Sustainable Surface Deformation Related with 2006 Augustine Volcano Eruption in Alaska Measured Using GPS and InSAR Techniques

Abstract

Augustine volcano, located along the Aleutian Arc, is one of the most active volcanoes in Alaska and nearby islands, with seven eruptions occurring between 1812 and 2006. This study monitored the surface displacement before and after the most recent 2006 eruption. For analysis, we conducted a time-series analysis on data observed at the permanent GPS(Global Positioning System) observation stations in Augustine Island between 2005 and 2011. According to the surface displacement analysis results based on GPS data, the movement of the surface inflation at the average speed of 2.3 cm/year three months prior to the eruption has been clearly observed, with the post-eruption surface deflation at the speed of 1.6 cm/year. To compare surface displacements measurement by GPS observation, ENVISAT(Environmental satellite) radar satellite data were collected between 2003 and 2010 and processed the SBAS(Small Baseline Subset) method, one of the time-series analysis techniques using multiple InSAR(Interferometric Synthetic Aperture Radar) data sets. This result represents 0.97 correlation value between GPS and InSAR time-series surface displacements. This research has been completed precise surface deformation using GPS and time-series InSAR methods for a detection of precursor symptom on Augustine volcano.

1. Introduction

Of the life-threatening natural disasters such as earthquakes, tsunamis, and volcanic activity, those created by orogenic movements are harder to predict and cause more property damage and human casualties. The natural disasters originating from volcanic activity can devastate cities, farmlands, and vegetation with flowing lava and pyroclastic flow, with follow-on effects like earthquakes, landslides, and tsunamis. In addition, the spread of volcanic ash may disrupt air travel through a wide area for an extensive period while causing respiratory problems for residents near the eruption site (Waythomas and Waitt, 1998). Although it is important to minimize the damage of a volcanic eruption by broadcasting warnings as early as possible, it is not easy to do so with current technology (Pieri and Abrams, 2005). The precursor signs of volcanic activity that have been known to us include increases in volcanic earthquake frequency and changes in volcanic gas emissions, crust shape, geothermal heat level, and hydrological conditions (Xu et al., 2012). Of the precursor signs, ground deformation caused by magma chamber movements observed hours or days before the eruption is one of the most common eruptive activities. It is possible to estimate the time of eruption based on change of the shape and form of ground deformation. In general, shallow magma chambers close to the surface, commonly around several km under the surface, cause bloating of the surface due to the mounting pressure within the chamber (Yazdanparast and Vosooghi, 2013).

To monitor and predict volcanic activities, a variety of sensors, including seismometer, clinometer, gravimeter, CGPS(Continuous Global Positioning System), and satellite image have been used. In particular, high-tech geodetic survey tools such as the GPS and InSAR have been used since the 1980s and have contributed to volcanic monitoring and research in meaningful ways (Lu, 2005). The recently developed InSAR technique, which uses two or more SAR(Synthetic Aperture Radar) images to generate maps of surface deformation or digital elevation, using differences in the phase of the waves returning to the satellite, can monitor displacements over several tens of square kilometers with several millimeter- or centimeter-precision, which is ideal for observing earthquake, volcano, glacier, and landslide movements (Jo et al., 2015; Lee et al., 2013; Lee, 2014). The remote-sensing technology like InSAR is widely used in volcano research because of its ability to make high-spatial-resolution displacement observations in mountains and other difficult to access areas (Jónsson et al., 2002; Pritchard and Simons, 2004).

Although GPS was originally developed for military purposes, it is now used more widely for civilian applications including land surveying, navigation, communication, and weather monitoring. With GPS devices, displacements of the earth’s crust and faults can be measured with millimeter-precision, which is useful for geophysics research. However, GPS devices that must be installed on the earth’s surface for observing surface displacements only from the station where the antenna is set up. Alternatively, InSAR data could provide the short spatial scale surface deformation such as every coherence points within SAR images practically lacking in GPS data (Wei et al., 2010). In particular, the use of GPS devices in volcanoes is limited since the antenna is easily damaged by eruptions and weather conditions may affect data collection (Cervelli et al., 2006; Lisowski et al., 2008). When comparing it to the InSAR technique’s relatively longer data collection cycle, the GPS method is superior in that it can measure surface displacements at the time of eruption and collect three-dimensional locational information unlike InSAR that provides the LOS(Line-Of-Sight) directions of the displacement only (Larson et al., 2010). As each observational technique has pros and cons, it is advisable to use multiple techniques in a complementary way to ensure a more precise outcome. As the two techniques of InSAR and GPS can offset problems of one approach by the other, their simultaneous use improves the precision of surface displacement monitoring (Lee et al., 2010).

In this study, ENVISAT data and GPS data, collected from June 2003 to August 2010, were simultaneously used to measure the movements before and after the eruption of Augustine volcano in Alaska, which had a record of discharge in 2006 (Berardino et al., 2002). This research alert us to important information of symptom for volcanic eruption through quantitative analysis from time-series surface deformation using GPS and InSAR techniques.

 

2. Study Area and Dataset

2.1 Study area

Augustine Island, the main locus of this study, is located in the southwestern Cook Inlet, about 290 km southwest of Anchorage. The island has a land area of 90 km2, with the highest peak reaching 1,260 m (Fig. 1(a)) (Waythomas and Waitt, 1998). The Aleutian Islands, of which Augustine volcano is a part, belong to a subduction zone at which the Pacific and North American plates meet, with as many as 57 volcanoes nearby (Michel et al., 2009). Of these, Augustine volcano is one of the most active (Beget and Kowalik, 2006; Miller et al., 1998). Since the first recorded eruption in 1812, the volcano has flared up in 1883, 1935, 1964-65, 1976, and 1986. After a 20-year hiatus, it again exploded in January 2006 (Power et al., 2006).

Fig. 1.Phenomena associated with 2006 eruption on Augustine volcano (a) GPS permanent stations were installed on Augustine Island. And AC59, used for reference station, are located near Augustine (b) Seismic hypocenters and surface temperature. Blue circles represent earthquake epicenters corresponding to seismic magnitude (Power et al., 2010)

Of the seven eruptions, the most violent 1883 episode caused a tsunami of 7.5-9.0 m as debris avalanche on the northern slope of the island poured onto the sea and caused repercussions as far as the Kenai Peninsula in English Bay 80 km away (Miller et al., 1998; Waythomas and Waitt, 1998). As for the 1976 eruption, which was closest in pattern to the 2006 event, there had been precursor signs nine months before the eruption with a notable increase in earthquake activity, including 13 volcanic eruptions for three days since January 22. Following a 12-day lull, volcanic activity increased slightly between February 6 and 15, which created lithic-rich pyroclastic flows and lava domes in the northern slope. From April 13 to 18, there were final eruptions that accompanied lava dome expansion and block-and-ash flows (Coombs et al., 2010; Larsen et al., 2010). The 1986 episode showed similar patterns to the 2006 event and the one in 1976, in that large pyroclastic flows gushed out in the beginning of the eruption and moved through the northern slope reaching a point 5 km from the northern coast (Miller et al., 1998).

The precursor signs for the eruption in 2006 first appeared in May 2005, with a steady increase in small tremors in and around the crater (Cervelli et al., 2006). Fig. 1(b) represents earthquake epicenters with blue circles and west-east (A-A’) and north-south (B-B’) cross section showing earthquake hypocenters associated with 2006 eruption at Augustine volcano, respectively. Red and blue cross represents depth of hypocenters before and after the 2006 eruption. Aster satellite image displays surface temperature after the 2006 eruption on pyroclastic flow deposit area at northern part of Augustine volcano. Aster satellite data need to be converted to radiance from DNs(Digital Numbers) for thermal bands. Also, Brightness temperature was converted using spectral radiance (Ghulam, 2009).

Moreover, Aster satellite image can display surface temperature after the 2006 eruption on pyroclastic flow deposit area at northern part of this island (Fig. 1(b)). After a brief lull, the seismic rate began rising rapidly from November 17, 2005, with the highest incidence and most severe tremors occurring in early January 2006, when the major eruption occurred (Figs. 2(a) and (b)) based on Alaska volcano observatory results.

Fig. 2.(a) Histogram showing the number of earthquakes per day from January 1, 2005 to December 31, 2011 (b) Magnitude of earthquakes from January 1, 2005 to December 31, 2011 (Power et al., 2010)

On December 2, the seismometer began detecting signs of phreatic explosions, followed by the discovery of a volcanic ash dusting mixed with weathered and glassy particles from December 12 during the active explosions of steam and gas at the peak of the volcano (Mattia et al., 2008). From January 11 in the following year, the explosion began in earnest, with 14 eruptions until January 28 (Michelle et al., 2010). The surface temperature data shown in Fig. 1(b) are collected from images taken from the Aster satellite about ten days after the major eruption. As shown in Fig. 1(b), the surface temperature in the northern part of the volcano was measured at a relatively high 80 ℃. This observation matches the preceding studies that the 14 explosions during the 2006 event caused the lava to flow through the northern slope (Wessels et al., 2010). The figure on the upper-left side in Fig. 1(b) is a diagram depicting the depth of small tremors that occurred before and after the 2006 eruption, with red points indicating pre-eruption and blue points showing post-eruption. The figure indicates that the pre-eruption tremors occurred near the surface while post-eruption tremors occurred much deeper in the crater.

As the volcanic activity went into a lull from early February 2006, the frequency of tremors also as a noticeable decline. During the period between late 2007 and early 2008, there were times when the color-coded level of concern rose to “yellow” as the volcanic activity resumed briefly. But it soon returned to a stable state and is continuing to be so as of 2016.

2.2 GPS data

In this study, surface displacements were monitored and analyzed before and after the 2006 event. To that end, GPS data were collected between January 2005 and December 2011 from 13 permanent GPS observation points installed in Augustine volcano by the PBO(Plate Boundary Observatory). All data thus collected are 30-second-interval 24-hour observations in the RINEX(Receiver Independent Exchange Format) format provided by the FTP server of UNAVCO.

To all observation points, the AC59 point was added so that it could be used as a base point when comparing data since the point is located outside the island and therefore is assumed to be impervious station from the volcanic impact. When selecting a reference point, one stable observation point was chosen near the island that had been run for a long time, understanding that the greater its distance from the volcano, the higher the margin of error. The AC59 station that had been used since September 2004 was selected. It is located on the North American plate about 24.5 km away from Augustine Island (Fig. 1(a)).

Other GPS observation points near the volcano were installed and operated by the Plate Boundary Observatory for the purpose of volcanic eruption monitoring. Beginning with the installation of the AUGL (renamed AV21 from 2006) observation station in September 2000, the GPS stations were installed at 12 points until September 2006. The six stations (AV11, AV16, AV17, AV18, AV19, and AV20) out of the total were installed after the 2006 eruption, which renders it impossible to find precursor signs before the explosion in 2006. As for the GPS stations AV03, AV04, and AV05 adjacent to the crater, all of them were damaged or burned during the 2006 episode. They have stopped collecting data ever since except the AV04 point which has been repaired. Fig. 1(b) shows a diagram indicating the permanent GPS observation locations used for this study.

To reduce errors in GPS observation data, we used data provided in file forms by the CODE(Center for Orbit Determination in Europe), one of the analysis centers for the IGS(International GNSS Service), and the CDDIS(Crustal Dynamics Data Information System) for GPS satellite precise ephemeris, satellite clock data, the earth’s rotational axis movement, and the earth’s ionosphere model. For the ocean tidal model, the FES2004 file was provided by OSO(Onsala Space Observatory).

2.3 InSAR data

To measure the degree of surface displacement at Augustine volcano through InSAR, ENVISAT data were acquired, the radar satellite operated by the ESA(European Space Agency). The data were collected include 22 images in 2229 tracks dated from June 2003 to August 2010 (Table 1).

Table 1.ENVISAT data and baseline information

 

3. Methodology

3.1 GPS processing

In general, GPS data processing can be done either through the relative or absolute positioning method. The relative positioning method determines the coordinates of the other observation point with millimeter-precision based on an observation point whose precise coordinates are already well known. In contrast, the absolute positioning method determines the coordinates independently by using signals received in the observation point from GPS satellites that can measure coordinates with centimeter-precision. To secure surface displacement data with millimeter-precision, this study adopted the relative method using constraints. For relative measurement, the earlier-mentioned AC59 observation point was used, with the coordinates for the observation point based on IGS08.

For precision baseline interpretation, Bernese GPS Software V5.0 developed by the Astronomical Institute of the University of Bern, Switzerland, has been used. Bernese is a software program widely used by scientists across the world for GPS network adjustment calculation, the earth’s ionosphere and troposphere modeling, and satellite orbit estimation. It is capable of determining precise integer ambiguity resolution even for extremely long baselines as far as 2,000 km (Dach et al., 2007).

Data processing using Bernese was performed in the order specified in Fig. 3 and includes the following processes. Firstly, formation of an optimal baseline through the OBS-MAX method that selects a set of baselines including maximum common observation data from all possible combinations. Secondly, estimation of relative tropospheric errors by fixing a point among observations to Niell model values to keep the stability of estimated values from deteriorating abruptly due to high correlations arising from the relatively short baseline distance around 10 km, when estimating tropospheric errors. Thirdly, determination of integer ambiguity numbers by way of the Quasi Ionosphere-free method that makes use of the L1 and L2 carrier signals.

Fig. 3.Diagram of data processing using Bernese software from GPS data

But the observation points AV04 (altitude of 915.95 m) and AV05 (altitude of 1,036.61 m) installed near the peak of the volcano experienced difficulties due to their antennas and mounts becoming covered with snow and ice (Cervelli et al., 2006). The snow and ice covering the devices in winter lead to delays in GPS signal transmission, which resulted in errors as large as 15 cm vertically from the observation point AV04. To precisely monitor surface displacement data, which is the main purpose of this study, therefore, we avoided using abrupt coordinate changes in the winter season by separating data with wide margins of error. In this case, the separation process was applied to data after the eruption with an assumption that the data observed just before the eruption would not be affected by the error as there would be no problem with snow and ice after the temperature rose. All data values with large margins of error have been discarded from the analysis.

3.2 InSAR processing

To avoid the coherence of interferogram due to snow and ice, SAR data were collected between mid-June and mid-October. We calculated the perpendicular baseline distance of all images as a way to create interferograms with a high coherence and ultimately created 25 interferograms, with the vertical baseline less than 400 m. All interferograms were created with the complex multi-look operation, with two looks in range and ten looks in azimuth directions. As a result, each pixel has an area of around 40 m by 40 m. The SRTM(Shuttle Radar Topography Mission) DEM(Digital Elevation Model) was used with about 30m resolution per second to remove the topographic phase contribution in the original interferogram.

Of the differential SAR interferometry images created from ENVISAT, only 16 interferograms selected as those with high positional errors because of a low coherence were excluded. As an initial reference point for time-series analysis, a point near the AV02 observation point was selected. Finally, the 16 interferograms were selected for more effective investigation of surface displacement patterns on the area were applied with the SBAS time-series surface displacement observation algorithm. With this, compensation for errors such as the positional errors due to the atmosphere, DEM errors and errors due to satellite orbits was possible. In general, in cases where abrupt surface displacements occur such as with volcanic eruptions, it is difficult to create differential interferometry images because the coherence of interferograms before and after the surface displacement is lowered. Even when interferograms can be created, it is easy for the irregularly occurring displacements in wide-ranging areas like a volcanic eruption to be excluded from analysis. For this reason, this study conducted a time-series analysis by dividing it into pre-eruption and post-eruption, with separate displacement distribution diagrams for the entire observation period including both pre- and post-eruption.

 

4. Results

4.1 GPS data analysis

For the 2,556 days from January 1, 2005 to December 31, 2011, data have been estimated by baseline interpretation, and a three-dimensional time-series data set has been created with the 12 noon observation time for each day at 12 permanent observation points in and around Augustine volcano as the observation standard. The daily three-dimensional geocentric coordinates thus created from the baseline interpretation were then converted into local plane coordinates, which are indicated in Fig. 4 with north and east and up divided into two parts. AV01 in Fig. 4 was the closest observation point among those that collected data before and after the eruption, which shows surface displacements relatively clearly.

Fig. 4.Time series of Augustine’s permanent stations. Light gray dots are outlier and the pink box is shown eruption period. Black line means trend of volcano behavior after eruption

As can be understood from the figure, the change in coordinates is seen visibly before the main eruption, which seems to be precursor signs. In particular, four months before the eruption a phenomenon of surface elevation of about 3 cm with an accompanying coordinate change of 3 cm to the south was observed. When Augustine volcano began erupting on January 11, 2006, the time-series data at the AV01 observation point showed an abrupt subsidence. Beginning in the second half of 2006, however, subsidence at the rate of 0.14 cm/year was observed without a clear horizontal coordinate change, which implies that the volcanic activity entered a stable phase.

In the case of the AV02 observation point, the surface displacement was almost identical to that of AV01, with a smaller amount of displacement as it was farther from the crater. For AV04, the bloating of the areas around the crater immediately before the eruption created a significant change in surface displacement to the southwest, the opposite direction from the crater. To examine this change in more detail, the coordinate change values for each direction (north, east, and up) were amplified at the moment of the eruption, which is indicated below the figure of AV01, AV02, and AV04. From AV05, which is closest from the crater, the displacement rate from the northeast was 15 cm. The amount of bloating was 5 cm, which was 5-10 cm higher horizontally and 5 cm lower vertically when compared to the AV04 observation point.

This implies that the surface displacement amounts between east and west differ, with the horizontal element higher in the east and the vertical element superior in the west. In contrast, the observation value at AV03 located in the north has a higher displacement value to the north horizontally. Comparison of the observation values for the observation points from AV11 to AV20 that were installed in mid-2006 after the eruption shows that the displacement values observed in AV18 and AV19 had values in opposite directions before the eruption. This may be because of the contraction of magma chambers as well as the compaction of the pyroclastic flow deposits.

Fig. 5 shows the time-series displacement values collected at AV04 and AV05, the observation points located closest to the crater of Augustine volcano. The two observation points were damaged on January 17 and 13, 2006, respectively, when the volcano erupted, which rendered it impossible to measure displacements. As they are located very close to the crater, however, it is possible to detect precursor signs more clearly than others. From the time-series graph, we can see that the gray line changes markedly from November 11, 2005. A day before the observation day, the time-series coordinates that exhibited a very slow linear pattern abruptly changed into very dynamic coordinates. The time series at AV05 exhibited a slight offset on November 17, 2005, which can be interpreted as a physical phenomenon when the crust broke off near the observation point ahead of the eruption (Fig. 5). Indeed, the earthquake activity histogram in Fig. 2 shows that the earthquake frequency increased significantly from mid-November. After this, the AV04 observation point showed a horizontal coordinate change to the southwest with an accompanying surface lifting of up to 15 cm while AV05 showed a horizontal coordinate change to the northeast with 10 cm of surface lifting (Fig. 5).

Fig. 5.Time series analysis of surface deformation at AV04 and AV05 stations

Fig. 6 shows the change in baseline length as time passed in the shape of a triangle (AV01-AV02, AV02-AV-03, and AV03-AV01) across north and south as well as to the west with the volcano’s crater in the center. Beginning in September 2005, the change in baseline length is clearly visible. When this is compared with the cumulative graph (Fig. 2(a)) for earthquakes that occurred in Augustine volcano since 2005 especially end of this year, the time period for the increase in earthquake activity and the change in baseline length in the two observation points corresponds quite closely. The increase in baseline length may be due to the expansion of the volcano as magma rose up just below the surface. The change in baseline length that cut across the crater can be used as an indicator of the degree of volcanic activity. As a result, we can see that each baseline changes in a similar pattern. The whole research period can be divided into five sub-periods, with separate analysis for each period.

Fig. 6.Time series of distance change between stations AV01-AV02, AV02-AV03, and AV03-AV01. Trend of distance are changed after August 2005. And the changing points agree with inflation point of earthquakes

The end results are listed in Table 2. Stage 1 is a time period in which the baseline length has not been changed by volcanic activity. A time-series analysis on the same period also reveals that there were few signs of volcanic activity that preceded the 2006 eruption. At Stage 2, in which steady lengthening of the baseline occurred for five months, surface lifting accompanied by a radial-shaped expansion horizontally with the crater at the center is noted. The speed of lifting is estimated at 9.7 cm/year, with that for AV04 and AV05 that are close to the crater showing rates of 8 and 11 cm, respectively. The reason the displacement rates are higher in observation points closer to the crater may have to do with the fact that the observation points are more subject to magma sources. For the relatively short subsequent two months at Stage 3, a full-fledged eruption occurred. Unlike the preceding stage, the third stage shows polar opposite results both vertically and horizontally, which includes an abrupt shortening of the baseline length and a rapid sink rate of 9.2 cm/year, as well as horizontal displacements around the crater (Table 2).

Table 2.Step of surface deformation of the 2006 Augustine volcano eruption

The surface displacement that includes the subsidence and horizontal movement was caused by the contraction of the volcano after the eruption. Stage 4 shows a pattern of increasing baseline length, with slight surface lifting and radial-shaped displacement around the crater. Compared to the second and third stages, the speed of surface displacement slowed significantly. Finally, at Stage 5 between October 16, 2006 and the end of 2011, a very slow sink rate was witnessed, with a few instances of large subsidence observed at some points. In particular, points AV04, AV18, and AV19 exhibited sink rates of 2.1, 4.0, and 3.3 cm, respectively (Fig. 4). According to a study by Wessels et al. (2010) that made use of thermal infrared ray aerial photos and Aster satellite thermal band images on the 2006 eruption, the northern slope where the AV18 observation point was located was the locale for most pyroclastic flows, just like in 1986 (Miller et al., 1998). According to “Deposits from the 2006 eruption of Augustine volcano” produced by Coombs and Michelle (Larsen et al., 2010), AV19 was also located in an area with sediments of pyroclastic flows.

Given these facts, the subsidence of the area is likely to be due to the compacting of the sedimentary pyroclastic flows rather than the outcome of magma-caused displacements. Therefore, the surface displacement patterns observed in Stage 5 do not seem to suggest additional eruptions after the 2006 episode.

4.2 InSAR data analysis

Fig. 7 shows the average speed of surface displacements occurring between 2003 and 2006 just prior to the eruption (Fig. 7(a)) and that between 2006 right after the eruption and 2010 (Fig. 7(b)). According to the measurement by SBAS method prior to the eruption in 2006, the northern slope exhibited a fast sink rate of 3.5 cm/year while all other slopes showed a slow uplift rate of 0.5-1.0 cm/year. The area with a high sink rate on the northern slope was due to the hardening of pyroclastic flows just like the surface subsidence observed in AV03 and AV18 GPS observation points (Fig. 1(a)) differently with a radial subsidence pattern from crater according to magma chamber’s shrink. The northern slope showed subsidence in the area with sediments of pyroclastic flows during the eruption in 1986, with continuous subsidence observed between 1992 and 2005 for a total of 40 cm (Lee et al., 2010). The uplifting observed in all other slopes may be interpreted as a sign of an impending eruption.

Fig. 7.Surface deformation averaging maps before (a) and after 2006 (b) eruption

InSAR measures the degree of displacement of LOS directions from the satellite to the surface from 2003 to 2010 in this study without dividing it into pre-eruption and post-eruption for measuring time-series surface deformation (Fig. 8). This time-series processing is appropriate to linear surface deformation measurements at high coherence areas.

Fig. 8.Time series surface deformation maps from June 18 2003 to August 25 2010 associated with 2006 eruption

4.3 Comparison of GPS and InSAR

In contrast, GPS observes three-dimensional displacements from the antenna location from 2005 to 2011 (Fig. 4), which makes it impossible to compare the observation values of SAR and GPS. To compare the displacement values of the two observation methods, it is necessary to recalculate the 3D (north, east, and up) displacements observed by GPS by projecting them onto the LOS directions of the satellite. Suppose that the unit vector of LOS direction as Vlos, each element of the unit vector can be expressed as in Eq. (1) below, where N0 , R0 and U0 are initial value of 3D displacements by projecting them onto the LOS directions of the satellite, θL and θT denote the satellite’s sight and track angles, respectively.

As the inner product of the unit vector is equivalent to the size of the projected vector, we can project the 3D surface displacement vector Vgps observed by GPS to the LOS direction using Eq. (2) as follows:

In addition, the deformation derived by interferometric techniques is relative to a reference area in the image, not an absolute value. Hence, an additional process that derived the value from GPS time-series analysis was added to InSAR results pre- eruption and post-eruption, respectively.

Fig. 9 compares the GPS-based surface displacement projected to the LOS direction with SBAS time-series result. Especially, we cannot match directly GPS and InSAR time-series results in case of station AV 04 in Fig. 9 because of low coherence interferograms by 2006 volcanic eruption. So we used time-series processing result from dividing it pre and post-eruption and reference point from GPS result at eruption time. From this, we can see that the degree of displacement through the two observation methods matches at the level of sub-centimeter scale for pre- and post-eruption with 0.97 correlation value (Fig. 10). It implies that the degree of displacement at Augustine volcano is within a reasonably reliable range.

Fig. 9.Time series analysis of crustal deformation detected by GPS and InSAR on the LOS component

Fig. 10.Correlation analysis between GPS and InSAR

 

5. Discussion

Fig. 4 shows a list of coordinate displacements from relative positioning through GPS data converted into local NEU (north, east, and up) coordinates. As can be seen from the time series, there were coordinate changes two months before the eruption in six observation points (AV01, AV02, AV03, AV04, AV05, and AUGL (renamed AV21 from 2006)) installed before the 2006 explosion. On November 11, 2005, there was a sudden offset in the time series of AV04 and AV05, with a visible increase in seismic activity. Since the offset, the speed at which coordinates changed increased rapidly at AV04 and AV05, with the detection of a weak change in the deformation rate in observation points far from the magma source. Since the eruption in 2006, the time series in each observation point shows a more stable state of displacement compared with the one before. At observation points AV18 and AV19, simultaneous horizontal movements and surface subsidence were recorded (Fig. 4).

To analyze surface displacement patterns more precisely at the time of eruption, changes in baseline lengths between observation points AV01, AV02, and AV03 were diagramed (Fig. 6). This figure shows clearly the change in baseline length before and after August 2005. When compared with the seismic histogram (Fig. 2(a)) that occurred at Augustine from April 2005 to March 2006, the time of increased seismic activity and changes in baseline length between the two points matches closely, with both seismic activity and baseline length recording their highest values in January 2006 when the main eruption occurred. It implies that the increase in baseline length is a precursor sign of a major eruption, which can be interpreted as the distance between the two points being moved farther apart due to the mounting pressure from magma that bloated the volcano.

The three-component time series in Fig. 11 shows changes in baseline distances (AV01-AV02, AV04-AV05, AV18-AV19, and AV01-AUGL (renamed AV21 from 2006)) among the observation points across the crater. From this, we can see that each baseline had similar patterns and the entire time series can be divided into the following five time periods: (1) 1/1/2005 – 8/14/2005; (2) 8/15/2005 – 1/10/2006; (3) 1/11/2006 – 3/16/2006; (4) 3/17/2006 – 10/15/2006; and (5) 10/16/2006 – 12/31/2011.

Fig. 11.Time series of distance change among the each station. Each pair of stations spanning the Augustine’s summit. And the section was separated based on distance change trends (Unit: cm)

Except for Stage 1, in which no clear baseline length change was detected, surface displacement vectors are calculated and diagramed for stages 2 to 5 (Fig. 12). The surface displacement vector for each observation point has been calculated by a straight fitting by means of least-squares adjustment by each interval. The map on the left side of Fig. 12 represents a horizontal displacement vector while the one on the right side represents a vertical displacement vector.

Fig. 12.Comparison surface deformation between each period. The left side figure means horizontal deformations and the right side means vertical deformations (a) 15.08.05-10.01.06 (b) 11.01.06-16.03.06 (c) 17.03.06-15.10.06 (d) 16.10.06-31.12.11

In Stage 2 (Figs. 12 (a) and (a’)), in which a pattern of gradually increasing baseline length has been shown, there was a radial-shaped horizontal pattern with the crater at the center, with an accompanying vertical surface uplifting. This surface deformation is believed to have been caused by the mounting pressure within the volcano as magma moved closer to the surface. It is possible here to interpret this surface change as a precursor sign of the eruption in January 2006. A surface lift at the average rate of 9.7 cm/year was detected, with an average of 3.9 cm of vertical surface displacement recorded during the same time. In particular, the observation points AV04 and AV05, which are close to the crater, experienced significant lifting (8 cm and 11 cm, respectively), as well as an average horizontal displacement of 9.5 cm. The difference may have been due to the fact that the two points adjacent to the crater were influenced more by the magma source. During the two months in Stage 3 (Figs. 12 (b) and (b’)), an abrupt shortening of baseline length and rapid sink rate of 9.2 cm/year occurred (Table 1). As there was an eruption in earnest during this stage, the observed subsidence may be largely due to the contraction of internal pressure within the volcano after the eruption. A negative vertical coordinate change as well as a horizontal surface displacement movement toward the crater was detected. During Stage 4 (Figs. 12 (c) and (c’)), a slight baseline length increase was detected after an abrupt baseline length shrinkage in Stage 3. Just like the second stage, there was a radial-shaped displacement toward the crater detected at this stage. Comparing to earlier stages 2 and 3, the uplift rate of 0.3 cm/year in Stage 4 is much slower than before (Table 1). Finally, in Stage 5 (Figs. 12 (d) and (d’)) from the time of the 2006 eruption to the present, a very slow sink rate has been noted, with relatively large episodes of surface subsidence reported at some observation points. Most instances of subsidence occurred near the crater, with AV18 and AV19 observed to have sunk by 4.0 cm and 3.3 cm, respectively. The AV18 observation point’s location is where most pyroclastic flows occurred in the 2006 event. The area was also the locus of pyroclastic flows in volcano eruptions before 2006 (Miller et al., 1998). According to “Deposits from the 2006 eruption of Augustine volcano” produced by Coombs and Michelle (Larsen et al., 2010), AV19 was also located near the area where there were sedimentary pyroclastic flows. Therefore, the subsidence was most likely caused by the hardening of the sedimentary pyroclastic flows rather than the outcome of magma-caused displacements. Therefore, the surface displacement patterns observed in Stage 5 do not seem to suggest additional eruptions after the 2006 episode.

 

6. Summary and Conclusions

Surface displacement patterns were analyzed before and after the 2006 eruption of Augustine volcano of Alaska as a way to determine precursor signs before an eruption, post-eruption earth crust changes, and prediction of a major eruption. To complement the limitations of one approach with another, GPS and InSAR techniques were used and observed surface displacements occurring between 2003 and 2011.

To that end, the permanent GPS data were collected between the years 2004 and 2011, as well as the ENVISAT SAR image data between 2003 and 2010 for time-series analysis of the observed data. The GPS observation data were then applied with the SBAS algorithm for a time-series analysis based on SAR images. By taking into account the features of SAR images that are not capable of creating differential interferometry images when large displacements such as volcanic eruptions occur suddenly, the data were analyzed by breaking down the period into pre and posteruption stages. The results are summarized as follows. First, we found that an average of 2.3 cm/year of surface uplift before the eruption were observed to be a precursor sign of the 2006 eruption. In particular, we confirmed an abrupt uplift and subsidence of the surface through GPS observation. By observing the change in baseline distance across the crater, we divided the observation period into five stages, with each stage showing a different surface displacement pattern. Second, the surface displacement patterns observed by GPS and SAR imaging have in general stabilized after the 2006 eruption. But in some areas a continuing trend of subsidence at the rate of 1.6 cm/year has been reported. In some areas with confirmed cases of subsidence, this may be due to hardening of the sedimentary pyroclastic flows from previous volcanic activities instead of the result of the current volcano eruption. We can draw a conclusion from this that the displacement patterns from 2006 to the present suggest that there are no precursor signs as of yet to suggest additional eruptions. Third, the CGPS displacements were projected to the LOS vector for comparing SBAS InSAR and GPS time series. As a result, the degree of surface displacement by means of the two observation methods matches the level of sub-centimeter scale. A correlation test indicated that the two methods exhibited a high degree of correlation at 0.97, which confirms the reliability of each observation method.

From this study, we found the comparison of GPS and satellite radar observation is adequate method to monitor volcanic activity for precursor signs and that permanent monitoring of surface displacement is also useful both in terms of cost and time. In particular, the combined use of the two techniques can compensate for the limitations of one method, allowing more precise displacement observations. We will conduct continuous surface monitoring for real-time surveillance that can signs of impending eruptions in the future.