1
IAVWOPSG/4-IP/6/ IAVWOPSG/4-IP/6
5/8/08
INTERNATIONAL AIRWAYS VOLCANO WATCH OPERATIONS GROUP (IAVWOPSG)
FOURTH MEETING
Paris, France, 15 to 19 September 2008
Agenda Item / 6: / Development of the IAVW6.5: / Development of eruptions source parameters (ESP)
Follow-up on IAVWOPSG/3 CONCLUSION 3/25:
ASHE Infrasound Study at Tungurahua, Ecuador
(Presented by Canada)
SUMMARYThis information paper reports on the findings of the ASHE Infrasound Project in Ecuador and presents some case studies. It was prepared by David Fee, Milton Garces, and Andrea Steffke at the University of Hawaii at Manoa.
SUMMARY
The primary objective of the Acoustic Surveillance for Hazardous Eruptions (ASHE) project was to demonstrate the possibility of using low frequency acoustic (infrasound) observations over broad regions for automatic, low-latency volcanic eruption notification to the Washington DC VAAC. Our results indicate the primary objective can be met for eruptions with a Volcanic Explosivity Index(VEI) of 2 or greater at distances of 40-250 km. For a station distance of 40 km, it appears possible to identify the acoustic signature of stratospheric ash injection and notify the VAAC within 5 minutes. At 250 km, notification latency grows to ~20 minutes. As confirmed during the July 12, 2008 eruption of Okmok Volcano, for the larger eruptions the results of this study may be extended to the 2000+ km range coverage provided by the IMS network.
Our eruption detection algorithms were trained using the Vulcanian to Plinian eruptions of Tungurahua Volcano, Ecuador. For VEI 2 or greater, a sustained volcanic signal was found to be a better indicator of substantial atmospheric ash injection than transient explosions. The eruptions on 5/12 2006, 7/14-15 2006 and 8/16-17 2006 were used as case studies. The ash-poor eruption of May 06 eruption served as an interesting example of continuous harmonic tremor punctuated by powerful explosions, but with little airborne ash. The July and August 2006 eruption erupted significant ash into the atmosphere which posed a serious threat to aviation in the region. The detection algorithms developed during the 2006 case studies were used to automatically notify the VAAC of large explosions and significant ash injection preceding and accompanying the 2/6/08 Tungurahua eruption, which lofted ash to ~15 km. Extensive work went into integrating the acoustic records with detailed analyses of satellite imagery for these events.
A very promising result of this project is the recognition that for large eruptions there may be a good correlation between acoustic energy and ash height. Further, there appears to be a potential to use coherent acoustic energy in the 0.05-0.25 Hz band as a discriminant for stratospheric ash injection (Figure 4; Figures 10-11). Careful comparisons between satellite and calibrated acoustic data are needed to confirm this relationship on a broader basis. We recommend this project be continued to assess the feasibility of using ASHE and IMS arrays to automatically identify eruptions with a VEI of 2 or greater at regional distances, and possibly smaller eruptions at closer distances.
Brief ASHE Timeline
February 2006:
Arrays deployed in Ecuador
March 2006:
Data received at ISLA (obtained missing data from mid-Feb)
April 2006:
ASHE data product website goes online at: Automatic explosion detection algorithm running. Website initially updated every 4 hours, then soon after updated every 1 hour.
March 8th, 2007:
Daily automatic email updates began summarizing Tungurahua volcano infrasound for the preceding 24 hours.
December 7th, 2007:
Website is updated every five minutes, and explosion and tremor energy
notification emails are sent immediately after detection by algorithm.
February 6th, 2008:
Email notification of onset, escalation, and end of significant eruption at Tungurahua
Per month Infrasound Citations by the Washington VAAC
March 07:6
April 07:5
May 07:3
June 07:1
July 07:11
August 07: 7
September 07: 4
October 07: 6
November 07: 10
December 07: 22
January 08: 24
February 08: 17
March 08: 4
Number of Tungurahua Explosions Detected at RIOE (37 km)from 2/6/06 to 6/10/08:
20,124
Number of Tungurahua Explosions Detected at LITE (251 km) since 2/6/06*:
>3500
*Note there have been extensive outages at LITE during the deployment that has decreased the number of detected explosions. LITE operations ceased in April 2008.
Infrasound Detection Methods
A combination of array and signal processing techniques were used to detect explosions and sustained signal (tremor and/or jetting) from Tungurahua. For the first portion of the experiment the processing techniques were run every hour, and then eventually shortened to every 5 minutes for station RIOE. In order to differentiate between signal and noise, the Progressive Multi-Channel Correlation (PMCC) array processing method was ran for both the RIOE and LITE arrays between 0.5-4 Hz, as this is the band where the majority of the volcanic signal was concentrated. To maximize signal detection, the processing was done using 10 second windows with 90% overlap, and 20 frequency bands.
The acoustic source energy (or acoustic energy) was calculated both for explosions and tremor to determine how energetic the volcanic signal was. The acoustic energy of a far-field source in free space can be estimated using:
EAcoustic=2πr2/ρc∫ΔP(t)2dt
where r=source-receiver distance, ρ=air density, c=sound speed, and ΔP=change in pressure. To better illustrate the volcanoes energetics, acoustic energy was converted to acoustic power by dividing by the energy value by the time interval. If not otherwise noted, the acoustic power levels quoted in this document were calculated over 15 minute time intervals. One caveat with calculating the acoustic energy is that the calculation assumes the change in pressure (ΔP) is entirely produced by the source of interest, and does not account for contributions from noise or clutter (other coherent sources). These unwanted contributors can often dominate, particularly in the far-field or during noisy periods. To minimize the effects of wind noise, the acoustic energy was calculated above 0.5 Hz. In addition, the acoustic energy was only calculated if the array processing (PMCC) results for that time period corresponded to significant acoustic signal arriving from within ±7° of Tungurahua (26°-40°). Using this method the acoustic energy for sustained signal (tremor or jetting) was estimated for a specified interval (Figure 1). If a high level of acoustic energy was detected over a given time interval, then a notification email was sent to the VAAC and other interested parties. Thus far, 1x108J over 1 hour is the selected energy threshold. If the energy level doubled, another notification was sent out. Once the energy level dropped below the threshold value, a final email notification was sent.
Figure 1. Acoustic energy calculation for the 2/6 eruption. Infrasound array data is high-pass filtered between above 0.1 Hz and split into 60 sec windows. The top panel shows the correlation value from PMCC for azimuths within +/-7 degs from Tungurahua. If the correlation value is above 0.6, the data segment is considered coherent and is colored black. The middle panel shows the filtered waveform. If the section is red, the signal correlation value is too low or its azimuth is off the target beam. If it is black, it is coherent and arriving from the Tungurahua azimuth, and the incident acoustic source energy is estimated. The bottom panel shows the cumulative acoustic source energy for the time interval.
The explosion detection algorithm for Tungurahua initially high-pass filters the data above 0.5 Hz and runs a STA/LTA (Short term average/Long term average) to determine the onset and end time of any impulsive signal. Two STA/LTA ratios are used, 2/5 and 3/40 seconds, to ensure both impulsive and somewhat more emergent explosions are detected. The explosion must also be recorded on all four channels. The explosion times are then correlated with the PMCC detections to ensure the signal is arriving from an azimuth within ±7° of Tungurahua. The coinciding PMCC family must have a minimumRMS amplitude >0.02 Pa and family size >15 pixels. If these conditions are met, then an explosion is registered and the time, duration, maximum raw amplitude, and acoustic source energy are recorded. The acoustic source energy is calculated for the explosion duration, and is then normalized by the energy of a reference event. The reference event had a peak-peak amplitude of ~1 Pa and ~5.65x108 J. By normalizing the explosion energy, we minimize source geometry, propagation, and topographical effects. Once the algorithms were operational, all explosions with an energy ratio greater than 5 caused a notification email to be sent out.
Infrasound Monitoring Limitations
During periods of low activity or high noise, the infrasound signal-noise levels are such that the employed algorithms are not nearly as effective. Since the arrays have insufficient wind noise reduction, the recordings during the middle of the day (~5 hours) are often saturated with wind noise and not useful for detecting low-level signals. However, during big explosions and the three large eruptions, signal levels were high enough to overcome the ambient wind noise. More regular maintenance of the arrays would also help raise the detection thresholds by making sure the sensor and site responses did not vary.
False alarms were fairly minimal, as the algorithms employed were correlated extensively with volcanological observations in case studies before notification services were initiated. Thunderstorms were one source of false alarms and may be misidentified as explosions, but not as tremor events. The ASHE project was also greatly assisted by the exhaustive and detailed monitoring of Tungurahua Volcano by the Ecuadorian Instituto Geofísico (IG), as well as their willingness to assist the ASHE project.
Large explosions can bias the acoustic energy/powere calculations, and a more effective way of separating explosions and tremor levels would be helpful. However, the relationship between explosions and tremor is not well understood. For example, large explosions followed by harmonic, gliding tremor was representative of gas-rich, ash-poor eruptions (e.g. May 06), and multiple episodes of jetting were preceding (and initiated?) by large explosions (e.g. towards the end of the 2/6/08 eruption, beginning of 7/14 eruption, and around 0435 on 8/17).
Case Study Observations and Correlations
ERUPTION CASE STUDY: 2008/2/6
Main Eruption: 0400-1430 UTC, Total Duration ~10.5 hours, VEI 2-3 estimated from ash heights. Refer to Figure 2 for a spectrogram, power levels, and ash heights for the 2/6 eruption.
Figure 2. 2/6/08 Eruption Infrasound. Top) Spectrogram between 0.1-10 Hz. Bottom) Black lines denotes acoustic power calculated every 15 mins. Green lines indicate the max and min inferred ash heights from satellite data. The red lines show the four infrasound notifications sent out, while the brown lines show the eight explosion notifications. The cyan diamond is the VAAC estimated ash height derived from satellite data and coarse atmospheric models. Reanalysis using the G2S atmospheric specifications (in progress) will reduce height uncertainty.
Brief Description of Eruptive Activity
According to news articles, on Feb. 6th, 2008 several hundred to 2,000 people were evacuated as pyroclastic flows and heavy ash fall inundated the areas surrounding Tungurahua. Based on information from the IG and satellite imagery evaluation, the Washington VAAC reported that ash plumes rose to estimated altitudes of 7.3-14.3 km (24,000-47,000 ft) a.s.l. and drifted S and NW. Ashfall was reported in areas downwind and to the SW and W, including Riobamba (30 km S). Precursory seismicity saturated local stations and presented similar patterns seen prior to intense episodes in July and August 2006 (source: Smithsonian's Global Volcanism Program).
The event began at 0420 UTC on February 6th, when tremor increased abruptly at all seismic stations. The tremor increased steadily until it reaches a maximum at 0611 UTC. The tremor then falls, until 0943 UTC where it begins again and continues until 1436 UTC, where it drops off rapidly. During these episodes of tremor great explosions took place, as well as very intense roars, felt earthquakes (including in the OVT), and pyroclastic flows that descended past the head of the gorges from Juive, Mandur and Cusúa. The most extensive were further east, and they flowed into the upper part of Cusúa. The emission columns were continuous; with a significant ash load extending to 10 km of height according to VAAC data. After this eruption, an activity of continuous emissions with ash contents of moderate-high. Further explosions were also registered, several of which were of great intensity. (source: IG).
ASHE notification and chronology:
Four activity (acoustic energy) updates were sent on 2/6. Below we summarize the infrasound activity, why the notifications were issued when they were, and list the relevant notification information.
0200-0400 UTC: Numerous small explosions, followed by larger explosion at ~0420 UTC. Jetting (high amplitude tremor) begins around 0425 UTC. Tremor between 0.5-1.5 Hz increases prior to 0425 UTC.
0434 UTC, Notification #1
First notification above 1x108 J threshold. Substantial activity in the “MOD” or “Moderate” level, but still below “HIGH’ energy level of 1x109 J. Notification issued at 2/6 0434 UTC (2/5 18:34 HST, the local time of the sent email). This notification time is only 4 minutes from the last recorded data segment, and ~6 mins from the actual time of recorded volcanic activity. Email notification below:
Date: Tue, 5 Feb 2008 18:34:23 -1000 (HST)
From:
To:
To:
To:
To:
To:
Subject: Volcano Infrasound Activity Update for 2008/2/6 03:30-04:30 UTC
DRAFT Infrasound Energy NOTIFICATION, ECUADOR ASHE PROJECT
TUNGURAHUA VOLCANO
DETECTIONS AT STATION RIOE, RANGE OF 37 KM, TIME IN UTC
MOD Energy Level, 1.28e+008 J
Time interval: 03:30-04:30 UTC
0439 UTC, Notification #2
Energy level doubles at next 5 minutes calculation, triggering another email alert. Note the energy level is still in the “moderate” energy range (4.41x108 J, 15 min power level= 1x106 W). Pertinent email text:
MOD Energy Level, 4.41e+008 J
Time interval: 03:34-04:35 UTC
0440-0500 UTC:
Constant jetting that decreases over time. Signal concentrated between ~0.1-2 Hz. First plume observed in satellite imagery, with plume detaching from the summit.
0500-0540 UTC:
Some jetting but at lower power levels. New ash plume rising.
0540-0655 UTC:
Jetting increases substantially and becomes more broadband (up to ~9x106W). Signal is constant, with notch at ~0.5 Hz. Large explosion at 0626 UTC. Plume continues to increase in size and height.
0700-0830 UTC
Another jetting episode, but spectral structure somewhat different than previous episode. Power levels peak at ~6x106 W. Signal gradually decreases over time. Harmonics present in higher frequencies around 0800-0830. Ash plume is starting to separate, and does so by 0815 UTC.
0843 UTC
Significant explosion, with amplitude of ~13 Pa and energy ratio of 96. This triggers and email notification.
Time=09:11:14, Raw Amp=7.99 Pa, Dur=7.0 sec, Energy Ratio=35.64
0911 UTC
Another large explosion with an amplitude of ~8 Pa and energy ratio of 35.64. Meteorological clouds make it difficult to discern ash cloud…if any.
0945-1115 UTC
Most energetic jetting of entire sequence, with power levels up to 1x107 W. Infrasound signal is broadband and fairly constant. Shuts off rapidly at 1115 UTC. New plume observed at 0945 and grows in width and length until 1145 image. This plume appears to stay attached to the vent during the rest of the sequence.
0959 UTC, Notification #2
Over the past 5 hours the hourly energy level has gone up and down, but stayed above 4.41x108 J (the energy level of the past notification). The “HIGH” threshold has already been attained, but no notification was issued. In hindsight this was a flaw in the system design and could have been useful. At 2/6 0955 UTC the energy level doubles from its previous value, triggering another update. Also, it would have been useful to include a phrase related to ‘energy level has double since past update’. Note we are now in the “HIGH” energy level and have been for awhile, but no alert was triggered since we already had triggered and the energy level had not doubled since the past update. As mentioned before, a new plume is observed in satellite imagery and significant eruption is noted by VAAC. Our satellite estimates put the max plume height at ~9.6 km. Max power level is ~1.1x107 W, highest of this eruption. Pertinent notification email text:
HIGH Energy Level, 4.69e+009 J
Time interval: 08:55-09:55 UTC
1120-1205 UTC
Approximately three shorter, less energetic (but still significant) jetting episodes. Power levels between 2-6x106 W.
1207-1345 UTC
Five-six jetting episodes that all begin with a large explosion. Duration and frequency content change somewhat, but fairly similar to each other in spectral shape and duration. Power levels vary between 2-4 x 106 W. Satellite image is cropped and difficult to discern maximum extent of plume. The VAAC reports a detached plume at 14.3 km moving south (Figure 3), with a smaller plume attached to the summit. We are in the process of obtaining new satellite data to confirm the estimated VAAC heigh, as the infrasound data would suggest a higher ash plume from the 1000-1100 UTC infrasonic power levels.
Figure 3. VAAC ash advisory image for the 2/6 Tungurahua eruption.
1345-1430 UTC
Final jetting episode of sequence, power level up to ~8 x 106 W. Signal drops off rapidly.
1529 UTC, Notification #4
After staying at the HIGH level, the volcano finally starts to quiet down and the energy level drops off rapidly. There is still some activity around 1400-1420 UTC, and this is included in the past hour of the energy calculation. The VAAC cites decreased seismic and infrasound activity in their advisory at 1633 UTC. Pertinent email text:
Activity level decreased
LOW Energy Level, 8.33e+006 J
Time interval: 14:25-15:25 UTC
Pertinent explosion update email text issued during the 2/6 eruption:
Time=06:26:17, Raw Amp=3.85 Pa, Dur=2.0 sec, Energy Ratio=13.83