Particles in the great Pinatubo volcanic cloud of June 1991: The role of ice

Abstract

1] Pinatubo's 15 June 1991 eruption was Earth's largest of the last 25 years, and it formed a substantial volcanic cloud. We present results of analysis of satellite-based infrared remote sensing using Advanced Very High Resolution Radiometer (AVHRR) and TIROS Operational Vertical Sounder/High Resolution Infrared Radiation Sounder/2 (TOVS/HIRS/2) sensors, during the first few days of atmospheric residence of the Pinatubo volcanic cloud, as it drifted from the Philippines toward Africa. An SO 2 -rich upper (25 km) portion drifted westward slightly faster than an ash-rich lower (22 km) part, though uncertainty exists due to difficulty in precisely locating the ash cloud. The Pinatubo clouds contained particles of ice, ash, and sulfate which could be sensed with infrared satellite data. Multispectral IR data from HIRS/2 were most useful for sensing the Pinatubo clouds because substantial amounts of both ice and ash were present. Ice and ash particles had peak masses of about 80 and 50 Mt, respectively, within the first day of atmospheric residence and declined very rapidly to values that were <10 Mt within 3 days. Ice and ash declined at a similar rate, and it seems likely that ice and ash formed mixed aggregates which enhanced fallout. Sulfate particles were detected in the volcanic cloud by IR satellites very soon after eruption, and their masses increased systematically at a rate consistent with their formation from SO 2 , which was slowly decreasing in mass during the same period. The initially detected sulfate mass was 4 Mt (equivalent to 3 Mt SO 2) and after 5 days was 12–16 Mt (equivalent to 9–12 Mt SO 2). Components: 12,820 words, 29 figures, 9 tables, 2 animations.

Full text

Particles in the great Pinatubo volcanic cloud of June 1991:
The role of ice
Song Guo
Department of Geological Engineering and Sciences, Michigan Tech nological University, Houghton, Michigan 49931, USA
Now at Department of Atmospheric and Oceanic Sciences, McGill University, 805 Sherbrooke Street West, Montreal,
Quebec, Canada H3A 2K6 (songguo@zephyr.meteo.mcgill.ca)
William I. Rose, Gregg J. S. Bluth, and I. Matthew Watson
Department of Geological Engineering and Sciences, Michigan Tech nological University, Houghton, Michigan 49931, USA
[1]Pinatubo’s 15 June 1991 eruption was Earth’s largest of the last 25 years, and it formed a substantial
volcanic cloud. We present results of analysis of satellite-based infrared remote sensing using Advanced
Very High Resolution Radiometer (AVHRR) and TIROS Operational Vertical Sounder/High Resolution
Infrared Radiation Sounder/2 (TOVS/HIRS/2) sensors, during the first few days of atmospheric residence
of the Pinatubo volcanic cloud, as it drifted from the Philippines toward Africa. An SO
2
-rich upper (25 km)
portion drifted westward slightly faster than an ash-rich lower (22 km) part, though uncertainty exists due
to difficulty in precisely locating the ash cloud. The Pinatubo clouds contained particles of ice, ash, and
sulfate which could be sensed with infrared satellite data. Multispectral IR data from HIRS/2 were most
useful for sensing the Pinatubo clouds because substantial amounts of both ice and ash were present. Ice
and ash particles had peak masses of about 80 and 50 Mt, respectively, within the first day of atmospheric
residence and declined very rapidly to values that were <10 Mt within 3 days. Ice and ash declined at a
similar rate, and it seems likely that ice and ash formed mixed aggregates which enhanced fallout. Sulfate
particles were detected in the volcanic cloud by IR satellites very soon after eruption, and their masses
increased systematically at a rate consistent with their formation from SO
2
, which was slowly decreasing in
mass during the same period. The initially detected sulfate mass was 4 Mt (equivalent to 3 Mt SO
2
) and
after 5 days was 12–16 Mt (equivalent to 9 –12 Mt SO
2
).
Components: 12,820 words, 29 figures, 9 tables, 2 animations.
Keywords: ash, ice, and sulfate particles; Pinatubo Volcanic Cloud; satellite remote sensing.
Index Terms: 0370 Atmospheric Composition and Structure: Volcanic effects (8409); 0305 Atmospheric Composition and
Structure: Aerosols and particles (0345, 4801); 1640 Global Change: Remote sensing.
Received 25 October 2003; Revised 2 March 2004; Accepted 22 March 2004; Published 8 May 2004.
Guo, S., W. I. Rose, G. J. S. Bluth, and I. M. Watson (2004), Particles in the great Pinatubo volcanic cloud of June 1991: The
role of ice, Geochem. Geophys. Geosyst.,5, Q05003, doi:10.1029/2003GC000655.
1. Introduction
[2] Mount Pinatubo is located at 1508
0
N,
12021
0
E in western Luzon, Philippines. After
weeks of precursory volcanic activity, a climactic
eruption occurred around 13:42 (local time) on
15 June 1991 and lasted for approximately 9 hours
[Wolfe and Hoblitt, 1996]. It is the largest eruption
G3
G3
Geochemistry
Geophysics
Geosystems
Published by AGU and the Geochemical Society
AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES
Geochemistry
Geophysics
Geosystems
Article
Volume 5, Number 5
8 May 2004
Q05003, doi:10.1029/2003GC000655
ISSN: 1525-2027
Copyright 2004 by the American Geophysical Union 1 of 35

of the past 25 years based on its eruption volume
(8.4– 10.4 km
3
total ejecta bulk volume estimated
by Scott et al. [1996]) and its aerosol perturbation
to the stratosphere [McCormick et al., 1995].
Huge masses of volcanic ash and gas, or mixtures
of them with hydrometeors, were directly emitted
into stratosphere and produced significant global
environmental, atmospheric, and climatic effects
for up to several years [McCormick et al., 1995;
Self et al., 1996; Robock, 2002]. The most
significant atmospheric effects include the follow-
ing: (1) The large stratospheric sulfate aerosol
loading with chemical and dynamic perturbations
affecting the stratospheric NO
2
, reactive chlorine,
and ozone concentrations and increasing the
stratospheric opacity [McCormick et al., 1995].
(2) The effects on the global radiative processes
causing the coexistence of cooling effects in the
troposphere (more solar radiation scattering back
to space) and warming effects in the stratosphere
(infrared absorptivity of stratospheric aerosols)
[McCormick et al., 1995]. Summer surface cool-
ing and winter surface warming were found in the
northern hemisphere 1 – 2 years after eruption
[Robock, 2002]. (3) The destruction of strato-
spheric ozone occurring due to both heteroge-
neous reactions occurring on the surface of
sulfate aerosol (similar to heterogeneous reactions
that occur within polar stratospheric clouds re-
sponsible for the Antarctic Ozone Hole [Solomon
et al., 1993]) and circulation changes after the
eruption [Kinne et al., 1992]. (4) Climatic effects
which lasted for up to several years after eruption
[Robock, 2002].
Table 1. Characteristics of TOMS, HIRS/2, and AVHRR Satellite Sensors Used in This Paper
TOMS
a
AVHRR
b
HIRS/2
b
Spatial Resolution: 50 km (nadir) Spatial Resolution: 4.4 km (GAC) Spatial Resolution: 17.5 km (nadir)
Channel Wavelength, nm Usage Channel Wavelength, mm Usage Channel Wavelength, mm Usage
1 308.6 SO
2
4 10.3 – 11.3 Ash, ice 5 13.97 ash, ice, sulfate
2 313.5 SO
2
5 11.5 –12.5 Ash, ice 6 13.64 ash, ice, sulfate
3 317.5 SO
2
7 13.35 ash, ice, sulfate
4 322.3 SO
2
8 11.11 ash, ice, sulfate
5 331.2 SO
2
, AI 9 9.71 ash, ice, sulfate
6 360.48 AI 10 8.16 ash, ice, sulfate
11 7.33 SO
2
c
a
McPeters et al. [1998].
b
Yu and Rose [2000].
c
Prata et al. [2003].
Table 2. Data Sets Used in This Study From AVHRR and HIRS/2
a
Date
NOAA-10
NOAA-11 NOAA-12
HIRS/2 AVHRR HIRS/2 AVHRR HIRS/2
6/15/91 06:30 06:30 10:20&12:06
10:26 17:58&19:41 18:08 10:53
6/16/91 00:52 07:26 07:28 10:00&11:44 01:18
11:44 20:05 20:05 12:20
6/17/91 00:12– 00:33 07:19&09:00 07:25– 08:43 11:21&13:01 00:39– 02:01
11:39– 13:05 19:16&20:59 19:16 – 20:59 12:07 – 13:31
6/18/91 01:55– 04:58 05:51&07:25&08:43 07:26– 08:43 12:41&14:34&16:08 02:21 – 05:25
12:41 – 16:08 20:47&22:22&23:59 20:47 – 21:22 12:53 – 16:05
6/19/91 01:58 – 03:41 08:38&10:20&12:04 07:25 – 08:43 12:21&14:05&15:44 02:20 – 05:25
12:41 – 16:08 20:47 – 21:22 12:53 – 16:05
a
Note: All times listed are UTC. The AVHRR and HIRS/2 sensors are on board the same platform. The AVHRR image times are the central pixel
sensing time if only one orbit involved or the central pixel sensing times of each individual orbit. The HIRS/2 image time is either the central pixel
sensing time if only one orbit involved or the central pixel sensing times of the starting and ending orbits if more than more orbit involved.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
2of35

[3] In spite of its significance, Pinatubo has not
been studied with volcanic cloud particle retrieval
methods. AVHRR data have been used to trace the
spatial development of the Pinatubo volcanic
clouds and to observe gravity currents [Long and
Stowe, 1994; Holasek et al., 1996; Self et al.,
1996]. Study of sequential AVHRR imagery was
challenging because of its scale and tropical loca-
tion: (1) The Pinatubo clouds are so widespread
that multiple orbits must be combined to see the
whole volcanic cloud; (2) it is sometimes difficult
to distinguish the Pinatubo volcanic clouds
from other interferences (e.g., tropical convective
clouds, upper tropospheric saharan dust, and
Typhoon Yunya), some of which also have nega-
tive BTD (Brightness Temperature Difference of
band 4 and 5) values; (3) only the central portions
of the volcanic ash clouds can be detected by
the split-window method (negative BTD value
scheme) as the high water vapor contents below
the ash cloud cause the positive shift of the BTD
values of the ash dominant pixels [Yu et al., 2002];
and (4) during the first 20 hours after the erup-
tion, parts of the ash cloud remained opaque to
infrared sensors owing to its high optical thickness.
[4] The separation of the gas-rich portions and ash-
rich portions of volcanic clouds have been ob-
served in most eruptions studied with satellite data
[Bluth et al., 1994; Rose et al., 1995; Schneider et
al., 1999; Shannon, 1997; Mayberry et al., 2002;
Constantine et al., 2000] with the notable excep-
tion of 1992 Mount Spurr eruption [Schneider et
al., 1995; Rose et al., 2001]. These separations
might be due to the fact that SO
2
rises higher in the
plume than the volcanic ash [Rose et al., 2000]
over the eruption vent; differences of wind direc-
tions and speeds at different altitudes will cause
horizontal separation.
[5] In this paper, we apply methods for IR retriev-
als of particles in the Pinatubo cloud for the first
several days of its atmospheric residence. Fourteen
AVHRR ash maps and twenty-eight HIRS/2
images the first five days after eruption are ana-
lyzed to study the particle properties within the
Pinatubo volcanic clouds using the AVHRR two-
band split-window retrieval [Wen and Rose, 1994]
and HIRS/2 multiband retrieval [Yu an d R ose,
2000], respectively. The atmospheric correction
method developed by Yu et al. [2002] is applied
to improve the AVHRR retrievals. The adverse
effects on the AVHRR two-band split-window
retrieval caused by the contrary effects of ash
and ice are studied. Simultaneous retrieval of
ash, ice, and sulfate properties using multiband
HIRS/2 data for the Pinatubo volcanic clouds
allows us to consider how the different particles
may interact. The retrieved ash properties and
derived ash removal rates are compared with those
of the 1982 El Chicho´n and other smaller erup-
tions. The movement of Pinatubo’s SO
2
and ash
clouds are also investigated with the NASA GSFC
Figure 1. Plot to demonstrate the proportional con-
tribution to optical depth of the various kinds of particles
in the Pinatubo volcanic cloud based on the modified
method of Yu and Rose [2000]. Three curves represent-
ing the optimum uniform sizes of the three particle types
are plotted along with the sums in parentheses. A similar
result using only two particle types (ash and ice) is also
shown for comparison in red. The points plotted are the
average values of optical depth observed in the HIRS/2
volcanic cloud of 11:44 UT (16 June 1991). The
residuals of the deviation between the points and the two
summed optical depth curves are used to find the best fit.
Note that the optimal sizes for ash and ice do not differ
much whether the 2 or 3 component match is attempted.
Note also the total optical depth for sulfate is small.
These observations show that the proportions of ice and
ash dominate the signal.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
3of35

isentropical wind trajectory results to estimate the
altitudes of the SO
2
-rich and ash-rich clouds.
2. Satellite Data
[6] The characteristics of multispectral satellite
sensors used in this paper are summarized in
Table 1.
[7] The Pinatubo fine ash (1–12 mm) cloud
remained visible to the AVHRR sensor for
104 hours after the start of eruption. Fine ash (1–
15 mm; reflecting a slightly different size discrim-
ination sensitivity from AVHRR) was detected by
HIRS/2 sensor for 111 hours. The 14 AVHRR and
28 HIRS/2 satellite maps used in this paper are
summarized in Table 2. The first three days’
TOMS SO
2
and TOMS AI (Aerosol Index, see
definition below) maps (TOMS AI is only visible
for the first 46 hours after eruption) are also used
to supplement the IR retrievals.
3. Methodology
3.1. Volcanic Ash Detection With
TOMS AI
[8] Volcanic ash and other types of aerosols can be
qualitatively detected by the TOMS AI (Aerosol
Index) method. AI is defined as the difference of
the measured spectral contrast between the
331.2 nm and 360.48 nm radiances and the spectral
contrast between the 331.2 nm and 360.48 nm
radiances calculated for a Rayleigh scattering at-
mosphere and Lambertian surface [Seftor et al.,
1997; Krotkov et al., 1999]. Absorbing aerosols
(volcanic ash, smoke, desert dust) usually have
Figure 2. Logarithmic plots of the residual from fit for
various two-component combinations of particles in the
Pinatubo volcanic cloud. A well-defined single residual
minimum for ash-ice mixtures is identified in the results
(Figure 2a). The plots with sulfate (Figures 2b and 2c)
show that the best fit represents a slight minimum point
along two trough-like valleys. This suggests that the
sulfate sizes and masses are not as well constrained as
the ash-ice relations. The actual retrieval we use is three-
dimensional, so visualization of ash-ice-sulfate relation-
ships is difficult. Figures 1 and 2 are included to help the
reader visualize how the multiband retrieval [Yu and
Rose, 2000] works.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
4of35

positive AI values while non-absorbing aerosols
(e.g., sulfate aerosol) usually have negative AI
values [Seftor et al., 1997].
3.2. AVHRR Two-Band Split-Window
Retrieval
[9] Two thermal bands (band 4, 10.3 to 11.3 mm,
and band 5, 11.5 to 12.5 mm) of AVHRR have been
Figure 3. Brightness temperature differences of band
4 minus band 5 (BTD) versus band 4 brightness
temperature (T4) for pixels from the 16 June 20:05
(GMT) Pinatubo volcanic cloud measured by AVHRR.
The negative BTD value region (15.9% of all pixels)
represents pixels dominated by ash particles. The upper
field (BTD > 3.0) region (4.1% of all pixels) represents
an ‘‘ice region’’ where ice is the dominant particle type
in the pixel region. The triangular region (1.7% of all
pixels) bounded by points B, C, and D represents an
atmospheric moisture/vapor correction region which
could reflect optically thin, ash-dominant pixels. The
region labeled ‘‘mixtures’’ (78.3% of all pixels)
represents pixels which appear to be the result of
significant masses of both ash and ice, consisting either
of both ice and ash or hybrid mixtures (icy ash balls).
The atmospherically corrected AVHRR two-band split-
window method can detect ash pixels below the line AE
(BTD < 0) and within triangular area BCD. The ice
signals might be counteracted by ash signals in the
AVHRR retrieval. Many large stratospheric volcanic
clouds have ash regions which extend to much lower
BTD values, as low as 15 to 20, and the lack of such
low values here is another sign of ice.
Figure 4. Brightness temperature differences (BTD)
of band 4 minus band 5 versus band 4 brightness
temperature (T4) from NOAA 12 AVHRR ash maps
from Rose et al. [1995]. (a) Rabaul eruption at 09:00
(GMT) on 19 September 1994, showing ice particles
within the volcanic cloud which have positive BTD
values. (b) Klyuchevskoi eruption at 06:40 (GMT) on
1 October 1994, showing typical ash dominant volcanic
clouds, which have negative BTD values.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
5of35

used to distinguish volcanic clouds from meteoro-
logical clouds [Prata, 1989a]. Band 4 minus band
5 brightness temperature differences (BTD) usually
have negative values for silicate particle dominated
volcanic clouds [Prata, 1989a; Wen and Rose,
1994; Schneider et al., 1999] and positive values
for meteorological clouds [Yamanouchi et al.,
1987]. The reason for this difference is mainly
due to the fact that silicate particles have a stronger
dispersive nature than water/ice particles [Prata,
1989b]. The effective particle size, the optical
depth of the volcanic cloud and the mass of fine
ash or ice in the volcanic cloud can be retrieved by
using the model developed by Wen and R o se
[1994]. The maximum uncertainty for the AVHRR
two-band split-window retrieval is 53% as summa-
rized by Gu et al. [2003].
[10] However, owing to the high water vapor
content in the moist tropical troposphere under
Figure 5. NASA TOMS maps: (a) SO
2
cloud map and (b) AI map, showing the SO
2
and ash clouds from the low-
level activity before the climactic eruption at 05:41 (GMT) on 15 June. Latitude and longitude grid spacing is 30.
The SO
2
values are represented in Dobson Units (DU), and AI is unitless. The image sensing time is approximately
the central pixel sensing time of the image in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
6of35

the volcanic clouds, the BTD values of some ash
particles (boundary, low BTD value regions) will
be shifted to positive values [Yu et al., 2002]. Thus
only the central parts of volcanic clouds formed
from Pinatubo can be detected by the traditional
negative BTD retrieval scheme. An atmospheric
moisture correction method, developed by Yu et al.
[2002] to identify the ash-dominant pixels which
might have a positive BTD value due to the water
vapor interference, is used here. The atmospheric
moisture correction method uses both the BTD
value and cloud temperature (T4) to identify the
volcanic ash cloud.
[11] The ice retrieval used in this paper is based on
Rose et al. [1995]. The theoretical calculated grid
plots representing the look-up tables are calculated
by using the same parameters in calculating the ash
look-up tables, except the refractive index and
density of ice are used.
Figure 6. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method [Yu et al., 2002] applied. Latitude and longitude grid spacing is 30. The SO
2
values are
represented in Dobson Units (DU), and BTD values are represented in degrees Kelvin. The image sensing time is
approximately the central pixel sensing time of the image in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
7of35

[12] In this paper, the volcanic ash is assumed to
contain mainly andesite particles with 54% SiO
2
due to the availability of refractive indices for this
composition. Bernard et al. [1996] found that most
of the erupted magma was dacite, with 64.5% SiO
2
.
The retrieval differences should be negligible based
on the use of basalt (53.25% SiO
2
) and rhyolite
(73.45% SiO
2
) for the comparison purpose. A
lognormal particle size distribution with a standard
deviation of 0.74 [Wen and Rose, 1994] is used.
Both the andesite and ice particles are assumed
to be spherical, with densities of 2.6 g/cm
3
and
0.917 g/cm
3
, respectively [Neal et al., 1994; Scott et
al., 1996]. The refractive indices of andesite and ice
particles are based on Pollack et al. [1973] and Kou
et al. [1993].
3.3. HIRS/
/2 Multiband Retrieval
[13] A multiband retrieval using six infrared bands
(bands 5– 10) of HIRS/2 data were used for esti-
mates (optical depth, effective radius, mass) of ice,
Figure 7. NOAA-12 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximate the central
pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
8of35

ash and sulfate particles in the Pinatubo volcanic
clouds, following the work of Yu and Rose [2000].
The retrieval compares observed HIRS/2 radiances
with a lookup table of simulated values based on
MODTRAN [Berk et al., 1989] calculations of
volcanic clouds containing reference mixtures of
particles at the same heights of the Pinatubo cloud.
We further improved the calculations of simulated
cloud particle mixtures for this study by using a
forward model developed by Watson et al. [2003],
which can calculate spectral transmission through
volcanic clouds caused by different solid species, or
mixtures of species. SO
2
cloud maps from TOMS
and TOVS data, ash cloud maps from AVHRR, and
TOMS AI are used to select volcanic cloud areas to
carry out multiband retrievals using HIRS/2 data.
The ice, ash and sulfate particles in each forward
model calculation are assumed to be spherical and of
uniform size. A total of 144, 627 unique mixtures of
ice, ash and sulfate were examined systematically
Figure 8. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
9of35

using typical size values for each species (diameter
ranges of 1– 30 microns for ice, 1 –15 microns for
ash, and 0.1 to 0.5 microns for sulfate). A non-
negative least squares analysis is then used to
search for the most compatible results, and this
is shown graphically for two-component mixtures
for one of the Pinatubo clouds in Figures 1 and 2.
The method produces a pronounced minimum of
likely ice and ash proportions (Figure 2a) and a
weak minimum for mixtures involving sulfate
(Figures 2b and 2c). Refractive index data for
andesite determined by Pollack et al. [1973] was
used for the Pinatubo ash. Yu and Rose [2000]
investigated and discussed the errors involved in
this method. The assumption of spherical size for
particles is incorrect, and the implications of this
are not explored here. Riley et al. [2003] have
shown that basaltic and andesitic volcanic ash is
more nearly spherical (aspect ratio of 1.5 – 1.7) than
rhyolitic ash (aspect ratio of 2.2), but the effects
of this on radiative transfer are uncertain. The
assumption of uniform size was determined to have
Figure 9. NASA TOMS maps: (a) SO
2
cloud map and (b) AI map. Latitude and longitude grid spacing is 30°. The
SO
2
values are represented in Dobson Units (DU), and AI is unitless. The image sensing times are approximately
the central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
10 of 35

a significant impact by causing an overall under-
estimate of mass by approximately 30 – 40%. The
assumption of negligible scattering is also debat-
able and untested. The mean (and maximum)
uncertainties for the HIRS/2 multiband retrieved
effective radius, optical depth are 40 (60)%, and 30
(45)%, respectively for ash and ice, and higher for
sulfate [Yu and Rose, 2000]. Mass estimates are of
course based on both the radius and optical depth
estimates and therefore have larger errors. Overall
there is much uncertainty in this technique and it is
discussed much more by Yu an d Ro s e [2000].
Usually, although both AVHRR and HIRS-2 data
are available on volcanic clouds from polar-orbiting
NOAA satellites, we use two band AVHRR data for
ash estimates. In this paper, however, we have used
HIRS-2 simultaneous retrievals of the properties of
silicate ash, ice, and sulfate particles because ash,
ice, and/or particles that are mixtures of the two
are thought to coexist with sulfate in the Pinatubo
Figure 10. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30°. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
11 of 35

cloud and this precludes meaningful two band
retrievals.
3.4. Wind Trajectory Modeling
[14] An isentropic wind trajectory model devel-
oped by scientists from NASA GSFC (Goddard
Space Flight Center) [Schoeberl et al., 1992, 1993]
is used in this study to estimate cloud height.
Shannon [1997] used the NASA wind trajectory
model to view the drifting of volcanic gas clouds in
three dimensions. By comparing cloud positions
derived from the satellite sensors, the NASA wind
trajectory model can constrain the height of
volcanic cloud, which is needed in retrieval algo-
rithms. The atmospheric temperature and wind
(direction and speed) profiles at different locations
(grid points of the model domain) are the key inputs
to run the trajectory model. No real time atmo-
spheric sounding data were found within 500 miles
around Mt. Pinatubo during the eruption period
in the GSFC DAO (Data Assimilation Office) or
Figure 11. NOAA-12 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30°. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
12 of 35

UARS UKMO (UK Meteorological Office) data-
bases. Therefore the inputs are based primarily on
the NCEP/NCAR (National Center for Environ-
mental Prediction/National Center for Atmospheric
Research) Reanalysis data set [McPherson et al.,
1979].
3.5. Other Cloud Maps Used in the Paper
[15] TOVS and TOMS SO
2
maps are also used in
this paper for SO
2
cloud height estimation, through
comparison with the wind trajectory results
at different levels between 20 and 30 km. The
generation of TOMS and TOVS SO
2
maps are
described in detail by Guo et al. [2004].
4. Results
4.1. Coexistence of Ash and Ice in the
Pinatubo Cloud
[16] Pinatubo’s volcanic clouds differ from other
volcanic clouds seen with IR sensors as described
below. Figure 3 shows the BTD (brightness tem-
Figure 12. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30°. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
13 of 35

perature differences of band 4 and 5) versus T4
(band 4 brightness temperature) plot of the 16 June
07:28 (GMT) AVHRR ash map (Figure 3). Similar
patterns of BTD versus T4 were found for the other
13 AVHRR Pinatubo ash maps during the first five
days after eruption. A majority of pixels have
positive BTD values, which, together with the
low temperatures, are consistent with the presence
of significant ice within the volcanic cloud. The
region labeled ‘‘ice-region’’ (BTD > 3.0) reflects
pixels dominated by ice. The BTD = 3.0 line
represents the maximum BTD value of ‘‘clear
air’’ pixels over the tropical ocean [Yu et al. ,
2002]. Pixels with BTD < 0 reflect the influence
of volcanic ash. The triangular region bounded by
points B, C, and D, which have BTD values
between 0.0 and 3.0 and T4 values between 250
and 280, is the atmospheric correction region for
Figure 13. NASA TOMS maps: (a) SO
2
cloud map and (b) AI map. A tiny amount of ash signals appears on the AI
map, which might be due to the increase of sulfate aerosol within the cloud (see discussion in text). Latitude and
longitude grid spacing is 30°. The SO
2
values are represented in Dobson Units (DU), and AI is unitless. The image
sensing times are approximately the central pixel sensing time of the starting and ending orbits in GMT as indicated in
the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
14 of 35

volcanic ash using the Yu et al. [2002] method.
This region is dominated by ash and the shift of
BTD value to values >0 is caused by lower
tropospheric water vapor and very low optical
depth. Pixels whose BTD values are between 0.0
and 3.0 and outside the atmospherically corrected
triangle region are interpreted as ‘‘mixtures’’ of ash
and ice (ash-ice hybrid, ash nucleated ice particles,
and pure hydrometeors). Two analogous plots from
Rose et al. [1995] are presented again in this paper
to show patterns seen in ash-dominated and ice-
dominated volcanic clouds, respectively (Figure 4).
Figure 4a is the retrieval grid of the 19 September
Rabaul (Papua, New Guinea) volcanic cloud at
09:00 (GMT), showing the ice-dominated volcanic
cloud which mainly has positive BTD values.
Figure 4b is the retrieval grid of the 1 October
1994 Klyuchevskoi (Kamchatka, Russia) volcanic
cloud at 06:40 (GMT), showing the ash-dominated
volcanic cloud with mainly negative BTD values.
Figure 14. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30°. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
15 of 35

Through comparison, we conclude that Pinatubo
clouds contain ice, ash and perhaps the mixtures of
the two (hybrid mixtures, or icy ash balls, or ash
nucleated icy balls) and the signals seen in the two-
band retrieval reflect hybrids of ash and ice. This
indicates that the ash retrievals done with AVHRR
data are unlikely to be as useful as they have been
for other eruptions, because we cannot assume that
ice has no effect on the BTD values. More work is
needed in the future in the conflicting effects of ash
and ice in the AVHRR two-band retrievals.
[17] An abundance of H
2
O sources also makes
the coexistence of ash and ice within Pinatubo
cloud possible. Rutherford and Devine [1996]
point out that for Pinatubo, 5.1 to 6.4 percent
of the preeruption magma is H
2
O. If the total
erupted bulk volume of 8.4–10.4 km
3
[Scott et
al., 1996] and bulk density of 1.5 g/cm
3
[Scott et
al., 1996] are used, then the magma released
643– 998 Mt H
2
O. Gerlach et al. [1996] estimate
a maximum of 491– 921 Mt water in the erupted
plume based on both modeling of the composi-
Figure 15. NOAA-12 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30°. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
16 of 35

tion of the pre-eruption vapor and degassing of
melts. The Pinatubo plume likely also entrained
abundant H
2
O from a very moist tropical atmo-
sphere as suggested by numerical simulation
conducted by ATHAM (Active Tracer High Res-
olution Atmospheric Model) [Herzog et al.,
1998; Guo et al., 2000]. The Pinatubo volcanic
cloud was also fed from significant co-ignimbrite
cloud formation which can further enhance en-
trainment [Dartevelle et al., 2002]. In addition,
there was an active tropical cyclone, Typhoon
Yunya, passing the Pinatubo area during the
eruption period. Thus very large amounts of
water vapor were convected with the volcanic
cloud which likely led to significant amounts of
ice, enough to overwhelm the effects of ash. The
various sources of H
2
O can not be individually
assessed in this study, but the larger ice propor-
tions mean that the two-band ash retrieval
could not be reliably used alone to estimate ash
Figure 16. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30°. The SO
2
values are represented in Dobson Units
(DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the central pixel
sensing time of the starting and ending orbits in GMT as indicated in the figures. Detection of the movement of
volcanic clouds is interfered with by a tropical cyclone developed in the Bay area between Thailand and India.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
17 of 35

masses, and ash maps based on this method are
also inadequate.
4.2. Two-Dimensional Maps of the
Pinatubo Ash and SO
2
Clouds
[18] In spite of the problems presented by ice, we
mapped the two-dimensional (2-D) distribution of
the Pinatubo ash clouds using two-band AVHRR
BTD maps for several days after the 15 June 1991
eruption. We also used TOMS AI data for mapping
ash clouds, because we think that presence of ice
causes the AVHRR maps to underestimate 2-D
areas. We compare these ash maps with SO
2
maps derived from TOVS and TOMS at the same
time to compare the cloud areas and positions
(Figures 5– 21, Animations 1 and 2). Overall these
maps demonstrate that SO
2
persists in the cloud
much longer than volcanic ash and the area repre-
sented by ash-dominant pixels decreases strongly
after 24 hours and is 20% of its maximum after
Figure 17. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30°. The SO
2
values are represented in Dobson Units
(DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the central pixel
sensing time of the starting and ending orbits in GMT as indicated in the figures. Detection of the movement of
volcanic clouds is interfered with by a tropical cyclone developed in the Bay area between Thailand and India.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
18 of 35

70 hours (Figure 23). Because of the dominating
effects of ice, the areas shown in the AVHRR maps
likely underestimate the area where ash exists in
the volcanic cloud.
[19] These maps offer little evidence that SO
2
-rich
and ash-rich portions of the Pinatubo volcanic
cloud separated as they did in the case of El
Chicho´n in 1982 [Schneider et al., 1999]. SO
2
clouds may be traveling to the west slightly faster
than ash (Figures 6– 21), due to the fact that wind
directions at all levels in the stratosphere are
consistently easterly but intensities increase with
altitude above the tropopause (Figure 27).
4.3. Ash Retrievals Using the AVHRR
Split-Window and HIRS/
/2 Multibands
[20] The mean particle effective radius, the optical
depth of the volcanic cloud and the mass of
fine ash were retrieved for each AVHRR map using
the retrieval model developed by Wen and Rose
Figure 18. NOAA-12 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
19 of 35

[1994] (Table 3), giving us data to assess changes
of ash particles during the first few days of the
Pinatubo volcanic cloud. The atmospheric correc-
tion method developed by Yu et al. [2002] was
applied to retrievals due to the moist tropical
troposphere (high water vapor contents below
the volcanic ash cloud). The BTD cutoff values,
band 4 and band 5 temperatures of underlying
surface, and cloud temperatures used in the atmo-
spheric correction method are summarized in
Table 4. Opaque cloud areas exist in the AVHRR
ash images collected during the first 20 hours
after eruption, so that the retrieved fine ash mass
for these images are only lower limits of the
volcanic fine ash mass for the first 20 hours after
eruption.
[21] Simultaneous retrievals of ash, ice and sulfate
properties (mass, mean effective radius, optical
depth) were conducted using HIRS/2 data and a
sulfate and silicate ash retrieval model developed
by Yu and Rose [2000]. Because they are based on
Figure 19. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
20 of 35

multiband data and retrieve all three particle types
simultaneously, the HIRS/2 results are probably
more robust and more indicative of what really
happened. Figure 1 shows the best fit, calculated
optical depth curves for mixtures of ash, ice, and
sulfate and mixtures of ash and sulfate with
observed optical depths at different wavelengths.
It shows that sulfate particles contribute less to the
total optical depth than ash and ice particles.
Figure 2 is best fit residual surface graph showing
differences between observed and calculated opti-
cal depths based on all combinations of ash, ice,
and sulfate particle sizes using the whole look-up
table. It indicates that sulfate particle size selection
has the least influence on the residual. Thus accu-
rate retrievals of sulfate properties are more prob-
lematic than those of ash and ice. Figures 1 and 2
are meant to give the reader insight into the
multiband retrieval. Simultaneous retrievals of par-
ticle properties (effective radius, optical depth,
mass) of fine ash (1 –15 mm), fine ice (or ash
covered by ice) (1–30 mm), and sulfate (0.1 –
Figure 20. NOAA-11 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
21 of 35

0.5 mm) particles are carried out using HIRS/2
data for up to 111 hours after eruption (Table 5).
Figure 22 shows the 16 June 11:44 (GMT) burden
maps of ash, ice, and sulfate from HIRS/2 multi-
band retrievals. More sulfate remains to the east of
the cloud while more ash is to the west of the
cloud.
[22] Generally the HIRS/2 mass results for ash are
20–50% higher than AVHRR results for the first
three days and then fall to values that are nearly the
same after 3 days (Figure 23 and Tables 3 and 5).
The uncertainties for AVHRR and HIRS/2 re-
trievals are ±53% [Wen and Rose, 1994; Gu et
al., 2003] and ±85% [Yu and Rose, 2000],
respectively, though the counteracting effects of
ash and ice challenge the validation of AVHRR
two band retrievals (the whole pixel is classified
as either ash only or ice only in the AVHRR two-
band retrieval). The ash masses estimated by
AVHRR data are low because ash is counteracted
by ice in part and because there are complexities
Figure 21. NOAA-12 maps: (a) TOVS SO
2
cloud map and (b) AVHRR BTD (ash cloud) map with atmospheric
correction method applied. Latitude and longitude grid spacing is 30. The SO
2
values are represented in Dobson
Units (DU), and BTD values are represented in degrees Kelvin. The image sensing times are approximately the
central pixel sensing time of the starting and ending orbits in GMT as indicated in the figures.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
22 of 35

in the retrieval caused by tropical clouds. The
AVHRR results after 18 and 19 June are unreli-
able because the ash signal is very weak. The
overall ash removal rates derived from both
sensors are similar, and thus may be indicative
of the real values.
[23] The total masses of fine ash from both HIRS/2
and AVHRR show decreasing trends with time
while the mean effective radii remain almost con-
stant (Figure 23 and Tables 3 and 5). The HIRS/2
values for ash are probably more accurate because
they incorporate and measure the influence of ice.
As the data are timed with respect to the start of
the eruption, the first three points are within
theeruptionperiodwithanincreasingtrend
(Figure 23). About 92% of the fine ash (1–12 mm)
detected by AVHRR and 89% of the fine ash (1 –
15 mm) detected by HIRS/2 were removed from
stratosphere 73 hours after eruption (about three
days) and about 99% of the detected fine ash was
removed from the atmosphere 110 hours after
eruption (Tables 3 and 5). The volcanic ash cloud
areas for both HIRS/2 and AVHRR retrievals
increased for the first 24 hours after eruption, and
then decreased dramatically to 20% of its maxi-
mum in 70 hours and to 5% of its maximum after
94 hours, during a period when the SO
2
cloud area
continuously increased (Figure 24 and Tables 3
and 5). The ash cloud areas based on HIRS/2
Table 3. Fine Ash (1– 12 mm) Retrieval Results for Pinatubo Volcanic Clouds Using AVHRR
a
Date
Time,
GMT
Hours After
Eruption
Ash Mass,
Mt
Mean
Effective
Radius, mm
Mean Particle
Radius, mm
Optical Depth
(11 mm)
Cloud Area
(10
6
km
2
)
6/15/91 06:30 0.82 25.75 9.56 3.20 2.65 0.665
10:20&12:06 5.53 33.74 10.21 3.41 2.44 1.312
17:58&19:41 13.14 33.98 9.50 3.18 1.82 2.042
6/16/91 07:16 25.58 20.44 7.90 2.64 1.48 3.880
10:00&11:44 29.18 19.14 8.82 2.95 1.65 3.085
20:05 38.40 15.20 8.38 2.80 1.32 3.535
6/17/91 07:19&09:00 50.48 10.05 9.59 3.21 1.56 3.267
11:21&13:01 54.50 12.11 8.32 2.78 1.46 2.760
19:16&29:59 66.94 6.91 8.67 2.90 1.45 1.649
6/18/91 05:51&07:25&08:43 73.60 2.71 8.43 2.82 1.12 0.780
12:41&14:34&16:08 80.73 3.28 8.67 2.90 1.53 0.295
20:47&22:22&23:59 88.70 3.72 9.48 3.17 1.30 0.246
6/19/91 08:38&10:20&12:04 100.67 1.58 8.30 2.78 1.22 0.137
12:21&14:05&15:44 104.36 0.89 8.69 2.91 1.54 0.107
a
Note: The image time is the central pixel sensing time if only one orbit involved or the central pixel sensing times of each individual orbit.
Hours after eruption is based on the middle of the sensing time in column 2 and the starting time of the eruption is 05:41 (GMT), 15 June 1991.
Table 4. Parameters Used in AVHRR Ash Retrieval (Table 3) With Atmospheric Correction
Date and Time, GMT
Band 4 Temperature
of Underlying Surface, K
Band 5 Temperature
of Underlying Surface, K
Cloud
Temperature, K
BTD
Value, K
6/15/91 06:30 285 283 200 <2.0
6/15/91 10:26&12:06 285 283 190 <2.0
6/15/91 17:58&19:41 285 282 200 <2.0
6/16/91 07:26 288 285 200 <1.0
6/16/91 10:00&11:44 285 282 200 <1.0
6/16/91 20:05 285 282 200 <0.5
6/17/91 07:19&09:00 290 288 200 <0.5
6/17/91 11:21&13:01 293 290 190 <0.5
6/17/91 19:16&20:59 285 282 190 <1.0
6/18/91 05:51&07:25&08:43 295 293 200 <0.5
6/18/91 12:41&14:34&16:08 288 285 200 <0.5
6/18/91 20:47&22:22&23:59 290 288 200 <0.5
6/19/91 08:38&10:20&12:04 288 285 200 <0.0
6/19/91 12:21&14:05&15:44 293 290 200 <0.0
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
23 of 35

retrieval are 20– 30% larger than those detected by
AVHRR 10 – 55 hours after eruption (Figure 24).
These differences reflect the significant amount of
ice interference to the AVHRR split-window
detection (Figure 25).
[24] On the basis of dense rock equivalent volume
of 3.7– 5.3 km
3
[Scott et al., 1996] and a dacitic
magma density of 2.4 g/cm
3
[Scott et al., 1996],
8880–12720 Mt of ash were erupted from Pina-
tubo on 15 June 1991. Therefore the fine ash mass
detected by the satellite is 0.6–0.9% of the total
ash mass, similar to the proportions determined by
Schneider et al. [1999] for the 1982 El Chicho´n
eruption. Most of the mass of erupted ash falls out
quickly because it is coarser than 100 mmin
diameter and is not seen by the satellite. The
retrieved mean effective radii from both AVHRR
model and HIRS/2 model are comparable to the
monomodal peak (11 mm in radius) found in distal
deposits collected further than 45 km from the vent
by Dartevelle et al. [2002].
4.4. Ice Retrievals Using the AVHRR
Split-Window and HIRS/
/2 Multibands
[25]Ice(1–30 mm in radius) properties were
retrieved using the refractive index and density
of ice using AVHRR data and following the work
of Rose et al. [1995] (Table 6). Higher masses
(2– 3 times) of ice than ash were detected in the
Pinatubo cloud and the ice masses decreased
rapidly, at rates that mimic the removal rates of
ash (Figure 25). The multiband HIRS/2 ice results
shown in Table 5 and Figure 25, like the ash
results, are probably more accurate than AVHRR
Table 5. Fine Ash (1– 15 mm), Fine Ice (1 –30 mm), and Sulfate (0.1 – 0.5 mm) Retrieval Results for Pinatubo
Volcanic Clouds Using HIRS/2 Data
a
Date Time, GMT
Hours
After
Eruption
Ash
Mass,
Mt
Ash
Effective
Radius, mm
Optical
Depth,
mm
Ice
Mass,
Mt
Ice
Effective
Radius, mm
Sulfate
Mass,
Mt
Sulfate
Effective
Radius, mm
Ash Cloud
Area
(10
6
km
2
)
6/15/91 06:30 0.82 30.85 8.62 1.31 46.03 23.37 3.33 0.15 0.739
10:26 4.75 37.61 8.36 1.51 62.47 25.54 3.52 0.16 1.045
10:53 5.2 48.57 8.55 1.53 80.95 23.75 2.38 0.21 1.537
18:07 11.58 45.55 8.28 1.64 84.44 24.77 3.52 0.20 2.329
6/16/91 00:52 19.18 37.47 8.90 1.15 52.01 23.75 8.07 0.15 3.512
01:18 19.62 41.94 8.97 1.45 59.75 23.94 7.22 0.14 2.923
07:28 25.78 28.35 7.66 1.30 40.32 27.47 6.67 0.14 4.071
11:44 30.05 38.24 7.31 1.69 46.53 24.50 5.81 0.16 3.719
12:20 30.65 26.59 6.84 1.12 49.48 26.73 9.57 0.15 3.595
20:05 38.4 30.54 9.33 1.01 42.84 26.02 6.30 0.14 4.019
6/17/91 00:12 – 00:33 42.69 37.00 7.55 1.38 32.93 27.74 8.66 0.16 3.376
00:39 – 02:01 43.65 18.66 7.31 0.94 35.27 25.93 7.77 0.13 3.511
07:25 – 08:43 50.38 27.70 8.30 1.01 39.86 26.16 7.24 0.14 3.618
11:39– 13:05 54.68 18.44 7.21 1.46 21.65 26.12 13.07 0.16 2.714
12:07 – 13:31 55.13 17.72 7.53 1.21 29.25 28.44 6.82 0.13 2.285
19:16 – 20:59 62.44 20.94 7.23 1.04 32.59 25.52 10.61 0.14 2.013
6/18/91 01:55 – 04:58 69.76 10.32 8.13 0.74 18.03 27.27 10.41 0.16 1.756
02:21 – 05:25 70.2 5.53 7.22 0.73 9.99 28.28 6.22 0.14 1.016
07:25 – 08:43 74.38 6.48 8.12 0.95 12.02 26.64 7.03 0.13 0.812
12:41 – 16:08 80.73 5.38 6.32 1.15 12.98 26.28 10.6 0.16 0.974
12:53 – 16:05 80.8 2.94 6.26 1.13 7.14 26.41 8.97 0.14 0.513
20:47 – 21:22 87.39 3.22 6.77 1.04 8.88 27.11 7.44 0.13 0.336
6/19/91 01:58 – 03:41 93.14 2.90 7.83 1.58 4.58 25.35 11.28 0.15 0.421
02:20 – 05:25 94.19 2.27 7.84 1.24 5.53 25.38 14.63 0.13 0.271
07:25 – 08:43 98.38 2.73 6.20 1.12 6.00 27.32 9.94 0.14 0.196
12:41 – 16:08 104.73 1.20 7.58 0.88 4.30 28.29 13.2 0.15 0.151
12:53 – 16:05 104.8 1.85 7.74 0.82 3.00 27.49 16.16 0.13 0.135
20:47 – 21:22 111.39 0.49 6.64 0.99 2.72 29.18 12.06 0.14 0.095
a
Note: The image time is either the central pixel sensing time if only one orbit involved or the central pixel sensing times of the starting and
ending orbits if more than more orbit involved. Hours after eruption is based on the middle of the sensing times in column 2 and the eruption
starting time is 05:42 (GMT), 15 June 1991.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
24 of 35

Figure 22
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
25 of 35

results as the coexisting ice, ash, sulfate particles
are retrieved simultaneously.
[26] The HIRS/2 mass results for ice are generally
higher than AVHRR results most of time, espe-
cially during the first three days (Figure 25 and
Tables 5 and 6). As with ash retrievals, the ice
masses estimated by AVHRR data are inaccurate
because ice is counteracted by ash in part and
because there are complexities in the retrieval
caused by tropical clouds. The ice masses from
both HIRS/2 and AVHRR show decreasing trends
over time while the mean effective radii remain
almost constant (Figure 25 and Tables 5 and 6).
Note that the first three points are within the
eruption period, and display an increasing trend
(Figure 25). About 90% of the ice detected by
AVHRR and HIRS/2 was removed from the
stratosphere 73 hours after eruption (three days),
and about 98% of the detected ice was removed
in 110 hours (Tables 5 and 6). The pattern of
decrease of both ice and ash was similar.
4.5. Sulfate Retrievals Using the
Multiband HIRS/
/2 Data
[27] HIRS/2 also includes retrieval data on sulfate
particles in the Pinatubo cloud (Table 5). The total
mass of sulfate increases with time while the mean
effective radius remains almost constant (0.14 –
0.20 ± 0.12 mm) (Figure 26 and Table 5). The
Figure 22. HIRS/2 maps: (a) ash burden, (b) ice burden, and (c) sulfate burden. Latitude and longitude grid spacing
is 30. The burdens are represented in Kt/km
3
. The image sensing times are approximately the central pixel sensing
time of the starting and ending orbits in GMT as indicated in the figures. Note the different scales used: the burdens of
sulfate are much less than those of ash or ice.
020 40 60 80 100 120
time after eruption onset (hour)
0
20
40
60
total fine ash mass (Mt)
eruption perioderuption period
total mass (AVHRR)
total mass (HIRS/2)
mean radius (AVHRR)
mean radius (HIRS/2)
0
5
10
15
20
25
mean radius (micron)
Figure 23. Fine ash mass and mean effective radius of
the 1991 Pinatubo volcanic ash cloud as a function of
time, using AVHRR split-window retrieval (ash particle
size 1– 12 mm) (Table 3) and HIRS/2 retrieval (particle
size 1– 15 mm) (Table 5). The ash masses decrease with
time, while the mean effective radius remains almost
constant. The error bars show the error range of the data
(proportional to the total) at a few specific points (see
text for discussion).
Figure 24. The 1991 Pinatubo volcanic ash cloud area
as a function of time based on AVHRR data (Table 3)
and HIRS/2 data (Table 5). The Pinatubo SO
2
cloud area
as a function of time based on TOMS data [Guo et al.,
2004]. The ash cloud area increases for up to 25 hours
after the start of the main eruption and then decreases
dramatically with time. The SO
2
cloud area increases
continuously over time. The error bars indicate the error
range of the data at a few specific points (see text for
discussion).
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
26 of 35

retrieved sulfate effective radii are similar to the
maximum pure sulfate aerosol radius (0.2 mm
[Turco et al., 1983]) and the effective radii of the
Pinatubo sulfate aerosols (0.1– 0.2 mm) just after
eruption by Russell et al. [1996]. About 12 – 15 Mt
sulfate was detected within the Pinatubo cloud
110 hours after the eruption (Figure 26). The
sulfate mass retrieved is significant (4Mt)
within an hour after eruption (exponential curve,
Figure 26), although the first four retrievals
(Table 5) are likely less reliable than later measure-
ments owing to high optical depths.
4.6. Analysis of Cloud Movements
[28]Holasek et al. [1996] reported the maximum
Pinatubo plume height could be as high as 39 km,
and the cloud heights were estimated at 20– 25 km
after 14 hours using cloud shadow and thermal
methods. Self et al. [1996] report the maximum
plume height could be >35 km and the plume
heights are 23– 28 km after 15 –16 hours using
temperature measurements. Read et al. [1993]
estimated that the peak SO
2
layer was at 26 km
using data from the Microwave Limb Sounder
(MLS) experiment on the Upper Atmosphere Re-
search Satellite (UARS). The neutral buoyant
regions of the Pinatubo aerosol were also estimated
by other measurements: 17– 26 km (lidar) by
Table 6. Ice (1– 30 mm) Retrieval Results for Pinatubo Volcanic Clouds Using AVHRR Data
a
Date Time, GMT
Hours
After
Eruption
Ice
Mass, Mt
Mean
Effective
Radius, mm
Mean
Particle
Radius, mm
Optical Depth
(11 mm)
6/15/91 06:30 0.82 44.76 20.01 6.69 2.23
10:20&12:06 5.53 59.91 22.39 7.49 3.84
17:58&19:41 13.14 60.76 26.09 8.73 2.36
6/16/91 07:16 25.58 43.54 26.89 9.00 1.83
10:00&11:44 29.18 28.34 25.44 8.51 1.67
20:05 38.40 29.49 29.02 9.71 1.48
6/17/91 07:19&09:00 50.48 15.86 29.97 10.03 1.59
11:21&13:01 54.50 21.88 29.86 9.99 1.47
19:16&29:59 66.94 14.90 27.87 9.32 1.94
6/18/91 05:51&07:25&08:43 73.60 4.78 29.06 9.72 1.14
12:41&14:34&16:08 80.73 5.97 29.77 9.96 1.64
20:47&22:22&23:59 88.70 5.83 29.99 10.03 1.30
6/19/91 08:38&10:20&12:04 100.67 2.88 30.00 10.03 1.23
12:21&14:05&15:44 104.36 1.58 30.00 10.03 1.59
a
Note: The image time is the central pixel sensing time if only one orbit involved or the central pixel sensing times of each individual orbit
connected by &. Hours after eruption is based on the middle of the sensing times in column 2 and the beginning time of the eruption is 05:41
(GMT), 15 June 1991.
020 40 60 80 100 120
time after eruption onset (hour)
0
10
20
30
40
50
60
70
80
90
100
total ice mass (Mt)
total mass (AVHRR)
total mass (HIRS/2)
mean radius (AVHRR)
mean radius (HIRS/2)
eruption perioderuption period
0
10
20
30
40
50
mean radius (micron)
Figure 25. Total ice (1–30 mm) mass and mean
effective radius of the 1991 Pinatubo volcanic ash cloud
as a function of time, using AVHRR split-window
retrieval (Table 6) and HIRS/2 retrieval (Table 5). The
total ice masses decrease with time, while the mean
effective radius remains almost constant. Ice masses are
179% to 213% of the ash masses shown in Figure 23
except for one point. The error bars indicate the error
range of the data at a few specific points (see text for
discussion).
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
27 of 35

DeFoor et al. [1992], 20 –23 km (balloon) by
Deshler et al. [1992], 17 –28 km (lidar) by Jaeger
[1992], and 17– 25 km (lidar) by Avdyushin et al.
[1993].
[29] To help reconcile cloud positions to atmo-
spheric conditions, forward isentropical wind
trajectory simulations (NASA Goddard Space
Flight Center) were run using Pinatubo’s eruption
start time. The overall movements of Pinatubo
SO
2
and ash clouds match the 25 km and 22 km
wind trajectories quite well, except during the first
15 hours (Table 7), meaning the central parts of the
SO
2
and ash clouds were 25 km and 22 km,
respectively. The wind directions are easterly at
all levels above the tropopause and the wind
intensity increases slightly with altitude. Thus
cloud heights can easily be determined as the
higher air parcels travel longer distances. The
inaccuracy for the wind trajectory results in
the first 15 hours after eruption is due to the
interference from the Typhoon Yunya as no real-
time atmospheric profiles were measured during
the eruption periods and the atmospheric profiles
Figure 26. Total fine sulfate (0.1– 0.5 mm) mass and
mean effective radius of the 1991 Pinatubo volcanic ash
cloud as a function of time, measuring using HIRS/2
data (Table 5). The total fine sulfate masses increase
with time, while the mean effective radius remains
almost constant. The error bars indicate the error
range of the data at a few specific points (see text for
discussion).
Table 7. Distances and Directions of Pinatubo Volcanic SO
2
and Ash Cloud Leading Edges and NASA GSFC Wind
Trajectory Results at 22 km and 25 km
a
Hours After
Eruption
b
SO
2
Cloud Leading
Edges
Ash Cloud Leading
Edges
Wind Trajectory
(25 km)
Wind Trajectory
(22 km)
Distance,
km
Direction,
degree
c
Distance,
km
Direction,
degree
Distance,
km
Direction,
degree
Distance,
km
Direction,
degree
0.82 1477.7
d
286.2
d
102.2
d
296.0
d
455.7 268.6 274.2 267.8
5.53 1555.3
d
273.4
d
998.2
d
295.3
d
919.8 270.1 625.4 270.0
13.14 1548.9
d
269.5
d
1288.9
d
279.0
d
1620.2 267.9 1199.8 268.1
22.04 2100.2 266.7 1446.4 282.0 1872.0 267.7 1232.2 267.5
25.58 3124.8 270.0 1842.8 273.1 2191.0 267.7 1325.7 267.1
29.18 3252.6 269.8 2107.0 271.6 2690.6 268.7 1650.3 267.5
38.40 3603.1 269.6 2551.4 269.8 3222.0 267.1 2018.1 268.7
50.48 3679.1 270.6 2913.0 269.5 3443.7 266.7 2685.8 269.5
54.50 3847.1 269.3 2696.0 270.8 3728.0 267.0 2932.8 270.3
66.94 5172.2 270.2 3020.2 267.7 4302.7 268.3 3308.3 269.3
73.60 5587.5 269.9 3681.0 273.5 5074.4 269.3 3763.9 268.8
80.73 5690.6 268.9 3820.5 274.7 5486.2 269.3 3827.2 270.0
88.70 6125.2 269.2 4278.2 277.2 5986.1 269.7 4184.7 270.7
100.67 7063.4 269.6 4974.7 275.0 6819.8 270.0 4924.5 273.0
104.36 7254.0 269.1 5226.6 274.5 7381.8 270.1 5211.6 272.3
a
All distances and directions are measured relative to Pinatubo volcano (15.13N, 120.35E).
b
One TOMS AI (16 June) and 14 AVHRR ash maps are used in the calculation; the ending times in the first column of the table correspond to
image sensing times.
c
Direction is labeled the same as wind direction; 0 degree indicates north direction and clockwise increase.
d
Mass weighted latitude and longitude locations are used instead of cloud leading edge locations due to opaque cloud area and clouds from
precursive activities.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
28 of 35

used in trajectory calculations are interpolated
from the measured profiles several hundred miles
away, possibly minimizing the influence of the
typhoon. The movement of the ash cloud over the
Bay of Bengal between Thailand and India also
does not match the trajectory results precisely
(Table 7 and Figure 27). The effects of a tropical
thunderstorm there may account for this slight
deviation. The SO
2
cloud moved about 7300 km
west and the ash cloud moved about 5200 km
west 104 hours after eruption (Table 7). The
locations of forward trajectory results at each
sensing time (Figure 27a) match the leading edge
cloud positions for each image at 22 km for ash
and 25 km for SO
2
(Figure 27b).
5. Summary of Main Results
[30] 1. Ice masses several times greater than ash
masses were present in the Pinatubo volcanic
cloud. These make accurate particle retrievals pos-
sible only with multispectral infrared data, such as
HIRS/2 and MODIS (of course MODIS did not
exist then).
[31] 2. Although it is difficult to locate the ash
cloud precisely, a slight vertical separation of the
SO
2
-rich volcanic cloud (25 km high) and ash-
rich volcanic cloud (22 km high) is consistent
with the apparently slightly faster westward
migration of SO
2
anomalies compared to ash. Lack
Figure 27. (a) Five days of NASA GSFC isentropic
wind forward trajectory results (22 and 25 km levels)
starting from 15 June 91 with each point representing
one image sensing time. (b) Leading edge SO
2
and gas
cloud locations with each point.
Figure 28. Total ash mass of volcanic ash clouds of
the 1991 Pinatubo (Tables 3 and 5), the 1982 El
Chicho´n [Schneider et al., 1999], the 1991 Hudson
[Constantine et al., 2000], and the 1992 Spurr [Rose et
al., 2001]. All of them decrease with time (see text for
discussion). The error bars indicate the error range of the
data at a few specific points (see text for discussion).
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
29 of 35

of windshear may have made detection of separa-
tion more difficult.
[32] 3. Both ash and ice decreased rapidly to a
small percentage (10%) of their peak masses
after 3 days of atmospheric residence in the
volcanic cloud.
[33] 4. Sulfate particle masses in the Pinatubo
clouds, retrieved using multispectral IR sensing,
increased by 2– 3 times (to 12 –15 Mt) during the
first 111 (five days) after eruption.
6. Discussion
6.1. How Much Ice is Too Much for
Accurate Two Band Retrievals?
[34] In the past, the two band IR split-window
retrievals on AVHRR, GOES, and MODIS have
been used for ash particle retrievals for many
eruptions. The Pinatubo case demonstrates that
ice can severely affect such retrievals and neces-
sitates a multispectral retrieval using HIRS/2 or
MODIS data if ice is abundant enough. Although it
is a non-marine volcano, Pinatubo’s environmental
variables were especially favorable for ice, because
of the moist tropical location and low elevation.
This example shows that when ice masses and ash
masses are similar in magnitude in volcanic clouds,
two-band retrievals are adversely affected.
[35] If two-band retrievals are done on a volcanic
cloud which also contains ice, the mass estimates
will be lower than the true values and the effective
radius values are also likely to be larger. We need
to develop ways to know when ice contents are
significant from the satellite data alone (see
Figures 3 and 4). One distinctive feature of the
Pinatubo cloud is that even though it is very large
and has very cold stratospheric temperatures and a
very large temperature contrast with the warm
Figure 29. Mean effective radius of volcanic ash
clouds of the 1991 Pinatubo (Tables 3 and 5), the 1982
El Chicho´n [Schneider et al., 1999], the 1991 Hudson
[Constantine et al., 2000], and the 1992 Spurr [Rose et
al., 2001] (see text for discussion). The error bars
indicate the error range of the data at a few specific
points (see text for discussion).
Table 8. Ash and Ice Removal Rates From Different Eruptions, Estimated With Remote Sensing
Volcano
Period of Measurement
(Hours After the End of Eruption)
Particle
Type
Mean Removal
Rate, Kt/hr
e-Folding
Time, hours Reference
1991 Pinatubo (HIRS/2) 4.1– 95.4 ash 482 24 this paper
1991 Pinatubo (HIRS/2) 2.6– 102.4 ice 819 30 this paper
1991 Pinatubo (AVHRR) 4.1– 95.4 ash 363 27 this paper
1991 Pinatubo (AVHRR) 4.1– 95.4 ice 648 27 this paper
1982 El Chicho´ n B 5.0– 68.0 ash 34.2 13 Schneider et al. [1999]
1982 El Chicho´ n C 7.1– 70 ash 98.7 15 Schneider et al. [1999]
1991 Hudson 2.0– 132.0 ash 21.8 30 Constantine et al. [2000]
June 1992 Spurr 12.5– 152.0 ash 2.35 143 Rose et al. [2001]
August 1992 Spurr 13.6– 83.4 ash 3.74 43 Rose et al. [2001]
September 1992 Spurr 8.0 – 70.0 ash 4.92 52 Rose et al. [2001]
2000 Hekla 6.2– 23.9 ice 48 8 Rose et al. [2003]
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
30 of 35

surface below, the brightness temperature differ-
ences are quite moderate and are rarely more than a
few degrees. We think this is due to the effects of
ice and possibly other components of the volcanic
cloud, and at least in the first two days, there could
be high dense meteorological clouds that suppress
the temperature contrasts.
6.2. Does Ice Enhance The Ash Fallout?
[36]Rose et al. [1995] first pointed out the
important role of ice in the ash-removal process,
and that ice-rich volcanic ash clouds may have a
shorter atmospheric residence time. Large
amounts of different kinds of hydrometeors (ice,
rain, hail, snow, sleet, etc.), especially ice, exist in
some volcanic clouds, such as the 1994 Rabaul
cloud [Rose et al., 1995], the 26 December 1997
Montserrat cloud [Mayberry et al., 2002], and the
2000 Hekla cloud [Rose et al., 2003]. One of the
principal roles of ice in the volcanic cloud is to
form ice-ash aggregates and thus accelerate the
ash fallout, as suggested by ATHAM simulation
results [Herzog et al., 1998; Textor, 1999] and
described by Rose et al. [2000]. Different particle
fallout velocities due to the different particle sizes
can cause gravitational collision, where large
particles collect smaller ones. Turbulent motions
of ash particles may increase the likelihood of
collision (R. Shaw, personal communication,
2003). The collisions result in particle aggregates
if binding forces between the particles are strong
enough. Moist particles have stronger short-range
surface tension forces than dry particles [Sparks et
al., 1997]. Therefore volcanic clouds containing
higher H
2
O content can expect to have a greatly
enhanced aggregation process. Fine ash particles
can also serve as active cloud nuclei to enhance
ice-ash aggregation as supported by ATHAM
simulation [Herzog et al., 1998] and ice can grow
on ash nuclei as described by Rose et al. [2003].
The ash aggregates usually have a higher surface
area relative to their masses and can be rapidly
filled up with ice by deposition of water vapor
and heterogeneous nucleation on the aggregates
[Wallace and Hobbs, 1997]. Because of the
abundant ice in the early Pinatubo volcanic cloud
suggested by our data analysis, there should be
water vapor from sublimation in the volcanic
cloud. We do not measure this nor do we
adequately consider the effects of this on our
results. The importance of this may extend to
other issues also, as water vapor should influence
SO
2
oxidation to sulfate and could cause interfer-
ence with SO
2
measurement using the IR (TOVS)
sensor.
[37] Figures 28 and 29 compare the retrieved fine
ash masses and mean effective radii for various
eruptions. Table 8 compares remote sensing data
which summarize the removal rates of ash particles
from various volcanic clouds. We note the follow-
ing in this table: (1) the e-folding times for ash
(24– 27 hours) and ice (27 –30 hours) at Pinatubo
are similar; (2) the e-folding times for the 1991
Hudson (21 hours), the 1982 El Chicho´n (13–
15 hours), and the 2000 Hekla (8 hours) eruptions
are shorter, indicating faster ash removal, while the
three 1992 Spurr eruptions have much longer
e-folding times (40– 140 hours), indicating slower
ash fallout. We need more data to see how these
data can be fully explained, and no simple expla-
nation emerges for the differences.
[38] The difference between e-folding times for
SO
2
(25days)andforash(24–27hours)is
striking, however. The similarity of removal rates
for ice and ash make it hard to imagine that ash and
ice were removed separately without aggregation.
Dartevelle et al. [2002] found one of the size peaks
of the fallen ash particles within 45 km of the vent
was 5.5 mm in radius. This is far too small to be
explained by simple fallout mechanism, and likely
Table 9. TOMS AI Values of the Pinatubo Volcanic
Clouds the First Week After Eruption
a
Date Time, GMT AI (Max Value) AI (Min Value)
6/15/91 03:26 7.5 0.1
6/16/91 2:01– 3:44 7.5 0.3
6/17/91 2:40– 5:45 7.0 1.3
6/18/91 4:14– 7:42 2.5 1.0
6/19/91 3:02– 7:45 1.5 1.0
6/20/91 3:38– 7:57 2.2 0.8
6/21/91 3:26– 8:15 1.7 1.0
a
Note: The whole volcanic cloud areas were accounted except when
the volcanic cloud was overlapped by dust storm or sand storm in the
regions of Middle East and Africa.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
31 of 35

reflects aggregation. Wiesn er et al. [2004] also
pointed out that only aggregation could explain
the fine-grained particles from the coast of Luzon
up to South China Sea, as the high-velocity east-
erly wind at upper-troposphere and stratosphere
would have transported fine-grained particles far
beyond South China Sea without aggregation.
6.3. Does the Increase in Sulfate Match
SO
2
Decrease Within the Whole Cloud?
[39] The sulfate mass retrieved is significant (4 Mt)
even within one hour after eruption (Table 5,
Figure 26). The observation of early sulfate based
on HIRS/2 data was also noted for the 1982 El
Chicho´ n eruption by Yu and Rose [2000], and
suggests that sulfate formation may be catalyzed
in eruption columns. The mass of sulfate particles
increases steadily for the 4 day period of analysis,
producing 8– 10 Mt additional sulfate mass. This
rate of increase can be compared to the SO
2
decrease determined by Guo et al. [2004] of about
4–8 Mt for the same time period. Given the mass
differences of sulfate and SO
2
, the two rates are
within 50% which suggests that conversion of SO
2
to sulfate is the main cause of sulfate increases seen.
[40] Non-absorbing aerosols, such as sulfate, can
produce negative AI signals [Seftor et al., 1997].
Negative TOMS AI signals existed on the Pinatubo
AI maps the first week after eruption (later the SO
2
cloud overlapped with dust storms and sandstorms
in the Middle East and Africa, making the negative
AI signals difficult to identify). The negative AI
signals were very weak (usually between 1 and 0)
(Table 9). The relatively weak sulfate signal might
be due to the larger particle sizes (0.15– 0.20 mm),
because only particles smaller than 0.05 mm can
produce strongly negative AI values [Rose et al.,
2003].
7. Conclusions
[41] We mapped the 2-D distribution of Pinatubo
ash clouds using two-band AVHRR ash maps and
TOMS AI maps. The cloud expanded for about
24 hours and had a maximum area of about 5 
10
6
km
2
and then decreased quickly to less than 1 
10
6
km
2
after about 3 days. Comparison of these
maps with SO
2
maps from TOMS and TOVS offer
little evidence that SO
2
-rich and ash-rich portions
of the Pinatubo volcanic cloud separated, though
the SO
2
cloud may be traveling to the west slightly
faster than the ash cloud. A slight vertical separa-
tion of SO
2
-rich (25 km high) and ash-rich
(22 km high) volcanic clouds is consistent with
the slight faster migration of SO
2
than ash westerly,
though uncertainty exists due to the difficulty in
precisely locating the ash cloud.
[42] The abundance of ice in the Pinatubo volcanic
clouds complicates the ash and ice retrieval results
by AVHRR two-band split-window method be-
cause of the contrary effects of ash and ice and
because of complexities caused by tropical clouds.
Multiband retrieval of ash, ice, and sulfate proper-
ties using HIRS/2 data is probably more indicative
of what really happened in this volcanic cloud.
[43] The multispectral remote sensing possible
from HIRS/2 [Yu and Rose, 2000] allows for a
more realistic sensing of particles in the case of
Pinatubo, where both ice and ash are abundant. It
also allows for retrieval of particle size, optical
depth, and masses of three different particle com-
positions (ash, ice, and sulfate). The ice and ash
masses in the Pinatubo cloud were about 80 Mt
(8– 16% of total magmatic H
2
O) and 50 Mt (0.6–
0.9% of total erupted ash) maximum, and declined
rapidly to less than 10% of their maximum within 3
days. We speculate that ash surfaces were enhanc-
ing ice formation and that fallout was also en-
hanced. The e-folding removal times estimated for
Pinatubo ice and ash were about 24 – 30 hours
which contrasts strikingly with comparable SO
2
e-folding (25 days).
[44] Sulfate particles were present in the Pinatubo
cloud even very soon after eruption and the masses
of sulfate increased during the first four days at a
rate consistent with the SO
2
decreases. Initially
detected sulfate mass was 3.3 Mt and then after
5 days was 12– 16 Mt.
Acknowledgments
[45]Song Guo acknowledges support from the U.S. National
Science Foundation (NSF EAR 01-06875) and a NASA
Graduate Student Fellowship. We acknowledge the help of
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
32 of 35

Leslie R. Lait at NASA Goddard Space Flight Center for the
use of a NASA wind trajectory model. We also thank A. J.
Prata from CSIRO in Australia for providing the raw TOVS
SO
2
retrieval codes. We also benefited from comments by A. J.
Prata, Hans Graf, Steven Self, and an anonymous reviewer.
References
Avdyushin, S. I., G. F. Tulinov, M. S. Ivanov, B. N. Kuzmenko,
I. R. Mezhuev, B. Nardi, A. Hauchecorne, and M. L. Chanin
(1993), Spatial and temporal evolution of the optical thickness
of the Pinatubo aerosol cloud in the northern hemisphere from
a network of ship-borne and stationary lidars, Geophys. Res.
Lett.,20, 1963 – 1966.
Berk, A., S. Bernstein, and D. C. Robertson (1989), MODTRAN:
A moderate resolution model for LOWTRAN 7, AFGL-TR-
89-0122, U.S. Air Force Geophys. Lab., Hanscom Air Force
Base, Mass.
Bernard,A.,U.Knittel,B.Weber,D.Weis,A.Albrecht,
K. Hattori, J. Klein, and D. Oles (1996), Petrology and geo-
chemistry of the 1991 eruption products of Mount Pinatubo,
in Fire and Mud: Eruptions and Lahars of Mount Pinatubo,
Philippines, edited by C. G. Newhall and R. S. Punongbayan,
pp. 767– 797, Philippine Inst. of Volcanol. and Seismol.,
Quezon City.
Bluth, G. J. S., T. J. Casadevall, C. C. Schnetzler, S. D. Doiron,
L. S. Walter, A. J. Krueger, and M. Badruddin (1994), Eval-
uation of sulfur dioxide emissions from explosive volcanism:
The 1982– 1983 eruptions of Galunggung, Java, Indonesia:,
J. Volcanol. Geotherm. Res.,63, 243 – 256.
Constantine, E. K., G. J. S. Bluth, and W. I. Rose (2000),
TOMS and AVHRR sensors applied to drifting volcanic
clouds from the August 1991 eruptions of Cerro Hudson,
in Remote Sensing of Active Volcanism,Geophys. Monogr.
Ser., vol. 116, pp. 45– 64, AGU, Washington, D. C.
Dartevelle, S., G. G. J. Ernst, J. Stix, and A. Bernard (2002),
Origin of the Mount Pinatubo climactic eruption cloud: Im-
plications for volcanic hazards and atmospheric impacts,
Geology,30, 663– 666.
DeFoor, T. E., E. Robinson, and S. Ryan (1992), Early lidar
observations of the June 1991 Pinatubo eruption plume at
Mauna Loa Observatory, Hawaii, Geophys. Res. Lett.,19,
187– 190.
Deshler, T., D. J. Hofmann, B. J. Johnson, and W. R. Rozier
(1992), Balloonborne measurements of the Pinatubo aero-
sol size distribution and volatility at Laramie, Wyoming
during the summer of 1991, Geophys. Res. Lett.,19, 199 –
202.
Gerlach, T. M., H. R. Westrich, and R. B. Symonds (1996),
Preeruption vapor in magma of the climactic Mount Pinatubo
eruption: Source of the giant stratospheric sulfur dioxide
cloud, in Fire and Mud: Eruptions and Lahars of Mount
Pinatubo,Philippines, edited by C. G. Newhall and R. S.
Punongbayan, pp. 415– 433, Philippine Inst. of Volcanol.
and Seismol., Quezon City.
Gu, Y., W. I. Rose, and G. Bluth (2003), Retrieval of mass and
sizes of particles in sandstorms using two MODIS IR bands:
A case study of April 7, 2001 sandstorm in China, Geophys.
Res. Lett.,30(15), 1805, doi:10.1029/2003GL017405.
Guo, S., W. I. Rose, G. J. S. Bluth, C. Textor, and H.-F. Graf
(2000), ATHAM model simulation of 15 June 1991 Pinatubo
volcanic plume, Eos Trans. AGU,80(48), Fall Meet. Suppl.,
Abstract V62C-10.
Guo, S., G. J. S. Bluth, W. I. Rose, I. M. Watson, and A. J.
Prata (2004), Re-evaluation of SO
2
release of the 15 June
1991 Pinatubo eruption using ultraviolet and infrared satel-
lite sensors, Geochem. Geophys. Geosyst.,5, Q04001,
doi:10.1029/2003GC000654.
Herzog, M., H.-F. Graf, C. Textor, and J. M. Oberhuber
(1998), The effect of phase changes of water on the devel-
opment of volcanic plumes, J. Volcanol. Geotherm. Res.,87,
55– 74.
Holasek, R. E., S. Self, and A. W. Woods (1996), Satellite
observations and interpretation of the 1991 Mount Pinatubo
eruption plumes, J. Geophys. Res.,101, 27,635 – 27,655.
Jaeger, H. (1992), The Pinatubo eruption cloud observed by
lidar at Garmisch-Partenkirchen, Geophys. Res. Lett.,19,
191– 194.
Kinne, S., O. B. Toon, and M. J. Prather (1992), Buffering of
stratospheric circulation by changing amounts of tropical
ozone: A Pinatubo case study, Geophys. Res. Lett.,19,
1927– 1930.
Kou, L., D. Labrie, and P. Chylek (1993), Refractive indices of
water and ice in the 0.65 to 2.5 micron range, Appl. Opt.,32,
3531– 3540.
Krotkov, N. A., O. Torres, C. Seftor, A. J. Krueger, W. Rose,
A. Kostinski, G. Bluth, D. Schneider, and S. J. Schaefer
(1999), Comparison of TOMS and AVHRR volcanic ash
retrievals from the August 1992 eruption of Mt. Spurr, Geo-
phys. Res. Lett.,26, 455 – 458.
Long, C. S., and L. L. Stowe (1994), Using the NOAA/
AVHRR to study stratospheric aerosol optical thickness fol-
lowing the Mt. Pinatubo eruption, Geophys. Res. Lett.,21,
2215– 2218.
Mayberry, G. C., W. I. Rose, and G. J. S. Bluth (2002),
Dynamics of volcanic and meteorological clouds produced
on 26 December (Boxing Day) 1997 at Soufrie`re Hills
Volcano, Montserrat, in The eruption of Soufrie`re Hills
Volcano, Montserrat, From 1995 to 1999, edited by T. H.
Druitt and B. P. Kokelaar, Mem. Geol. Soc. London,21,
539– 555.
McCormick, A. P., L. W. Thomason, and C. R. Trepte (1995),
Atmospheric effects of Mt. Pinatubo eruption, Nature,373,
399– 404.
McPeters, R. D., et al. (1998), Nimbus-7 Total Ozone Mapping
Spectrometer (TOMS) Data Product User ’s Guide, NASA
1998-20,6895, Goddard Space Flight Center, Greenbelt, Md.
McPherson, R. D., K. H. Bergman, R. E. Kistler, G. E. Rasch,
and D. S. Gordon (1979), The NMC operational global data
assimilation system, Mon. Weather Rev.,107, 1445 – 1461.
Neal, C. A., R. G. McGimsey, C. A. Gardner, M. L. Harbin,
and C. J. Nye (1994), Tephra-fall from the 1992 eruptions of
Crater Peak, Mount Spurr Volcano, AK: A preliminary re-
port on distribution, stratigraphy and composition, U.S.
Geol. Surv. Bull.,2139.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
33 of 35

Pollack, J. B., O. B. Toon, and B. N. Khare (1973), Optical
properties of some terrestrial rocks and glasses, Icarus,19,
372– 389.
Prata, A. J. (1989a), Observation of volcanic ash clouds in 10 –
12 (m window using AVHRR/2 data, Int. J. Remote Sens.,
10, 751– 761.
Prata, A. J. (1989b), Infrared radiative transfer calculations for
volcanic ash clouds, Geophys. Res. Lett.,16(11), 1293–
1296.
Prata, A. J., W. I. Rose, S. Self, and D. M. O’Brien (2003),
Global, long-term sulphur dioxide measurements from
TOVS data: A new tool for studying explosive volcanism
and climate, in Volcanism and the Earth’s Atmosphere,Geo-
phys. Monogr. Ser., vol. 139, edited by A. Robock and
C. Oppenheimer, pp 75– 92, AGU, Washington, D. C.
Read, W. G., L. Froidevaux, and J. W. Waters (1993), Micro-
wave limb sounder measurement of stratospheric SO
2
from
the Mt. Pinatubo volcano, Geophys. Res. Lett.,20, 1299 –
1302.
Riley, C., W. Rose, and G. Bluth (2003), Quantitative shape
measurements of distal volcanic ash, J. Geophys. Res.,
108(B10), 2504, doi:10.1029/2001JB000818.
Robock, A. (2002), The climate aftermath, Science,295,
1242– 1243.
Rose, W. I., D. J. Delene, D. J. Schneider, G. J. S. Bluth, A. J.
Krueger, I. Sprod, C. McKee, H. L. Davies, and G. G. J.
Ernst (1995), Ice in the 1994 Rabaul eruption cloud: Impli-
cations for volcano hazard and atmospheric effects, Nature,
375, 477– 479.
Rose, W. I., G. J. S. Bluth, and G. G. J. Ernst (2000), Integrat-
ing retrievals of volcanic cloud characteristics from satellite
remote sensors: A summary, Philos. Trans. R. Soc. London,
Ser. A,358, 1585 – 1606.
Rose, W. I., G. J. S. Bluth, D. J. Schneider, G. G. J. Ernst,
C. M. Riley, L. J. Henderson, and R. G. McGimsey (2001),
Observation of volcanic clouds in their first few days of
atmospheric residence: The 1992 eruptions of Crater Peak,
Mount Spurr volcano, Alaska, J. Geol.,109, 677 – 694.
Rose, W. I., et al. (2003), The February-March 2000 eruption
of Hekla, Iceland from a satellite perspective, in Volcanism
and the Earth’s Atmosphere,Geophys. Monogr. Ser.,vol.
139, edited by A. Robock and C. Oppenheimer, pp. 107–
132, AGU, Washington, D. C.
Russell, P. B., et al. (1996), Global to microscale evolution of
the Pinatubo volcanic aerosol derived from diverse measure-
ments and analyses, J. Geophys. Res.,101(D13), 18,745 –
18,763.
Rutherford, M. J., and J. D. Devine (1996), Preeruption pres-
sure-temperature conditions and volatiles in the 1991 Dacitic
Magma of Mount Pinatubo, in Fire and Mud: Eruptions and
Lahars of Mount Pinatubo,Philippines,editedbyC.G.
Newhall and R. S. Punongbayan, pp. 751– 766, Philippine
Inst. of Volcanol. and Seismol., Quezon City.
Schneider, D. J., W. I. Rose, and L. Kelley (1995), Tracking of
1992 eruption clouds from Crater Peak vent of Mount Spurr
volcano, Alaska, using AVHRR, in The 1992 Eruptions of
Crater Peak Vent, Mount Spurr Volcano, Alaska, edited by
T. E. C. Keith, U.S. Geol. Surv. Bull.,2139, 27 – 36.
Schneider,D.J.,W.I.Rose,L.R.Coke,G.J.S.Bluth,
I. Sprod, and A. J. Krueger (1999), The first four days of
volcano/atmosphere interaction following a stratospheric
eruption as observed with TOMS and AVHRR, J. Geophys.
Res.,104, 1037 – 1050.
Schoeberl, M. R., L. R. Lait, P. A. Newman, and J. E. Rosenfield
(1992), The structure of the polar vortex, J. Geophys. Res.,97,
7859– 7882.
Schoeberl, M. R., S. D. Doiron, L. R. Lait, P. A. Newman, and
A. J. Krueger (1993), A simulation of the Cerro Hudson SO
2
cloud, J. Geophys. Res.,98, 2949 – 2955.
Scott, W. E., R. P. Hoblitt, R. C. Torres, S. Self, M. M. L.
Martinez, and T. Nillos (1996), Pyroclastic flows of the June
15, 1991 climactic eruption of Mount, Pinatubo, in Fire and
Mud: Eruptions and Lahars of Mount Pinatubo, Philippines,
edited by C. G. Newhall and R. S. Punongbayan, pp. 545–
570, Philippine Inst. of Volcanol. and Seismol, Quezon
City.
Seftor, C. J., N. C. Hsu, J. R. Herman, P. K. Bhartia, O. Torres,
W. I. Rose, D. J. Schneider, and N. Krotkov (1997), Detec-
tion of volcanic ash clouds from Nimbus 7/total ozone map-
ping spectrometer, J. Geophys. Res.,102, 16,749 – 16,759.
Self, S., J.-X. Zhao, R. E. Holasek, R. C. Torres, and A. J.
King (1996), The atmospheric impact of the 1991 Mount
Pinatubo eruption, in Fire and Mud: Eruptions and Lahars
of Mount Pinatubo, Philippines, edited by C. G. Newhall and
R. S. Punongbayan, pp. 1089 – 1115, Philippine Inst. of Vol-
canol. and Seismol., Quezon City.
Shannon, J. (1997), Three dimensional evolution of volcanic
gas and ash clouds from Crater Peak, Mt. Spurr, M.S. thesis,
Mich. Technol. Univ., Houghton.
Solomon, S., R. W. Sanders, R. R. Garcia, and J. G. Keys
(1993), Increased chlorine dioxide over Antarctica caused
by volcanic aerosols from Mount Pinatubo, Nature,363,
245– 248.
Sparks,R.S.J.,M.I.Bursik,S.N.Carey,J.S.Gilbert,
L. Glaze, H. Sigurdsson, and A. W. Woods (1997), Volcanic
Plumes, John Wiley, New York.
Textor, C. (1999), Numerical simulation of scavenging pro-
cesses in explosive volcanic eruption clouds, Ph.D. thesis,
Max-Planck Inst. for Meteorol, Hamburg, Germany.
Turco, R. P., O. B. Toon, R. C. Whitten, P. Hamill, and R. G.
Keesee (1983), The 1980 eruption of Mount St. Helens:
Physical and chemical processes in the stratospheric clouds,
J. Geophys. Res.,88, 5299– 5319.
Wallace, J. M., and P. V. Hobbs (1997), Atmospheric Science:
An Introductory Survey, Academic, San Diego, Calif.
Watson, I. M., W. I. Rose, V. J. Realmuto, and G. J. S. Bluth
(2003), Forward modeling of volcanic aerosols transmissions
at different latitudes: Quantifying the effects of varying tro-
posheric water vapor on ash detection, Eos Trans. AGU,
84(46), Fall Meet. Suppl., Abstract V12E-05.
Wen, S., and W. I. Rose (1994), Retrieval of particle sizes and
masses in volcanic clouds using AVHRR band 4 and band 5,
J. Geophys. Res.,99, 5421– 5431.
Wiesner, M. G., A. Wetzel, S. G. Catane, E. L. Listanco, and
H. T. Mirabueno (2004), Grain size, areal thickness distribu-
tion and controls on sedimentation of the 1991 Mount Pina-
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
34 of 35

tubo tephra layer in the South China Sea, Bull. Volcanol.,66,
226– 242, doi:10.1007/s00445-003-0306-x.
Wolfe, E. W., and R. P. Hoblitt (1996), Overview of the erup-
tions, in Fire and Mud: Eruptions and Lahars of Mount
Pinatubo, Philippines, edited by C. G. Newhall and R. S.
Punongbayan, pp. 415– 433, Philippine Inst. of Volcanol.
and Seismol., Quezon City.
Yamanouchi, T., K. Suzuki, and S. Kawaguchi (1987), Detec-
tion of clouds in Antarctica from infrared multispectral data
of AVHRR, J. Meteorol. Soc.,65, 949 – 961.
Yu, T., and W. I. Rose (2000), Retrieval of sulfate and silicate
ash masses in young (1 to 4 days) eruption clouds using
multiband infrared HIRS/2 data, in Remote Sensing of Active
Volcanism,Geophys. Monogr. Ser., vol. 116, edited by
P. Mouginis-Mark, J. Crisp, and J. Fink, pp. 87 – 100, AGU,
Washington, D. C.
Yu, T., W. Rose, and A. Prata (2002), Atmospheric correction
for satellite-based volcanic ash mapping and retrievals using
‘‘split window’’ IR data from GOES and AVHRR, J. Geo-
phys. Res.,107(D16), 4311, doi:10.1029/2001JD000706.
Geochemistry
Geophysics
Geosystems
G3
G3
guo et al.: june 1991 pinatubo volcanic clouds 10.1029/2003GC000655
35 of 35