 m), a predominantly andesitic stratovolcano, is one of the most active and dynamic volcanoes in Kamchatka, Russia (Figure 1) [Izbekov et al., 517

3 A) B) Deployment 212 Deployment Camera Observation Point Karymsky 55 Latitude ( o N) km km Longitude ( o E) Figure 1. Location map. (a) Karymsky Volcano marked by a red triangle within Russia s Kamchatka Peninsula. (b) ALI satellite image of Karymsky Volcano, with locations of the remote-sensing instruments (blue diamond), the 211 infrasound array (yellow circles) and 212 infrasound array (red circles). A summit plume obscures the vent and south side of the edifice. 24]. Its current eruptive cycle began on 2 January 1996 with an explosive summit eruption of andesitic ash and gas, followed 12 h later by a phreatomagmatic eruption of basalt to basaltic andesite (52 wt.% SiO 2 ) from a new vent within the Akademia Nauk caldera lake located 6 km south of Karymsky s main edifice [Izbekov et al., 24]. While the eruption from Akademia Nauk only lasted 18 h, the eruption from Karymsky Volcano continues through the time of this writing. Activity at Karymsky Volcano since 1996 has consisted of discrete, explosive eruptions of ash and gas, frequently described as Vulcanian to Strombolian in appearance, along with various styles of degassing often accompanied by audible chugging or jetting/roaring, and periodic effusion of blocky lava flows [Johnson et al., 1998; Johnson and Lees, 2; Fischer et al., 22; Izbekov et al., 24]. From 1996 through 1999 activity at Karymsky Volcano was fairly regular, with ashrich explosions occurring every 4 6 h and producing plumes up to 9 m above the vent, and lava extruding on a nearly continuous basis [Izbekov et al., 24]. Juvenile material erupted between 1996 and 1999 was described as a porphyritic andesite, with a nearly constant whole rock SiO 2 concentration of 61.9 wt.%, containing vol.% phenocrysts of plagioclase, clinopyroxene, orthopyroxene, and magnetite [Izbekov et al., 24]. Activity since 1999 has been irregular, with daily to monthly explosive eruptions producing ash clouds to 6 km above sea level (ASL) [KVERT, 212]. Lava extrusion since 1999 has become sporadic, with reports of lava extrusion in September/October 22, April 27, and December 29 [KVERT, 212]. Additionally, on several occasions in 25 and 26, scientists observed an 2 3 m diameter lava dome within Karymsky s summit [KVERT, 212]. The range of activity and frequent eruptions make Karymsky Volcano an ideal target to employ a multiparameter observational data set to help characterize eruptive behavior and constrain volcanic processes. 3. Methods [7] Field campaigns were conducted at Karymsky Volcano August 211 and 17 3 July 212. Volcanic activity was recorded using the following methods: (1) SO 2 emission rate was quantified using a scanning FLYSPEC ultraviolet (UV) spectrometer system [Horton et al., 26], (2) a NicAIR thermal IR camera was used to remotely detect and quantify both SO 2 and ash [Prata and Bernardo, 29], (3) a forward looking infrared (IR) radiometer (FLIR) thermal imaging camera was used to record high sample frequency thermal observations of the volcanic emissions and hot eruption products [Spampinato et al., 211], and (4) infrasound was recorded using National Center for Physical Acoustics (NCPA) digital 518

4 NicAIR FLIR FLYSPEC Figure 2. Photo of the 211 experimental setup. Remote-sensing instruments including IR camera, FLIR camera, and FLYSPEC (labeled) are seen in the foreground, with Karymsky Volcano seen in the background. microphones (Figures 1b and 2) [Fee and Matoza, 213]. All instruments were deployed within 4 km of Karymsky s summit (Figure 1b). Several factors contribute to measurement error for the various remote-sensing instruments such that throughout this study we focus on the relative changes in the observed parameters over time and with respect to the various activity types. On several days during the field campaigns sampling conditions were favorable such that accurate remotesensing measurements were acquired. Occasionally, strong winds blew the plume down the flanks of the edifice and directly toward the remotesensing instruments such that accurate data were not acquired. We used the FLIR and NicAIR imagery to provide a visual record of the emission activity, to assist in interpretation of the remote data sets, and to identify measurements collected under poor sample conditions for exclusion from analysis. The equipment and methods employed for each technique are described in the following sections and appendix, with instrument parameters listed in Table FLYSPEC UV Spectrometer System [8] A FLYSPEC scanning UV spectrometer system (Figure 2) [Horton et al., 26] was used to measure SO 2 column density within Karymsky s plume in an application of the Lambert-Beer law [e.g., Platt and Stutz, 28]. Measurements of UV absorption by the volcanic plume are fit to a calibration curve generated from field measurements of cells containing known concentrations of SO 2 viewed in front of the background (SO 2 -free) sky in the nm wavelength region to calculate SO 2 column density [Horton et al., 26]. Repeated SO 2 column density measurements (ppmm, where 1 ppmm SO g m 22 SO 2 )[Gerlach, 23] are collected in a series across the plume, perpendicular to plume motion to obtain SO 2 cross-sections, which are then Table 1. Instruments and Methods Employed in This Study Instrument Spectral Region Sample Frequency Parameter Detected Years Deployed Deployment Mode References FLIR A32 IR Camera lm 5 Hz Temperature 211, 212 Stationary recording [Spampinato et al., 211] NicAIR IR Camera 7 14 lm 5s SO 2 emission rate; SO 2 and ash mass 211 Stationary recording [Prata and Bernardo, 29] FLYSPEC UV nm 2 s to 3 min SO 2 emission rate 211, 212 Horizontal scans [Horton et al., 26] spectrometer Microphones.2 25 Hz 125 Hz (211) 25 Hz (212) Infrasound (Pressure) 211, Sensor array [Fee and Matoza, 213] 519

5 integrated over the plume width and multiplied by the plume speed to derive SO 2 emission rates (supporting information A 1 ). The FLYSPEC was deployed to collect SO 2 column density measurements via horizontal scans within 15 2 m above Karymsky s vent to allow minimal time delay between the SO 2 emission rates, infrasound, and thermal radiation energy produced. The plume scan duration, combined with the time to make an individual column density observation (dependent on UV radiation intensity) resulted in an SO 2 emission rate temporal resolution of 2 s to 3 min. Note that this relatively low temporal resolution can inhibit comparison with high-temporal resolution geophysical data sets, especially for the highly dynamic activity observed at Karymsky Volcano. [9] Total error in FLYSPEC SO 2 column densities and derived emission rates are estimated to be 64 and 656%, respectively, for ash-free plumes [Lopez et al., 213] (Appendix A). We note that the presence of ash in plumes is known to significantly scatter and attenuate radiation [Millan, 198; Andres and Schmid, 21], which can decrease the pathlength of UV radiation that passes through the plume, and result in an underestimation of SO 2 column density, though the actual amount of underestimation cannot be constrained with current techniques (C. Kern, personal communication, 212). Therefore, for SO 2 column densities and derived emission rates collected for plumes containing ash (e.g., ash explosions and explosive eruption activities described in sections 4.1 and 4.4), we expect values to be significantly underestimated relative to ash-free plumes, and depict these values with black-diagonal stripes to differentiate them from better constrained measurements. Additionally, poor signal-to-noise measurements (i.e., noncoherent plume shapes) interspersed with quality measurements (e.g., Gaussian curve plume shapes such as in Figures 4 and 8) are distinguished from low quality measurements by a gray color. More details of the FLYSPEC methods employed and error calculations can be found in Appendix A NicAIR IR Camera for Detection of Ash and SO 2 [1] An improved multispectral IR imaging camera (NicAIR) originally described in Prata and Bernardo [29] is used to calculate ash masses, 1 Additional supporting information may be found in the online version of this article. Relative absorption (SO 2, ash and ice) SO 2 absorption F1 (8.6 µm) F2 (1. µm) F3 (11. µ m) Ash absorption Ice absorption Wavelength (µ m) Figure 3. Relative absorption for SO 2 gas (red line), silicate ash (blue line), and ice particles (green line) as a function of wavelength. The dotted black lines show the response functions for three NicAIR filters. Note that the change in silicate absorption across the 1 12 mm window region is in the opposite sense to that of ice. as well as supplementary measurements of SO 2 column density, derived emission rates, and plume ascent speeds when FLYSPEC measurements were not available (Figure 2) (A. J. Prata and C. Bernardo, Retrieval of SO 2 from a groundbased thermal infrared imaging camera, submitted to Atmospheric Measurement Techniques, 213). The NicAIR uses a commercially available thermal IR camera core, with a pixel array detector, a field-of-view (FOV), and an uncooled microbolometer with good temperature sensitivity in the region of 8 12 lm. Three filters centered at 8.6, 1, and 11 lm with bandwidths from.5 to 1. lm, and a broadband filter with a bandwidth from 7 to 14 lm, are used to detect and quantify very fine ash (1 16 lm radii) and SO 2 column densities (g/m 2 ), exploiting their characteristic IR absorption/emission features and permitting discrimination from meteorological clouds using radiative transfer calculations (Figure 3) (Prata and Bernardo, submitted manuscript, 213). When all four filters are in operation, a maximum temporal resolution of one frame every 5 s is achieved. [11] The acquired time series image data of ash and SO 2 column densities are used to calculate plume ascent speeds, SO 2 emission rates, and ash masses. Abundant ash within plumes can significantly attenuate IR radiation, in some cases making the plume opaque and preventing SO 2 and ash column densities from being accurately retrieved. Therefore, in the cases of ash-rich plumes, we consider the calculated ash masses to be conservative minimum estimates, and only evaluate SO Filter response function 511

6 detected under ash-free conditions. Estimated error in SO 2 column density and derived emission rates are expected to be <5%, while errors in ash column density and derived masses may approach 1% in worst-case scenarios of ash-rich plumes. More details on the NicAIR methods employed can be found in Prata and Bernardo [29], Prata and Bernardo (submitted manuscript, 213), and Appendix B. [12] The NicAIR has several important advantages over the FLYSPEC, specifically it is able to collect measurements 24 h a day, and it provides a two dimensional image of the plume, such that SO 2 emission rates can be calculated at a much higher temporal resolution (1 per 5 s). Unlike the NicAIR, the FLYSPEC has the advantage of being an established technique with a relatively simple and computationally fast analysis scheme, and was available for use during both field campaigns. On at least one occasion (e.g., section 4.2) insufficient UV radiation prevented SO 2 detection by the FLYSPEC, while the NicAIR was able to detect and quantify SO 2. We conduct a preliminary comparison of SO 2 column density measurements collected on 16 August 211by both the NicAIR and FLYSPEC to ensure that these data sets are comparable, and find a good qualitative correlation. More details of the comparison methods and observed correlation can be found in Appendix C FLIR Thermal Imaging Camera [13] A FLIR model A32 thermal imaging camera was used to acquire 5 frames per second (fps) thermal imagery over a broadband wavelength region from 7.5 to 13 mm to estimate pixelintegrated temperature between 22 and 35 C from measured radiance (Figure 2) [e.g., Spampinato et al., 211]. This thermal camera has a FOV and uses a pixel focal plane array detector and an uncooled microbolometer. Pixel area was calculated using the detector resolution, along with sample slant distance of 378 m and a camera inclination angle of 17 that resulted in an image-center pixel resolution of m 2. Data were analyzed using FLIR ThermaCam Researcher Professional software, which uses distance to source, emissivity, ambient temperature, and ambient relative humidity (both measured using a hand-held thermometer/hygrometer in the field) along with the LOWTRAN radiative transfer model to convert measured radiance into temperature according to Planck s law [Spampinato et al., 211]. An emissivity value of.98 was to chosen to represent a mixed-phase plume composed predominantly of condensed water vapor and silicate ash assuming (1) a nontransmitting source such that emissivity reflectance [Spampinato et al., 211] and (2) the average reflectance of andesite and water from the ASTER Spectral Library for the 8 14 lm wavelength region [Baldridge et al., 29]. Uncertainties in the calculated temperatures typically result in derived temperatures being lower than atsource kinetic temperatures [Spampinato et al., 211], such that all temperatures reported here are considered estimates and data evaluation focuses on relative change. [14] Time series measurements of (1) maximum pixel temperature directly above Karymsky s vent, (2) thermal radiation energy released from individual explosions and/or degassing events [Marchetti et al., 29], and (3) cumulative thermal radiation energies were calculated to facilitate comparison between volcanic emissions and infrasound, and to help characterize the observed activity. Additionally, plume ascent speeds for time periods with corresponding FLYSPEC SO 2 measurements were calculated using FLIR image data, known distance to source, and plume parcel-tracking methods, with error estimated to be 63% [Williams-Jones et al., 28]. Details of these methods can be found in Appendices A and D Infrasound Data and Methods [15] During the 211 campaign, infrasound was continuously recorded at a 125 Hz sample rate using a four-element array of NCPA digital microphones with flat response between.2 and 25 Hz. The array was installed approximately 4 km southeast of Karymsky s summit vent in areas with moderate vegetation to reduce wind noise (Figure 1b). Microphones were distributed in a centered-triangle array to permit source azimuth and trace velocity (propagation velocity across the array) determination, allowing volcanic infrasound to be distinguished from noise and infrasound produced by other sources. During the 212 field campaign, five NCPA digital microphones were deployed in a more distributed, network-like configuration between 2 and 4 km northwest of the active vent (Figure 2). Each stand-alone digital microphone consists of a piezo-ceramic acoustic sensor, onboard digitizer, and GPS connected to a battery. The microphones were able to record pressure signals between 6125 and 675 Pa for 211 and 212, respectively. 5111

7 A: Discrete Ash Explosions.5.5 Pressure (Pa) :15: 2:2: 2:25: 2:3: 2:35: 2:4: 2:45: 2:5: 2 SO 2 ER (kg/s) Max Temp (C) FLYSPEC B: Pulsatory Gas Emissions.5.5 Pressure (Pa) :5: 21:1: 21:15: 21:2: 21:25: 21:3: 21:35: 21:4: 2 SO 2 ER (kg/s) Max Temp (C) NicAIR C: Gas Jetting.5.5 Pressure (Pa) Max Temp (C) 22:3: 22:35: 22:4: 22:45: 22:5: 22:55: 23:: 23:5: 2 SO 2 ER (kg/s) FLYSPEC D: Explosive Eruption.5.5 Pressure (Pa) Max Temp (C) 1:4: 1:45: 1:5: 1:55: 2:: 2:5: 2:1: 2:15: 2 SO 2 ER (kg/s) FLYSPEC Figure 4. Multiparameter observations of the four activity types. Subplots show data corresponding with each activity type, including: (a) ash explosions, (b) pulsatory degassing, (c) gas jetting, and (d) explosive eruption. Images on the left of the figure depict the visual characteristics of each activity type. Infrasound pressure (Pa; top), maximum temperature ( C; middle), and SO 2 emission rate (kg/s; bottom) subplots show respective data sets observed over each of the 4 min analysis periods. Width of SO 2 emission rate bars represents the duration of each scan and resulting emission rate. Black bars represent high signal-to-noise measurements, gray bars represent low signal-to-noise measurements such that values are not accurate, and blackdiagonal striped bars represent measurements that have a high signal-to-noise ratio but are likely underestimated due to the presence of ash. SO 2 emission rates for Figures 4a and 4d are likely underestimated due to the presence of ash. Additionally, SO 2 emission rates for Figure 4b were collected using the NicAIR IR camera, while remaining measurements were collected with FLYSPEC; and SO 2 emission rates for Figure 4c were obtained in 212. In Figure 4d, photos depicting both the quiescence prior to explosive eruption and the explosive eruption are shown. Note that scales are fixed for each parameter evaluated to facilitate data comparison. Some infrasound pressure and/or maximum temperature values exceed the upper scale limit in Figures 4a and 4d. 5112

8 LOPEZ ET AL.: CHARACTERIZATION OF VOLCANIC ACTIVITY Night Ash Explosions Fog Gas Jetting and Pulsatory Degassing Gas Jetting and Pulsatory Degassing Pulsatory Degassing Explosive Eruption Hybrid Pulsatory Degassing to Hybrid Noise Hybrid Day Hybrid Clouds Data Outage Pulsatory Degassing to Hybrid Hybrid Clouds Hybrid Data Outage Gas Jetting Gas Jetting to Hybrid Gas Jetting Gas Jetting Gas Jetting UTC Hour Figure 5. Infrasound-based timeline. Top eight traces each depict 1 day of infrasound data collected during the 211 field campaign, while the bottom trace represents one example day from the 212 field campaign. Sample date is labeled on the primary y axis, infrasound pressure up to 64 Pa is shown on the secondary y axis, and UTC hour is shown on the x axis. Periods of interest for this study are marked by colored rectangles; specifically the blue, gray, and yellow rectangles (labeled) represent ash explosions, pulsatory degassing, and gas jetting analysis periods, respectively. The red rectangles show the six explosive eruption events detected by infrasound in 211, with the event analyzed labeled on 17 August. Visual observations of activity type for certain time periods, or other parameter of interest, are labeled above the infrasound trace when available. [16] To quantify the eruption energetics at Karymsky Volcano, and for comparison with volcanic emissions data at other volcanoes, we calculate the acoustic energy and reduced infrasonic pressure for individual eruptive events, and cumulative infrasound energy over specified time periods following the methods described in Johnson and Ripepe [211] and Fee and Matoza [213]. Spectrogram estimates are obtained using Welch s modified periodogram method and are used to characterize different styles of activity. More details of these methods can be found in Appendix E. 4. Multiparameter Characterization of the Observed Activity [17] Four types of volcanic activity were observed during the field campaign, which we define in the approximate order in which they were first observed as: (1) ash explosions, (2) pulsatory degassing, (3) gas jetting, and (4) explosive eruption. Activity representing a mix or hybrid of ash explosion, pulsatory degassing, and gas jetting activity was also seen, and frequently followed explosive eruptions and/or occurred during transitions between activity types. No lava effusion was observed during either field campaign, in contrast to field campaigns in [Johnson and Lees, 2]. Additionally, no aerial observations of the vent were obtained during either field campaign to constrain vent geometry and/or identify possible lava extrusion within the vent. During the 211 field campaign, all four activity types were observed, including multiple hours of ash explosions, pulsatory degassing, and gas jetting, along with frequent periods of hybrid activity, and seven explosive eruption events, reflecting an active and 5113

9 A) B) C) D) Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz) :15: 2:2: 2:25: 2:3: 2:35: 2:4: 2:45: 21:5: 21:1: 21:15: 21:2: 21:25: 21:3: 21:35: 22:35: 22:4: 22:45: 22:5: 22:55: 23:: 23:5: 1:4: 1:45: 1:5: 1:55: 2:: 2:5: 2:1: 2:15: Amplitude (db) Amplitude (db) Amplitude (db) Amplitude (db) Figure 6. Infrasound spectrograms for the four activity types, including (a) ash explosion, (b) pulsatory degassing, (c) gas jetting, and (d) explosive eruption. The y axis shows the frequency in Hz. The color bar represents signal amplitude in db. The x axis shows UTC time. Note the frequency range for Figures 6a, 6b, and 6d are between.1 and 2 Hz (same scale), while Figure 6c is between.1 and 6 Hz (extended scale). A helicopter signal at 15 Hz can be seen in D from 1:38 to 1:4. dynamic system. During this time period, all four activity types were clearly detected acoustically and several were observed by the various remotesensing instruments. In contrast, the activity in 212 was dominated by gas jetting, with only a few minor ash explosions observed. We select four 4 min time periods that best demonstrate the four activity types, and have the most complete multiparameter observations. In the following sections, we characterize the four activity types using observations of ash mass, SO 2 emission rate, temperature, thermal radiation energy, reduced infrasound pressure, acoustic energy, and plume ascent speeds. We report values for maximum, mean, and one standard deviation above/below the mean, when possible. Figure 4 shows representative visual images, infrasound pressure, maximum temperature, and SO 2 emission rates for each of the selected activity types. An infrasound-based timeline depicting data and activity from the 211 field campaign along with one example day from the 212 field campaign can be seen in Figure 5. Infrasound spectrograms for each activity type are displayed in Figure 6, permitting frequency-based interpretations. Table 2 summarizes the observations for the various activity types Ash Explosions [18] Activity referred to as ash explosions was visually characterized as consisting of discrete ash explosion occurring every 4 min on average, that either jet or roil out of the vent, and produce plumes to 5 15 m (above-vent) altitudes (Figure 4a). Explosions were often accompanied by an audible crack at the onset. The 4 min example period representing typical ash explosion activity spans from 2:1 to 2:5 (all times as UTC) on 15 August 211. During this time period, 26 events were observed in the FLIR imagery, with a maximum event duration of 4.3 min and a mean duration of min. In most cases, events were discrete, with periods of quiescence up to 1 s between events, though on approximately seven occasions new events began before the prior event had ceased. Maximum and mean peak 5114

10 Table 2. Multiparameter Characteristics of the Observed Activity Description Ash Explosions Pulsatory Degassing Gas Jetting Explosive Eruption Date (UTC) 15 Aug Aug Jul Aug. 211 Time (UTC) 2:1 2:5 21: 21:4 22:3 23:1 1:38 2:18 Temperature ( C) Max Peak 5 23 Max Peak 5 16 Max Peak 5 8 Max Peak >35 Mean Peak Mean Peak Mean Peak Thermal radiation Max a Max Max Max > energy (J) Mean a Mean Mean Cumulative thermal a radiation energy (J) Ash mass (kg) NA None observed None observed >69, b SO 2 emission rate (kg/s) Max Mean Max c Mean c Max Mean Max Mean SO 2 mass (kg) NA Max c NA NA Mean c Ascent speed (m/s) Max 59. Mean Max d 5 8. Mean d Max Mean Max Mean Infrasound onset Impulsive Emergent Emergent Impulsive Reduced infrasound Max 521,5 Max 55 Max 516 Max >5, pressure (Pa) Mean Mean Mean Acoustic energy (J) Max Max Max Max > Mean Mean Mean Cumulative infrasound energy (J) One standard deviation above/below the mean values are shown. a Background clouds are biasing thermal radiation measurements high for ash explosion activity. b This value is from an explosive eruption event on 22 August 211 at 8:21 UTC, and is used as a proxy for the ash mass of the explosive eruption presented here. c These measurements were collected using the NicAIR, all other SO 2 emission rates from FLYSPEC. d Ascent speeds calculated from the NicAIR. All other ascent speeds from FLIR. temperatures of 23 and C, along with maximum and mean thermal energies for individual events of J and J, were observed during this time period. Maximum and mean SO 2 emission rates of >1.3 kg/s and > kg/s, respectively, were calculated for this time period from FLYSPEC data. Unfortunately, only two out of four potential NicAIR filters were selected at this time (8 lm and broadband), such that ash mass retrievals were not possible. Maximum and mean plume ascent speeds of 9. and m/s were calculated using parcel-tracking methods with FLIR data. Infrasound signal onsets were impulsive, with individual events producing maximum and mean reduced infrasonic pressures of 21,5 Pa and Pa, respectively. Infrasound from these events consisted of relatively highamplitude, sustained signal coincident with visible ash jetting. The infrasound onsets were broadband in frequency (.1 2 Hz), followed by jetting focused between.1 and 5 Hz (Figure 6 and Table 2). Peaks in both temperature and infrasound pressure occurred nearly coincidentally for most events of this time period, though no obvious correlation in peak amplitude between these data sets was observed. Cumulative acoustic energy and cumulative thermal radiation energy for this analysis period were J and J, respectively Pulsatory Degassing [19] Pulsatory degassing activity consisted of individual pulses of volcanic gas emissions with little or no ash, low altitude (1 2 m) plumes, and no audible sound at a distance of 4 km. The selected time period representing typical pulsatory degassing activity was 16 August 211 from 21: to 21:4 (Figures 4b and 5). During this time period, 17 degassing pulses were identified in the FLIR imagery, with at least three of these truncated by subsequent events. Mean event durations of 1.5 min were shorter than periods of quiescence between events (2.9 min). Maximum and mean peak temperatures of 16 and C, and mean thermal radiation energy of J were observed during this time period. Poor plume geometry (blown-over plume) from relatively strong winds prevented SO 2 emission rate calculations from FLYSPEC data using the horizontal scan geometry. However, vertical SO 2 cross-sections made down-wind through the blown-over plume were estimated by the NicAIR IR camera and distinct pulses in SO 2 emission rates with a mean value of kg/s were 5115

11 observed. No obvious temporal correlation between SO 2 emission pulses and infrasound pressure and/or maximum temperature were apparent, likely due to the temporal delay between gas emission at the vent and SO 2 measurement down-wind (Figure 4b). The infrasound signals for pulsatory degassing pulses have emergent onsets and cigarshaped codas that taper at both ends (Figure 4b). Maximum and mean peak reduced infrasound pressures were 5 Pa and Pa, and maximum and mean acoustic energies were J and J, respectively. Dominant infrasonic frequencies were between 1 and 2 Hz (Figure 6). Plume speeds, in this case reflecting horizontal motion of the blown-over plume, were calculated from NicAIR imagery with maximum and mean values of 8. and m/s (Table 2). Cumulative acoustic energy and thermal radiation energy for this analysis period were J and J, respectively Gas Jetting [2] Gas jetting activity consisted of continuous degassing, with pulses of more vigorous degassing overprinting the background emissions and producing strong audible jetting or roaring. Like pulsatory degassing, little or no ash emissions were associated with this activity type (Figure 4c). While gas jetting activity was observed during the 211 field campaign (most easily recognized by its audible roar), poor sampling conditions including ground-hugging plumes and/or plumes traveling directly toward the instruments prevented accurate temperature and SO 2 retrievals in 211, therefore, we use measurements collected under favorable conditions in 212 for this analysis and assume that these measurements are representative of typical gas jetting activity. We consider this assumption to be valid for comparing changes in relative SO 2 emission rates with other data sets; but caution that the absolute SO 2 emission rate values may be significantly different between the 2 years due to potential changes in deeper processes such as magma supply, degassing depth, and conduit permeability. The time period selected to represent typical gas jetting activity was 21 July :3 to 23:1. We selected event durations for this activity type based on periods of more vigorous degassing, as determined using FLIR maximum temperature waveforms, and find 16 pulses during the 4 min analysis period (Figure 4c). Pulses of more vigorous degassing had an average duration of 3.5 min. Periods of weaker degassing between pulses were relatively short with an average duration of 2 s. Vigorous degassing produced relatively high maximum and mean ascent speeds (12.6 and m/s, respectively). Maximum and mean peak temperatures associated with gas jetting were 8 C and C, respectively, and the mean thermal radiation energy produced was J. Infrasound signals were emergent exhibiting distinct cigar-shaped amplitudes, with maximum and mean reduced infrasound pressures of 16 and Pa, respectively. Mean acoustic energies were J. Gas jetting activity exhibited a strong component of high frequency sound; in fact the signal was focused above the infrasound band into the audible region between 15 and 6 Hz (Figure 6). A general temporal agreement was observed between waxing and waning of infrasound pressure and maximum temperature (Figure 4c). Cumulative acoustic and thermal radiation energy for this analysis period were J and J, respectively Explosive Eruption [21] Explosive eruption activity was characterized by periods of relatively long-duration (3 min to >1 h) quiescence with no visible emissions, followed by an explosive eruption producing ash-rich plumes to >2 m and centimeter to meter (or greater) sized ballistics that rolled down the flanks of the edifice (Figure 4d). Explosive eruptions viewed at night showed abundant incandescent material that mantled the edifice following eruption, indicative of the involvement of juvenile magma. Seven explosive eruption events were observed during the 211 field campaign and we select 17 August 211 from 1:38 to 2:18 as the representative time period. This eruption was preceded by 4 min of quiescence during which emissions were not visible in FLIR imagery (no ascent speeds calculated), but SO 2 was clearly detected by the FLYSPEC resulting in an average SO 2 emission rate of kg/s (assuming the average ascent speed of 5.4 m/s calculated for the preceding hour) (Figure 4d). No infrasound or elevated temperatures were detected leading up to the eruption. The explosive eruption commenced at 1:55 and had a duration of 1.7 min, at which point the initial eruption was truncated by a secondary eruption according to the thermal imagery and infrasound data sets (Figure 4d). The initial explosive eruption, as well as two secondary eruption pulses, reached peak temperatures in excess of 35 C, the FLIR temperature saturation range. 5116

12 Average temperatures for this time period are not considered as the actual explosive eruption only comprises a small portion of the 4 min analysis period. The eruption produced a very high amplitude, impulsive infrasonic pressure signal, which clipped the microphones at 6125 Pa at a distance of 4 km, and resulted in a reduced infrasound pressure of >5 KPa. The estimated acoustic energy for this event was > J. Abundant ash produced in the first few minutes of the eruption decreased UV signal-to-noise such that SO 2 was not detected (gray bars in Figure 4d). Following this time period, SO 2 cross-sections were clearly identified, allowing SO 2 emission rates to be calculated; however, these values are likely underestimated due to the presence of ash in the plume. The maximum and mean SO 2 emission rates calculated for the entire analysis period were 2.3 kg/s and kg/s, respectively. Unfortunately, only two of four NicAIR IR camera filters were being used during this time period, preventing the retrieval of ash masses for this eruption. A maximum plume ascent speed of 178 m/s was calculated within the first second of the explosive eruption, using the distance traveled by the plume s leading edge as seen in consecutive FLIR images. Cumulative acoustic and thermal radiation energy for this analysis period were J and J, respectively. These observations are summarized in Table Discussion 5.1. Comparison of the Multiparameter Observations for the Four Activity Types [22] In this section, we (1) compare the multiparameter character of the four activity types, (2) distinguish temporal correlations among the multiparameter observations within individual activity types, (3) make inferences on volcanic activity based on these observations, and (4) identify unique data trends to enable remote detection and discrimination of various styles of volcanic activity Ash Explosions [23] Ash explosions exhibited the second highest maximum and mean temperatures, reduced infrasound pressure, acoustic energies, and cumulative acoustic energies of the four types (Figure 4a and Table 2). This suggests that ash explosions represent the second highest energy events, following explosive eruption activity. High peak temperatures and thermal radiation energies were observed during ash explosions compared with those for pulsatory degassing and gas jetting activity (Table 2), which is consistent with visual observations that Karymsky s ash explosions emit a greater quantity of fragmented material (Figure 7) [Marchetti et al., 29]. The cumulative acoustic energy observed for ash explosions (1 7 J) is two orders of magnitude greater than that of the degassing activity observed at Karymsky Volcano (1 5 J), and suggests a relatively high conduit overpressure during ash explosions, which may lead to a high magma fragmentation level. We contend the conduit is essentially sealed prior to ash explosions and that significant acoustic energy is rapidly transferred to the atmosphere when the confining pressure is released during explosion. While no ash masses were obtained for this activity type we can assume that the ash mass for these events will be greater than for the degassing activity types and smaller than the explosive eruption events, according to visual observations. Infrasound signal onsets for ash explosions are broadband and impulsive, with the second highest acoustic energy and mean reduced infrasound pressure observed. The high-amplitude infrasound coda is distinct and corresponds with a longduration gas-thrust phase for each event. Variations in SO 2 emission rate during this time period were observed, however, no correlations between SO 2 emission rate and maximum temperature or infrasound pressure were apparent, likely the result of poor temporal resolution SO 2 measurements collected under low UV (early morning) conditions. We note that the thermal and infrasonic signals observed for ash explosions are sufficiently unique such that it would be possible to distinguish this activity type from the other activity types using either one of these data sets, as well as draw conclusions on the style of activity and associated products Pulsatory Degassing and Gas Jetting [24] Pulsatory degassing and gas jetting activity show several similarities to each other, and share several differences when compared to the other activity types (Figures 4b and 4c and Table 2). Temperatures, SO 2 emission rates, and reduced infrasound pressures all had similar mean values between pulsatory degassing and gas jetting, and significantly larger maximum values for pulsatory degassing relative to gas jetting. Similar order of magnitude cumulative acoustic energies were observed for these types of activity, which were two to four orders of magnitude lower than observed for ash explosions and explosive 5117

13 Thermal Radiation Energy (J) Santiaguito Fuego Stromboli Ash Explosions Pulsatory Degassing Gas Jetting Explosive Eruption Hybrid Acoustic Energy (J) Figure 7. Thermal radiation energy (J) and acoustic energy (J) for each activity type. Distinct clusters among the various activity types can be seen. The mean thermal and acoustic energies calculated by Marchetti et al. [29] for other volcanoes that exhibit small explosive eruption behavior are also shown. We note that explosive eruption thermal radiation and acoustic energies are underestimated relative to other types due to instrument saturation and field-of-view limitations; while thermal radiation energy associated with ash explosions is overestimated relative to other types due to the presence of background clouds in the thermal imagery. eruption. Acoustic onsets were emergent for both types, which suggests that both degassing styles occurred under conditions of low vent overpressure, which we attribute to relatively permeable conduit conditions. On several occasions, minor quantities of hot volcanic bombs (<5 within the 4 min analysis period) were observed visibly and/or in FLIR thermal imagery to be erupted during the pulsatory degassing study period that were not observed during the gas jetting study period. We speculate that the significantly higher thermal energies observed for pulsatory degassing relative to gas jetting could be explained by the eruption of minor quantities of volcanic bombs, which is consistent with observations of degassing at Karymsky Volcano in 28 by Lopez et al. [211]. We caution, however, that the emission of minor quantities of volcanic bombs during gas jetting activity may also occur, but were not observed in the 4 min study period we selected. Overall the temperatures and infrasonic pressures associated with the degassing types are significantly lower than observed for the other activity types involving the eruption of abundant pyroclastic material. We note that the primary difference between pulsatory degassing and gas jetting activity as observed from our data set is in the acoustic frequency content. In particular, little to no 1 Pressure (Pa) SO 2 Emission Rate (kg/s) FLYSPEC :43 :57 1:12 1:26 1:4 1:55 2:9 2:24 211/817 UTC Time Figure 8. Infrasound pressure and SO 2 emission rates surrounding an explosive eruption.so 2 emission rates are elevated (>2 kg/s) and pulsatory approximately 1 h before the eruption at 1:55. In the 2 min before the eruption, SO 2 emission rates decrease to.14 kg/s, but remain detectable through the eruption onset. A large peak in both infrasound pressure (>125 Pa) and SO 2 emission rate corresponds with the initial eruption onset, however, abundant ash in the minutes following prevented accurate SO 2 retrievals. Following the explosive eruption pulses of SO 2 emission rates and infrasound can be seen. Poor signal-to-noise emission rate data are colored gray, ash-rich measurements are marked with black diagonal lines. 5118

14 infrasound (<2 Hz) was detected during gas jetting, rather it was associated with an audible roar and exhibited abundant high frequency acoustic energy (>2 Hz; Figure 6 and Table 2) that was not as evident during pulsatory degassing. Noise from volcanic jets has been observed at a range of frequencies [e.g., Woulff and McGetchin, 1976; Matoza et al., 29], and more work is required to determine the physical implications of the high frequency acoustic signals. We propose that pulsatory degassing and gas jetting activity can be distinguished from ash explosions and explosive eruption using the infrasound data according to their emergent onsets and significantly lower reduced infrasound pressure and acoustic energies. Pulsatory degassing and gas jetting can then be distinguished from each other based on their acoustic frequency content Explosive Eruption [25] Explosive eruption activity is distinct from other activity types with respect to all parameters evaluated and is clearly identified using infrasound and/or thermal data sets. In particular, saturated peak temperatures (>35 C), very high reduced infrasound pressures (>5 KPa), and ascent speeds up to 178 m/s, were all significantly larger than observed for other activity types. The infrasound onset was extremely impulsive and high pressure, saturating the sensors with pressures in excess of 125 Pa at a distance of 4 km (P red > 5 KPa). These pressures were much greater than those observed in previous infrasound studies at Karymsky Volcano [Johnson et al., 24b] and similar order of magnitude to values observed at Tungurahua Volcano, Ecuador (P red 5 15 KPa, at 1 m) [Fee et al., 21; Johnson and Ripepe, 211]. The acoustic energies for explosive eruption activity are significantly higher than observed for ash explosions, as well as in comparison to other volcanoes that exhibit small explosive eruptions (e.g., Fuego and Santiaguito Volcanoes, Guatemala; Figure 7) [e.g., Ripepe et al., 25a; Marchetti et al., 29; Johnson and Ripepe, 211] and suggests a high vent overpressure and/or a high degree of magma fragmentation [Fee and Matoza, 213]. The relatively high cumulative thermal radiation energy values in comparison with Karymsky s degassing activity types, as well as with other volcanoes, likely indicates the eruption of either a greater quantity of material or more highly fragmented material (Figure 7) [Ripepe et al., 25; Marchetti et al., 29]. Similarly, the maximum temperature observed for explosive eruptions was over 1 C greater than observed for ash explosions, which may indicate the involvement of hotter material or the involvement of relatively more abundant hot material. Unlike the other data types, clear trends in SO 2 emission rates can be seen with volcanic activity during the explosive eruption analysis period. Some of the highest SO 2 emission rates observed during the field campaigns (2.3 kg/s) were seen 75 min prior to the type example explosive eruption (Figure 8), though we caution that this may be biased by the fact that SO 2 is likely underestimated during ash-rich activity. These values dropped significantly in the 2 min prior to eruption down to.14 kg/s. These low but clearly detectable SO 2 emission rates corresponded to apparent vent sealing during which emissions were not visible by eye or in the thermal imagery and no infrasound was detected from the volcano. This observation supports previous findings by Fischer et al. [22] that Karymsky s vent does not seal entirely prior to eruptions. During the initial explosive eruption significant ash attenuation prevented accurate SO 2 measurements; however, after 1 min, SO 2 could again be detected and exhibited a temporal correlation with peaks in infrasound and thermal radiation energy, corresponding with secondary eruptions (Figure 4d). It appears that the SO 2 emission rates increased 1 min prior to peaks in infrasound and thermal radiation energy associated with the secondary eruption pulses following the initial eruption, though this could be an artifact of ash dissipation allowing SO 2 to be more accurately detected (Figure 4d; 2: 2:15 UTC). This increase in SO 2 emission rate prior to secondary explosions could alternatively result from a reduction in conduit pressure following each eruption which induces volatile exsolution [Carroll and Webster, 1994]. Improved temporal resolution SO 2 emission rate data, and/or more accurate constraints on absolute SO 2 column density values in the presence of ash, are required to better relate trends in volcanic emissions to the dynamic volcanic activity observed at Karymsky Volcano. We note that while ash masses could not be calculated for this explosive eruption event, an ash mass of >69, kg was estimated for a different explosive eruption event at 8:21 on 22 August 211 using the NicAIR camera (Figure 9). In this analysis, the plume remained partially opaque such that we consider this ash mass to be a minimum estimate and note that these measurements are limited to ash particles in the size fraction of 1 16 lm (i.e., very fine ash), representing only a portion of the pyroclastic material emitted during an explosive eruption. 5119

15 22 August 211 8:21 UTC; Total ash mass >69.2 t Vertical distance (km) Ash mass loading (g m -2 ) Horizontal distance (km) Figure 9. Ash mass retrieval for an explosive eruption. Data acquired at 8:21 on 22 August 211. Vertical and horizontal distances are labeled on the y axis and x axis, respectively. The color bar scale shows the ash column density (g m 22 ) for each pixel. The retrieved ash mass for this image is >69, kg. Opaque portions of the plume (seen as low ash column density regions within the plume center) prevent accurate mass retrievals such that this mass should be considered an underestimate Proposed Models Previous Models [26] Previous studies have used observations of seismicity, infrasound, and volcanic emissions data from Karymsky Volcano to infer shallow vent processes [Johnson and Lees, 2; Fischer et al., 22; Lees et al., 24; Johnson, 27]. These studies largely proposed increased gas pressure combined with vent sealing as the eruption trigger mechanism [Johnson et al., 1998; Fischer et al., 22; Ozerov et al., 23]. We note that activity at Karymsky described in these studies was slightly different than observed during our field campaigns in 211 and 212. In particular, observations from field campaigns in 1997 and 1999 were dominated by small, explosive, ash-rich events (similar to our ash explosions) that occurred 6 times per hour with impulsive infrasonic signals [e.g., Johnson et al., 1998], and gas chugging events that were associated with waxing and waning SO 2 emission rates [Fischer et al., 22], as well as distinct acoustic and seismic waveforms and frequency content [Johnson et al., 1998; Johnson and Lees, 2; Lees et al., 24]. Fischer et al. [22] used high-temporal resolution SO 2 emission rate measurements (1 per 5 s) to identify decreases in SO 2 emission rates prior to ash explosions, followed by increases in SO 2 emission rates following the explosions. Additionally, Johnson et al. [1998] identified impulsive infrasonic signals at the time of explosion, both of which support the vent-sealing eruption trigger. Ozerov et al. [23] proposed a model in which the conduit is topped by a relatively short length, high viscosity magma plug, with the remaining conduit filled by a low viscosity, compressible magma that is continuously fed from depth. Dense, angular, ash, and bombs, along with scratched bomb surfaces interpreted to have formed during extrusion, support a highly viscous and volatile depleted magma in the upper conduit [Johnson et al., 1998; Fischer et al., 22; Ozerov et al., 23]. In their model, Ozerov et al. [23] proposed that the two types of activity observed from 512