Growth of the Soufrire Hills Dome: Fusion of Thermal Infrared Spaceborne Data with a Multiparameter

1
Growth of the Soufriére Hills Dome Fusion of
Thermal Infrared Spaceborne Data with a
Multi-parameter Database
Sally S. Kuhn and Michael S. Ramsey
Image Visualization and Infrared Spectroscopy
(IVIS) Laboratory (http//ivis.eps.pitt.edu)
Department of Geology and Planetary Science
University of Pittsburgh, Pittsburgh, PA 15260,
USA
Abstract
Emplacement processes of the Soufriére Hills dome
(Montserrat) can be discerned using thermal
infrared (TIR) wavelengths, which are sensitive
to changes in temperature flux and emissivity
variations over time. With the Advanced
Spaceborne Thermal Emission and Reflection
Radiometer (ASTER), one cloud-free image is
captured every three months on average, with
increased frequency in 2002, when the volcano was
a high priority target. Montserrat Volcano
Observatory weekly reports from 1999 to present
(available online) were also ingested into a
multi-parameter, searchable database. These data,
which detail specific volcanic activity, were
compared against the ASTER data. The database
fields include SO2 flux, GOES-derived radiance
measurements, description of dome growth and
collapse, and intensities of pyroclastic flows,
rockfalls, fumarolic activity, and seismic
activity. This database provides a unique
cross-reference for the interpretation of the
spaceborne data, as well as highlighting
observable trends in each of the volcanic
activity types. We hope to apply this methodology
in the prediction and monitoring efforts of
active dome hazards elsewhere.
FIGURE 1. Map of the Caribbean. Detail
topographical hazard map of Montserrat shows
Englishs Crater outlined in yellow. The box
represents the same area of each ASTER image
below.
FIGURE 7. Graphic representation of numerical
data. Weekly totals of rockfall, hybrid events,
long-period earthquakes, volcano-tectonic events,
sulfur dioxide, and long rockfall signals were
plotted from December 31st, 1999 to April 11th,
2003. Sulfur dioxide output is reported in
ranges, with the maximum and minimum numbers
reported together. Green lines on all graphs
represent explosive events (orange on sulfur
dioxide graph), blue lines represent ASTER
capture.
Results and Conclusions
TABLE 1. Linear spectral deconvolution results
showing aerial percentage of each endmember and
corresponding emissivity spectra for the hottest
and background temperature pixels.
Linear spectral deconvolution results are
reported in Table 1.The total percentage of
obsidian and blackbody should equal 100, however,
for most pixels containing an anomaly, this is
not the case. Anomalous pixels for the April
13th, 2002 image are favorable, along with pixels
at or near background temperature. Petrographic
analysis demonstrates a 5-15 glass content for
samples thought to have resided in the dome for a
significant amount of time, and 25-30 for
samples collected after explosive eruptions
Sparks et al, 2000.The week including the April
13th, 2002 capture shows a higher number of
rockfalls and mild dome growth, but the imagery
suggests relatively uniform heat distribution
over each pixel and a 27-64 glass content.
Possibly a third endmember, i.e. andesite, should
be considered. Another hypothesis to explain the
results (Figure 5) suggests a pixel with a
smaller percentage of elevated temperature
results in a radiance curve derived from two
blackbody curves, one for the cooler temperature
and one for the hot temperature Dozier, 1981.
Further investigation using sub-pixel analysis is
required in order to account for anomalous
emissivity spectra. The temperatures from the
ASTER instrument versus GOES are somewhat
inconclusive. ASTER temperatures differ from 1.6
ºC to 13.7 ºC and do not plot consistently higher
or lower than the GOES temperatures. To further
investigate temperature inconsistencies between
ASTER and GOES, more work should be performed in
other areas, such as Bezymianny and Unzen.
Statistical results show that some variables cor
relate on a weekly temporal scale, while others
do not. Results for correlation 1 based on
regression and ANOVA tests reveal a best line of
fit as y353.986 0.164x, a p-value of 0.002,
and an f-value of 10.375. No correlation was
found with 4, except when hybrid activity was
separated into two groups, below 25 and 26 and
above. Fishers exact test shows a p-value of
0.008 between explosive behavior and a high
number of hybrid events. Correlation 6 shows
that the average number of rockfalls is fewer
with high values of hybrid events with a t-value
of -3.2 and a two-tailed p-value of 0.002. All
other variables show no correlations.
FIGURE 4. ASTER system response functions
(far left) in thermal wavelengths. Linear
spectral deconvolution endmembers (left), showing
blackbody in blue and obsidian in red.
FIGURE 2. (Above) Subsets of the six ASTER images
showing dome anomalies in the thermal infrared
wavelength regions. The first three images (2000
and 2001) were captured at approximately 1040 PM
local time, while the 2002 images were captured
at approximately 1040 AM local time. Note the
saturated pixels within the dome anomalies.
Scale 5.4 km wide.
FIGURE 3. (Left) Subset of GOES image (left) and
a resampled ASTER image (right) for comparison.
Red brackets indicate top and bottom of island.
GOES image spans 296 km across. ASTER is 60 km.
FIGURE 5. Illustration showing separated radiance
curves for an example pixel with 10 at 300 C (
Thot) and 90 at 25 C (Tcold).

Introduction
Soufriére Hills (FIGURE 1), an active andesitic
dome, has undergone three different eruptive
phases since reactivation in 1992. The first
began in 1995 and continued through 1998, with
cycles of growth and collapse. From early 1998 to
late 1999, the duration of the second phase, is
characterized by no extrusion, but dome collapse
and small explosions. The third phase, from late
1999 to present, shows renewed extrusion, with
two major collapses Montserrat Volcano
Observatory, 2002. An understanding of extrusion
behavior, particularly an adequate degassing of
volatiles, can provide information on the
triggers of explosivity. Presuming that thermal
emission spectra combine linearly in thermal
infrared wavelengths, aerial percentages of two
end-members (obsidian and vesicles) can be
estimated using a linear spectral deconvolution
technique Ramsey and Fink, 1999. Variations in
surface texture between the two end-members
specifically provide insight of emplacement time
or extrusion rate, volatile content, and internal
structure of the dome. ASTER is the first high
resolution, multi-spectral (FIGURE 4) spaceborne
instrument capable of discerning these features.
ASTER does not, however, provide any information
about the eruptive phase of the volcano. The GOES
satellite, imaging approx. every 15 minutes, does
provide information on the eruptive phase Harris
et al, 2001 if tracked long term (at least a
week). Further, a multi-parameter database offers
a complementary set of data unobtainable from
spaceborne satellites. The fusion of these two
datasets with ASTER imagery provide a foundation
for analysis of surface texture.
Data Sets and Methods

• Six Level 2 05 (emissivity) and 06 (kinetic
  temperature) ASTER image products were chosen
  based on the presence of an anomaly and relative
  absence of clouds (FIGURE 2)
• For each image, the emissivity spectra were
  unmixed against the endmembers of obsidian and
  blackbody for the four hottest pixels, and
  plotted in FIGURE 4.
• Level 2 (atmospherically corrected) and Level 1 B
  (non-atmospherically corrected) images were then
  resampled to 4x4 km pixel sizes to mimic GOES
  resolution (FIGURE 3). Temperatures of the pixel
  that includes the dome were plotted in FIGURE 6
  for all three data sets. GOES band 5 (11.5-12.5
  µm) temperatures span one hour prior and hour of
  ASTER image capture.
• The online Montserrat Volcano Observatory (MVO)
  weekly reports from December of 1999 to April of
  2002 were converted into a database to serve as a
  look-up table for ASTER imagery.
• Numerical data were graphed with a smoothing
  curve, explosive events, and ASTER capture to
  illustrate increasing or decreasing trends in
  activity (FIGURE 7).
• Statistical tests were conducted on numerical and
  categorical data based upon previous correlations
  determined by other researchers on smaller
  temporal scales. The following positive
  correlations were chosen for statistical
  analyses
• 1. Gas venting and rockfall Luckett et al,
  2002
  
• 2. Rockfall and long-period events Cole et
  al, 1998
  
• 3. Hybrid events should precede dome collapse
  Neuberg et al, 1998
  
• 4. Hybrid events should precede major
  explosions Neuberg et al, 1998
  
• 5. Long-period events should precede major
  explosions Miller et al, 1998
  
• 6. Hybrid events should precede rockfall
  White et al, 1998
  
• 7. Hybrid events linked to violent degassing
  White et al, 1998
  
• 8. Long-period events should precede and
  follow large hybrid events White et al, 1998
  
• 9. Collapses preceded by long-period and
  hybrid events Neuberg et al, 2000
  
• 10. Volcano-tectonic events should be low
  during dome growth Miller et al, 1998

References
Cole et al., Pyroclastic flows generated by
gravitational instability of the 1996-97 lava
dome of Soufriére Hills Volcano, Montserrat.
Geophys. Res. Lett., 25, pp. 3425-3428, 1998
Dozier, J., A method for satellite identificatio
n of surface temperature fields of sub-pixel
resolution. Remote Sens Environ, 11, pp. 221-229,
1981 Druitt, T. H., and Kokelaar, B. P. (eds).
The Eruption of Soufriére Hills Volcano, from
1995 to 1999. Geological Society, London,
Memoirs, 21, 595-602, 2002 Harris, A.J.L., Pilg
er, E., Flynn, L.P., Web-Based Hot Spot
Monitoring using GOES What it is and How it
Works. Advances in Environmental Monitoring and
Modeling, 1, pp. 3-31, 2001 Miller et al., Seis
micity associated with dome growth and collapse
at the Soufriére Hills Volcano, Montserrat.
Geophys. Res. Lett., 25, pp. 3401-3404, 1998
Montserrat Volcano Observatory, http//www.mvo.
ms/ Neuberg et al., Results from the broadband
seismic network on Montserrat. Geophys. Res.
Lett., 25, pp. 3661-3664, 1998
Neuberg et al., Models of Tremor and low-frequen
cy earthquake swarms on Montserrat. J. Volcanol.
Geotherm. Res., 101, pp. 83-104, 2000
Ramsey, MS and Fink, JH. Estimating silicic lava
vesicularity with thermal remote sensing a new
technique for volcanic mapping and monitoring.
Bull Volcanol., 6132-39, 1999
Sparks et al., Control on the emplacement of the
andesite lava dome of the Soufriére Hills
volcano, Montserrat by degassing-induced
crystallization. Terra Nova, 12, p. 14, 2000
White et al., Observations of hybrid seismic eve
nts at Soufriére Hills Volcano, Montserrat July
1995 to September 1996. Geophys. Res. Lett., 25,
pp. 3657-3660, 1998
FIGURE 6. Comparison of GOES band 5
(non-atmospherically corrected), ASTER L2
(atmospherically corrected), and ASTER L1B
(non-atmospherically corrected) temperatures of
the pixel that includes the dome. GOES data was
not available for December 28th, 2000.