Lorenzo Guerrieri

8/10/2019 Lorenzo Guerrieri

1/20

Evolution of the 2011 Mt. Etna ash and SO 2 lava fountain episodes using

SEVIRI data and VPR retrieval approach

Lorenzo Guerrieri, Luca Merucci, Stefano Corradini, Sergio Pugnaghi

PII: S0377-0273(14)00397-7

DOI: doi:10.1016/j.jvolgeores.2014.12.016

Reference: VOLGEO 5470

To appear in: Journal of Volcanology and Geothermal Research

Received date: 29 September 2014

Accepted date: 23 December 2014

Please cite this article as: Guerrieri, Lorenzo, Merucci, Luca, Corradini, Stefano, Pug-naghi, Sergio, Evolution of the 2011 Mt. Etna ash and SO2 lava fountain episodes usingSEVIRI data and VPR retrieval approach,Journal of Volcanology and Geothermal Research(2015), doi:10.1016/j.jvolgeores.2014.12.016

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.
http://dx.doi.org/10.1016/j.jvolgeores.2014.12.016http://dx.doi.org/10.1016/j.jvolgeores.2014.12.016http://dx.doi.org/10.1016/j.jvolgeores.2014.12.016http://dx.doi.org/10.1016/j.jvolgeores.2014.12.016


2/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

Evolution of the 2011 Mt. Etna ash and SO2lava fountain

episodes using SEVIRI data and VPR retrieval approach

Lorenzo Guerrieri, Luca Merucci, Stefano Corradini, Sergio Pugnaghi

AbstractIn this paper an estimation is made of the temporal evolution of volcanic ash and sulfur dioxide (SO2)

emissions from Mt. Etna during its eruption phases. The retrieval is performed using MSG-SEVIRI (Meteosat

Second Generation - Spinning Enhanced Visible and Infra Red Imager) images in the TIR (Thermal InfraRed)

spectral range. The ash and SO2 plume abundance maps are computed using the Volcanic Plume Removal

(VPR) procedure originally applied to MODIS (Moderate Resolution Imaging Spectroradiometer) sensors on

board the NASA Terra and Aqua satellites. As test cases, two 2011 lava fountain episodes were considered. The

set of parameters required by VPR for the Mt. Etna volcano, Volz type particles, and the SEVIRI sensor are

presented. Once the parameters have been computed, the VPR approach requires as input only the SEVIRI-TIR

radiances of the bands centered at 8.7, 10.8, and 12.0 m, together with the plume temperature and altitude. TheVPR returns maps of plume particles effective radius, aerosol optical depth at 550 nm, and columnar abundance

of ash and SO2. A new procedure for estimating wind speed and direction is also presented. Since the ash and

SO2abundance maps, and wind speed at the plume altitude are known, it is possible to reconstruct the ash and

SO2fluxes emitted during the eruption through time. The VPR procedure, applied to TIR SEVIRI data, allows

fast and reliable ash and SO2 retrieval with high temporal resolution during both day and night, and is thus

suitable for operational use during a volcanic crisis.

KeywordsRemote Sensing, Thermal-infrared radiation (TIR), volcanic ash and SO2 retrieval, multispectral satellite

data.

L. Guerrieri (Corresponding Author), S. Pugnaghi

Dipartimento di Scienze Chimiche e Geologiche

Universit di Modena e Reggio Emilia

c/o DIEF (ex DIMA)

Via Vignolese 905a, I-41125 Modena (Italy)

Tel./Ph. +39 059 205 6217

e-mail: [email protected]

L. Merucci, S. Corradini

Istituto Nazionale di Geofisica e Vulcanologia (INGV)

Via di Vigna Murata 605, Roma, Italy.

1 Introduction

Volcanic plumes ejected into the atmosphere during explosive eruptions interact with human activities on various

levels. Plumes are generally composed of a mixture of ash and gases (Sparks et al. 1997; Oppenheimer et al. 2011), and

represent a major threat that can be mitigated only by early and rapid geolocation (detection), quantitative

characterization, tracking and forecasting of trajectories (Rose et al. 2009). On a local scale, the release of volcanic ash

and gases can affect human and animal health, while intense ash fall-out can severely damage infrastructures and

disrupt ground traffic (Delmelle et al. 2002; Mather et al. 2003; Horwell et al. 2006).

On a regional and global scale, buoyant volcanic ash clouds transported by the wind are a major hazard for air traffic

(Casadevall 1994; Miller et al. 2000; Hufford et al. 2000) and can have an important impact on climate (Robock 2000;

Langmann et al. 2010; Solomon et al. 2011; Bourassa et al. 2012). Observations of passive volcanic degassing canprovide useful information for forecasting impending eruptions (Sparks 2003; Duffell et al. 2003; Aiuppa et al. 2007),

while flux measurements of ash and sulfur dioxide (SO2) emitted during explosive events can give insights into the


3/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

processes driving the volcanic activity and provide improved source terms to ash deposition and transport models

(Merucci et al. 2011; Theys et al. 2013; Boichu et al. 2013).

The wide range of effects related to volcanic emissions can often be conveniently monitored and investigated using the

data provided by Earth Observation satellites (Prata 2009; Thomas et al. 2010). Thermal infrared (TIR) multispectral

imaging radiometers, on low or geostationary Earth orbit platforms (LEO and GEO), allow the detection and retrieval of

ash and SO2 burdens in volcanic clouds if provided with appropriate channels. This is the case with the MODIS

(Moderate Resolution Imaging Spectroradiometer) sensors on the polar LEO satellites TERRA and AQUA, and the

SEVIRI (Spinning Enhanced Visible and Infrared Imager) sensors on the MSG geostationary platforms.

A number of volcanic ash and SO2detection and retrieval algorithms have been developed and applied successfully to

MODIS and SEVIRI data and these results represent an established reference for novel approaches (Prata 1989; Wen et

al. 1994; Realmuto et al. 1994; Pugnaghi et al. 2002; Corradini et al. 2009). Recently the authors presented a new

retrieval scheme for ash and SO2retrieval called volcanic plume removal (VPR)(Pugnaghi et al. 2013), specifically

designed to obtain fast, user-friendly, but also reliable retrieval of volcanic ash and SO2plume parameters from MODIS

TIR data. The proposed procedure retrieves some of the most important parameters of volcanic plumes containing ash

and SO2, these being the SO2 columnar abundance maps, ash mass, aerosol optical depth (AOD550), and effective

radius () maps, with plume altitude and temperature as the only additional inputs required at run time. Similar resultscan be obtained from the same datasets with ash and SO2 retrieval schemes based on look-up tables (LUT) of the

corrections terms computed by simulating the atmosphere containing ash and SO2plumes with radiative transfer models

(RTM). Generally these methods, albeit theoretically sound and reliable, are difficult to implement because they require

independent knowledge of the atmosphere at the time of data acquisition in the target area (i.e. pressure, temperature,

and humidity vertical profiles collected by radiosondes or from meteorological models), specific skills on radiative

transfer and RTM codes, and they also need extensive computation time.

Since quick and reliable detection and characterization of volcanic ash clouds potentially hazardous to air traffic is

extremely valuable to support decisions during volcanic crises, new retrieval schemes have recently been proposed

(Picchiani et al. 2011; Pugnaghi et al. 2013; Piscini et al. 2014).

The VPR method is easy to use and useful for application in the monitoring of active volcanoes because with only

minor training it can be implemented by the staff on surveillance shifts, while still giving fast and reliable results. The

most time consuming computations, and the parts in which specific expertise is needed, are anticipated in a preliminary

phase during which the procedure is set up for a given location, ash type, and sensor.In this work the VPR approach is applied to SEVIRI data collected during two brief lava fountain events that occurred

on Mt. Etna in 2011.

In Section 2 the VPR procedure is first described in its key features, namely the plume transmittance model, the ash andAOD550, and the SO2columnar content estimations. Then the high data frequency of SEVIRI (one image every 15

minutes, up to 5 minutes in rapid scan mode, available over a wide area of the Northern Hemisphere) is exploited to

define a method to retrieve wind speed and direction at the plume altitude, and to reconstruct the flux at the volcano

vents. The parameters required to run the procedure for the Volz (1973) ash type and the three SEVIRI instruments

currently on board the MSG platforms (Meteosat-8, -9 and -10) are computed and tabulated.

In Section 3 two Mt. Etna eruption test cases are described and, following the same scheme described in Pugnaghi et al.

(2013), the VPR results are closely compared with the results obtained by applying the ash and SO2retrieval schemes

(Corradini et al. 2009) based on LUTs computed by MODTRAN RTM code (Berk et al. 1989) to the same datasets. Aclassification of the cloud ash particles based on the retrieved effective radius is also presented.

Finally, wind speed and direction are estimated at the plume height from a sequence of images, and a reconstruction of

the fluxes emitted at the source is presented and discussed.Finally, the conclusions are drawn in Section 4.

2 VPR procedure

In this section the VPR procedure applied to SEVIRI data is described. Once the ash and SO2columnar abundance

maps have been estimated, the emitted fluxes can be computed since the wind speed at the plume altitude can be

established using the new flux reconstruction procedure. The required inputs of wind speed and direction at the plume

altitude can easily be derived from the sequence of SEVIRI images itself, if the plume structure allows the massmaximum position (ash or SO2) to be located and tracked in a sequence of images.


4/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

The entire procedure has five steps: the first three regard the VPR, and the last two regard flux reconstruction. The first

step computes the transmittance maps of the 8.7, 10.8, and 12.0 m bands; the second step computes the ash particle

effective radius () and the aerosol optical depth at 550 nm (AOD550) maps from the transmittance maps of the 10.8and 12.0 m bands; the third step computes the SO2abundance map from the transmittance maps of the 8.7 m band;

the fourth step computes wind speed at the plume altitude by detecting the position of the maximum mass (ash or SO2)

in the plume map; the fifth step extrapolates the flux of ash and SO2at the vents for each image considered.

2.1 VPR: plume transmittance

The VPR procedure (Pugnaghi et al. 2013) starts with the computation of a new image by replacing the radiance

values in the plume region with those obtained from a simple linear regression of the radiance outside the edges of the

plume (see Section 3 for a description of the technique used to retrieve a suitable plume mask). Thus the plume is

removed and the new virtual image created by the VPR procedure shows what the sensor on the satellite would have

seen if the plume was not present.

The two images, the original and the one generated without the plume, allow an initial estimation of plume

transmittance for the TIR bands centered at 8.7, 10.8, 12.0 m:

(1)

where is the radiance measured by the sensor, is the radiance obtained by removing the plume (the virtualradiance computed in the first VPR step), is the air mass factor which accounts for the path slant throughthe plume, is the sensor zenith angle, is the Planck function at the temperature Twhich is slightly modifiedrelative to the temperature of the atmosphere at the mean plume height, and is a rough empirical estimation of the

plume vertical transmittance due to scattering, which is specific for each aerosol particle type. For the Mt. Etna volcano,

with Volz (1973) type particles and MODIS or SEVIRI but, if , is recomputed with . The modified temperature T accounts for the layer above the plume, which is ignored in the VPR plume

model. For all three SEVIRI instruments , when is the air temperature at the mean plumealtitude in km. and are the input data for the VPR procedure. The final plume transmittance for each of thethree bands considered, is computed with a simple polynomial relationship from the previously described. The

polynomial relationship takes into account the approximations performed in the simple underlying plume model used in

Eq. (1).

(2)

The parameters for the Mt. Etna volcano and Volz type particles have been computed by fitting the MODTRANsimulated plume transmittances versus the transmittances obtained with Eq. (1), and are reported in Table 1 for each

band and each SEVIRI sensor. They derive from 76032 scenarios considered in the radiative transfer code simulations,including: 12 month profiles (the monthly mean profiles of atmospheric pressure, air temperature and relative humidity

measured at Trapani over the last 30 years; Trapani, on the western tip of Sicily, is the WMO upper-air station closest to

the Mt. Etna volcano); 4 plume altitudes ( km), with a constant thickness of 1 km; 6 aerosol opticaldepthsAOD550(0, 0.078, 0.156, 0.313, 0.625, 1.250); 8 effective radii (0.785, 1.129, 1.624, 2.336, 3.360, 4.833,6.952, 10.000 m); 3 sensor zenith angles (39.6, 44.0, 48.9 deg); 11 SO2columnar abundances (0-10 g m

-2in steps of 1

g m-2

). The radiances and were computed assuming a constant sea surface emissivity (0.98) and a monthly meansea temperature of the area of interest from NOAA (see Pugnaghi et al. 2013).

Finally, if the plume is very transparent (if ), then the total transmittance in the 8.7 m band isrecomputed using ,ignoring any ash effects in that wavelength.

Satellite Band MSG-1 2 (8.7 m) -0.053 0.663 0.735 -0.371 0.995

4 (10.8 m) -0.056 0.670 0.473 -0.098 0.999


5/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

5 (12.0 m) -0.061 0.610 0.535 -0.098 0.998

MSG-2 2 (8.7 m) -0.053 0.664 0.735 -0.371 0.995

4 (10.8 m) -0.056 0.670 0.474 -0.098 0.999

5 (12.0 m) -0.060 0.599 0.550 -0.103 0.998

MSG-3 2 (8.7 m) -0.052 0.660 0.741 -0.374 0.995

4 (10.8 m) -0.056 0.669 0.476 -0.099 0.999

5 (12.0 m) -0.061 0.606 0.541 -0.100 0.998

Table 1Cubic polynomial relationship coefficients of Eq. (2) computed for the three 8.7, 10.8 and 12.0 m bands for

all the current SEVIRI aboard MSG-1, -2, -3. is the correlation coefficient

2.2 VPR: effective radius and AOD550

The total plume transmittance of the window bands at 10.8 and 12.0 m is due only to ash extinction and can be

written as:

(3)

where is the air mass factor and AOD550is the aerosol optical depth at 550 nm, while is a function of the effectiveradius () of the particle considered, representing the slope of the linear relationship which stands between the AOD atthe wavelengths considered and the AOD at 550 nm (see fig. 4 in Pugnaghi et al. 2013).

The ratio of the logarithms of the plume transmittances of the bands at 10.8 and 12.0 m is:

(4)

The ratio expressed in Eq. (4), in the particle size range (0.8-10 m) considered in the VPR scheme, is a monotonicfunction of the effective radius ,which consequently can easily be estimated. Next, the slope can be obtainedfrom , and finallyAOD550is calculated from Eq. (3).The slope can also be retrieved in the same way, when with AOD550known, the ash transmittance at 8.7 mis obtained using Eq. (3); this is necessary for SO2retrieval (see next section 2.3).The , and values for each SEVIRI sensor, are reported in Table 2. These valuesfor the Mt. Etna volcano and the optical properties of the Volz (1973) type particles were computed by fitting the

MODTRAN plume transmittances obtained with the above quoted simulations.

Finally, when the particles effective radius and the aerosol optical depth AOD550 of each pixel have beenestimated, the ash mass can be retrieved using the Wen and Rose (1994) simplified formula.

0.785 1.129 1.624 2.336 3.360 4.833 6.952 10.000

MSG-1 1.604 1.575 1.466 1.317 1.165 1.047 0.984 0.968

0.150 0.330 0.623 0.966 1.235 1.348 1.333 1.272

0.158 0.339 0.615 0.934 1.203 1.329 1.314 1.250

MSG-2 1.639 1.609 1.494 1.337 1.176 1.051 0.984 0.967

0.151 0.331 0.623 0.966 1.236 1.348 1.333 1.272

0.158 0.340 0.618 0.938 1.207 1.330 1.314 1.250

MSG-3 1.605 1.577 1.469 1.320 1.167 1.048 0.984 0.968

0.150 0.329 0.621 0.964 1.234 1.348 1.333 1.273


6/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

0.159 0.342 0.622 0.943 1.211 1.331 1.314 1.249

Table 2values as a function of , for all the current SEVIRI aboard MSG-1, -2, -3.

2.3 VPR: SO2columnar content

The total plume transmittance for band 8.7 m estimated using Eq. (1) is the product of the SO 2absorption and ash

extinction. Therefore, to establish the contribution only from sulfur dioxide absorption, it is necessary to estimate the

ash transmittance at 8.7 m and divide the total transmittance of the plume by this value:

(5)

The SO2columnar content can be computed inverting the following relationship:

(6)

where is always the air mass factor, is the SO2absorption coefficient which depends on the plume temperature andpressure (height) and is an exponent of the concentration which depends only on the plume temperature:

(7)

(8)

The values of and parameters were empirically derived from the MODTRAN simulations and are reported in Table3 and Table 4 respectively.

Parameter MSG-1 MSG-2 MSG-3

-5.534e-5 -6.097e-5 -5.629e-5

-1.327e-4 -1.327e-4 -1.323e-4

4.099e-2 4.101e-2 4.079e-2

Table 3Parameters to compute the SO2absorption coefficient depending on the plume temperature andpressure (height). See Eq. (7).

Parameter MSG-1 MSG-2 MSG-3

3.522e-4 3.527e-4 3.530e-4 9.804e-1 9.804e-1 9.802e-1

Table 4Parameters to compute the SO2concentration exponent depending on the plume temperature. See Eq.(8).

Fig. 1 shows the distributions of the differences between the procedure results (only for the SEVIRI aboard MSG-2) and

the input values used in the MODTRAN simulations. More than 80% of the AOD550cases exhibit a difference less

than 0.125 (a), about 70% of the differences are less than 0.5 m (b), and more than 60% of the differences areless than 0.5 g m

2(c).


7/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

(a) (b) (c)

Fig. 1Frequency (%) of the differences VPR-MODTRAN (SEVIRI-MSG2) forAOD550(a), (b) and cs(c).

2.4 Wind speed at the plume altitude

The eruptions of the Mt. Etna volcano selected as test cases for the proposed procedure exhibit a fairly clear peak in

the spatial mass distribution maps of both SO2and ash. This peak has frequently been observed in recent short duration

Etna events characterized as lava fountains, and can be easily identified and tracked in a series of images. The first part

of the flux reconstruction procedure detects the peak (or peaks, since the SO 2 and ash peaks are not necessarily

collocated) in the abundance maps computed with the VPR procedure for each SEVIRI image, and then tracks the

subsequent positions in a sequence of images. From the position of the peak, both its distance from the top of the

volcano (craters area) and its angle relative to North are derived. If the wind direction is constant, the peak speed can be

estimated considering the 15 minute time interval between two SEVIRI image acquisitions (reduced to 5 minutes in

Rapid Scan mode). The peak speed is assumed to be the whole plume speed, which is the true wind speed at plume

altitude. Obviously this method may not be appropriate for very long plumes, e.g. from a long continuous eruption. Inthese cases, different parts of the plume may be travelling at different speeds. After a smoothing operation performed

with a mean filter over a 3 by 3 pixel area, the position of the peak in the abundance maps is assumed to be the center of

mass of a chosen square area (about 20 km side length) around the maximum in the map. This is slightly better (less

sensitive to fluctuations) than choosing the raw position of the maximum. The peak speed at a specific time is obtained

by means of linear regression of the peak distances from the top of Mt Etna in the three hours around the image

considered: 1.5 hours before and 1.5 hours after the time of the image. As the SO 2and ash peaks are usually not formed

simultaneously, the computed speed of these peaks may be quite different, in particular during the peak formation

period. Clearly, the images from the beginning and end of the eruption have fewer data before or after them.

This method for estimating wind speed at the plume altitude has the advantage that the plume speed is estimated

directly from the images, and so is not sensitive to uncertainty in the plume altitude.

2.5 Flux Reconstruction at the volcano vents

The wind speed estimated as described in section 2.4, is subsequently used to compute the temporal trend of the flux

emitted by the volcano for each image of the SEVIRI sequence (see below). If the wind speed cannot be reliably

retrieved as previously described, then an input wind speed is used, for example the one obtained from upper air

measurement. In particular, an input wind speed is used rather than the computed one if: 1) there are limited data for the

time interval considered: at least three peak positions are required in the three hour period with a time difference of at

least one and a half hours; 2) the percentage standard deviation about the linear regression is too wide: only the data

with an absolute percentage difference (data fit) lower than 10% are used for the final linear regression; 3) the wind

direction changes excessively: more than 20 degrees in the three hours considered.

This part of the procedure determines, for each image considered, the flux of both SO 2and ash emitted from the

vents. For each pixel of the plume the distance from the top of the volcano (known position) is determined. Then the

average wind speeds () in the time intervals preceding the image acquisition time are computed considering the


8/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

acquisition time and the previously estimated wind speed evolution. The average wind speeds and the distance () ofeach pixel from the volcano vent, allow computation of the plume age at each pixel, i.e. the time at which the SO2orash contained in any given pixel of the plume was emitted.

Finally, integrating the retrieved mass over the pixels at the same distance, that is along an arc of length land ray centered at the volcano vent, and knowing the wind speed at the flux of the SO2or ash emitted by the volcano

through a cylindrical surface (ray ) and orthogonal to the image plane is obtained as:

(9)

where is the wind speed at the time when the mass (SO2or ash) per unit surface of the pixel in the position has been emitted from the vents.

From the recent typical Etna lava fountains in the period 2011-2013, the temporal trend of the flux of both SO 2and ash

emitted from the vents is expected to be a form of normal distribution curve describing the explosive event. It is worth

noting that all the flux curves reconstructed from these eruption sequences of SEVIRI images have to overlap each

other.

3 Results and Applications

The VPR procedure described so far was applied to the TIR SEVIRI data of two different Mt. Etna eruptions that

occurred on 12 August 2011 and 23 October 2011. These test cases were selected from among several Etna lava

fountain episodes occurred during 2011. Both events lasted a few hours and the eruption plumes reached an altitude of

about 5-7 km and were clearly recorded by the SEVIRI sensor for several hours. A fixed SEVIRI window containing

the whole of the island of Sicily and a wide region of the Mediterranean Sea south-east of Mt. Etna was chosen in this

work and the re-sampled pixels have an area of about 14 km2. All the times shown below are in hours UTC. Finally, for

each SEVIRI image, a so-called plume mask was created to identify the pixels contaminated by the presence of the

volcanic plume. The choice of the plume mask is very important for the correct application of the VPR procedure, and

so plume detection was conducted in a semi-automatic way, with the supervision of an operator. The method is basedon a percentage threshold of the mean radiance value without the plume for a defined sequence of bands. Finally, the

operator can manually refine the plume mask by deleting some possibly spurious pixels. The SO2 and ash retrievals

were performed only in the pixels of the plume mask.

The first case considered (12 August) was a day-time event, while the second case (23 October) occurred at night. On

the 12 August the brightness temperature difference (BTD) between bands at 10.8 and 12.0 m indicates the presence

of water droplets and/or ice particles in addition to ash. The same test, performed on the 23 October images, reveals

only the presence of ash particles in the plume. Finally, both the cases clearly show the presence of sulfur dioxide in the

two volcanic clouds.

As regards the VPR-MODIS retrieval (Pugnaghi et al. 2013), the VPR-SEVIRI retrievals were also compared with the

results obtained using the well established LUT approach (Corradini et al., 2009). In the LUT procedure, ash retrieval

(AOD550, and ash mass) is based on the computation of look up tables simulated with a radiative transfer model and

relies on the BTD (Prata 1989; Wen et al. 1994; Yu et al. 2002; Corradini et al. 2008), while SO 2estimation is realized

using a chi square procedure applied to the SEVIRI channel centered at 8.7 m. The LUT required for the ash and SO 2

retrievals were computed using MODTRAN RTM. Differently from the VPR approach, that computes the coefficients

by using the monthly mean atmospheric profiles (P, T, RH) measured at Trapani WMO over the last 30 years, the LUT

procedure needs a single atmospheric Trapani radiosounding (P, T, RH) for a specific date and time, as close as possible

to the satellite image acquisition. This difference between the two methods is relevant: while VPR, using eq. (1) and (2),

retrieves plume transmittances that are weakly dependent on atmospheric conditions, LUT works directly with the TOA

radiances that are strongly affected by atmospheric parameters. As ash characteristics, the Volz (1973) ash optical

properties were used as RTM inputs both for LUT and VPR. Only for LUT, the sea surface temperature was computed

from the inversion of the radiative transfer equation, considering a spectral emissivity estimated from the convolution

between the JPL seawater TIR emissivity (http://speclib.jpl.nasa.gov/) and the SEVIRI response functions. The

influence of ash on the SEVIRI 8.7 m band, during SO2retrieval, was corrected following the procedure described by

(Corradini et al. 2009).


9/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

The comparison of the results of the LUT- and VPR-based procedures presented here mainly concerns the trends for the

total mass of the SO2and ash particles. Only the plume region containing ash particles and SO 2was considered because

the VPR procedure is not yet capable of analyzing plumes with water/ice droplets. On October 23 the entire volcanic

cloud was analyzed, while on August 12 ash particles were present only in about 50 % of the volcanic cloud and the

remaining part of the plume was excluded from the comparison.

As mentioned above the ash mass is retrieved using the Wen and Rose (1994) simplified formula, which is based on the

mean AOD550 and effective radius of the particles of the analyzed pixel. The retrieved effective radii enable

classification of the plume particles into fine and coarse particles. All particles with an effective radius of less than or

equal to 5 m were classed as fine particles, while the radii greater than 5 m were classed as coarse particles. The

temporal trend obtained from the VPR and LUT procedures for these two classes is described below.

Finally, an initial test of the VPR procedure with flux reconstruction was carried out for August 12, in which the peak

positions of aerosols and sulfur dioxide are well detected. The instantaneous wind speed and direction were computed

from the consecutive peak positions, and then the SO2 and ash fluxes emitted from the volcanic vents were

reconstructed as functions of time.

3.1 VPR-LUT comparison: test case 12 August 2011

On August 12, at about 7:00 UTC, the tenth Etna lava fountain episode of 2011 began (INGV 2011a). The

Strombolian explosive activity of the South-East crater increased rapidly and after about 8:30 the fountain phase was

quite intense. The lava fountain reached a height of several hundred meters above the vent, while the associated ash

column reached an altitude of about 7 km (a.s.l.) and produced a consistent fall out of ash debris under the plume, in

particular over the village of Zafferana Etnea on the eastern flank of the volcano. The lava fountain activity terminated

between 10:30 and 11:00 with residual pulses of ash emission continuing for about other 30 minutes.

The SEVIRI data shows that the volcanic plume reached the seashore at about 10:00 UTC, when the fountain activity

above the craters was at a maximum. On August 12, there is a delay of about an hour between the time of emission at

the craters and the detection of the plume over the sea.

The plume top altitude was established directly from the SEVIRI images by looking for the coldest pixels in the plume

in the band centered at 10.8 m. The brightness temperature of the coldest pixel was compared with the temperaturevertical profile measured at midday at Trapani (western tip of Sicily), which is the closest WMO upper air station

(number 16429). Images from the video-surveillance system of the Mt. Etna volcano, recorded during the eruptive event

of August 12, were analyzed to assess the highest altitude reached by the plume column during the maximum of the

paroxysmal phase. The comparison between the maximum plume top altitude evaluated from the SEVIRI data of about

7 km, and the slightly lower 6.5 km assessed by means of the INGV cameras deployed on the flanks of Etna shows

close agreement, within the associated 500 m measurement error (Scollo et al. 2014).

It is known (Prata 1989) that the BTD is generally negative in the presence of ash particles and positive in the presence

of water droplets or ice crystals. In this work, as in previous papers regarding Mt. Etna emission retrieval (Corradini et

al. 2009), a detection threshold of 0.17 K was applied. Consequently, all the pixels with a BTD 0.17 K were

considered to be mostly contaminated by ash. On the basis of this test, the first part of the volcanic emission (the part

furthest from the vents) is mainly formed by liquid water droplets (or more probably ice crystals) and SO2, while the

subsequent emission, after about 10:00 UTC, comprises mainly ash and SO2.

This paper aims to compare the VPR and LUT retrieval of only ash and SO 2, rather than water or ice and SO2, so only

the part of the plume with ash and SO2was analyzed. For example, Fig. 2a shows the BTD for the image recorded at

11:12 and Fig. 2b shows the corresponding plume mask chosen for the ash and SO 2retrievals. In Fig. 2a the volcanic

cloud is clearly identifiable: BTD negative values (black plume) identify mainly ash, while positive values (white

plume) identify mainly water/ice droplets. It is important to note that this decision results in Mt. Etnas SO2 emissions

being significantly underestimated, and the values presented here should not be considered as the total SO2emission for

the August, 12 event.

The VPR procedure is designed for rapid use during early volcanic crises, when only remotely sensed images and a few

other data are available, and a quick quantitative assessment of the presence of ash in a volcanic cloud can offer

improved aviation security. However, future developments of the VPR will include assessment of the mass of liquid

water/ice droplets with or without sulfur dioxide.


10/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

(a) (b)

Fig. 2BTD computed from the 12 August 2011 SEVIRI image, recorded at 11:12 UTC (a); ash and SO2volcanic plumemask (b).

Fig. 3 (top image) shows a comparison of the SO 2and ash total mass, this being the accumulated mass of SO 2/ash over

all the pixels in a map, computed using the LUT and VPR procedures. Retrieval was achieved by analyzing 19 images

collected by the SEVIRI radiometer aboard the MSG-2 satellite, from 10:12 UTC to 14:42 UTC on August 12 2011.

The agreement between the two procedures is quite clear, even though for both SO 2 and ash the total mass curve

computed with the VPR is a little lower than the LUT plot.


11/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

Fig. 3 SO2 and ash total mass trends (top) and percentage of presence in the volcanic cloud of the two classes of

particles (fine with and coarse with ) considered (bottom) computed using the LUT andVPR procedures. The label Image time refers to the image acquisition time.

Initially the mass trend shows an emission of about 1200 tons of ash and about 1600 tons of SO2in one hour, which

means more than 300 kg of ash and more than 400 kg of SO 2per second. The maximum of the peak is at about 11:30;

this was the end of the event but because of the aforementioned delay relative to emission time, the end of the event


12/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

happened at least 30-60 minutes before, as indicated in the INGV bulletin (INGV 2011a). In less than three hours both

the plumes of ash and SO2 had disappeared, or at least the SEVIRI radiometer was no longer able to detect their

presence over the sea. The fall out and/or dispersion was very fast.

Fig. 3 (bottom image) shows the trends of the percentage of presence in the volcanic cloud of the two classes of

particles considered: fine ( ) and coarse ( ). There is an obvious rapid decay throughtime of the coarse particles and a corresponding increase in fine particles. The VPR-LUT comparison reveals a

reasonably close agreement.

3.2 VPR-LUT comparison: test case 23 October 2011

The seventeenth lava fountain episode of 2011 took place on the 23 October. (INGV 2011b). The explosive activity

started at 17:13 from the so-called new cone on the eastern flank of the South-East crater of Mt. Etna. It continued

and increased in intensity till 18:26 when a new eruptive vent (18:36) open on the flank of the cone giving added

strength to the lava fountains. The fountains reached 300 m above the crater after 19:00 UTC. After 20:30 both the

effusive and explosive activity reduced drastically. At 21:00 there was some Strombolian activity that completely

disappeared after 21:15.

SEVIRI only detected the first track of the volcanic cloud over the sea starting from 19:00 UTC when the fountainactivity at the crater was at its maximum, due to the known delay between time of emission from the crater and image

time. In the SEVIRI-TIR images the plume is clearly visible until midnight.

(a) (b)

Fig. 4BTD computed from the 23 October 2011 SEVIRI image recorded at 21:27 UTC (a); ash and SO 2 volcanic

plume mask (b).

A plume top altitude of 6 km a.s.l. was estimated using the coldest pixel technique, but no comparison with a ground-

based camera estimation was possible because the event occurred during the night. On the basis of the BTD test, the

plume was mainly composed of ash particles and SO2therefore the whole plume was analyzed with the LUT and VPR.

For example, Fig. 4a shows the BTD for the image taken at 21:27 UTC and Fig. 4b shows the plume mask chosen for

the ash and SO2plume.

Fig. 5 (top image) shows a comparison of ash and SO2 total mass computed with LUT and VPR for the event of 23

October 2011. Retrieval was achieved by analyzing 20 images collected by the SEVIRI radiometer aboard the MSG2

satellite, from 19:12 UTC to 23:57 UTC on October 23 2011. As previously noted, SEVIRI detected the volcanic cloud

over the sea only after 19:00 UTC (image time), while the maximum cumulative mass of both SO 2and ash was reached

around 21:00-21:15 UTC (image time). From the end of the volcanic emissions (the maximum cumulative mass) until

the last image considered on 23 October 23:57 UTC, no mass variations of either ash or SO2were measured over about

three hours.


13/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

The LUT and VPR trends were similar but clearly the VPR ash mass was systematically lower than the LUT ash mass.

Less obvious but equally clear is the effect on the SO 2mass trends. In this case the SEVIRI images appear noisy in the

area of sea around the volcanic cloud. This effect is mainly due to numerous thin and almost transparent meteorological

clouds affecting the whole area. In order to avoid the problem of clouds, the LUT procedure uses a cloud mask

algorithm to retrieve the sea surface temperature in a region surrounding the plume. Therefore, if some small

meteorological clouds are also present below the plume the LUT procedure interprets their effect as ash. Conversely,

since the VPR does not use a cloud mask algorithm, the presence of meteorological clouds near the plume also affects

the radiance retrieval, used in Eq. (1). The different behavior of the two procedures in relation to meteorologicalclouds is thought to be the explanation why VPR retrieval is lower than LUT retrieval.

Fig. 5 (bottom image) shows the percentage trends of the two particle classes considered in the volcanic cloud. The

decreasing trend through time of the coarse particles appears slower than that of the 12 August episode (Fig. 3), and the

VPR-LUT percentage comparison also exhibits a greater divergence.


14/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

Fig. 5 SO2 and ash total mass trends (top) and percentage of presence in the volcanic cloud of the two classes of

particles (fine with and coarse with ) considered (bottom) computed using the LUT andVPR procedures. The label Image time refers to the image acquisition time.


15/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

3.3 Wind velocity and direction retrieval: 12 August 2011 test case

This section presents and discusses the results of the determination of short-term variations in wind speed and

direction, obtained from the VPR maps of SO2and ash. In section 3.1, with reference to the 12 August test case and

aiming to compare the VPR and LUT results, the SO2and ash mass retrieved by the two procedures were analyzed only

for the part of the volcanic cloud containing only ash and SO 2. Here instead the whole volcanic cloud is considered,including all the pixels with SO2,ash particles, and ice crystals (evaluated as ash), because this procedure aims first to

clearly detect the mass distribution peak position (for wind speed determination) rather than the mass values

themselves.

The SEVIRI image is geolocated, and so each pixel position has specific latitude and longitude coordinates. The

evaluation of wind speed and direction at the plume altitude is conducted simply by determining the position of the

gravity centers of the aerosol particle peaks (ash/ice) and of the sulfur dioxide emitted from the volcano, as described in

section 2.4. The retrieved wind speeds and directions obtained from the SO2and aerosol maps may differ if the two

plumes are not collocated.

Fig. 6 (top image) shows the evolution of the distance of the gravity centers of the SO 2and ash/ice peaks from the Mt.

Etna craters computed using the SEVIRI images recorded from 9:42 to 16:57 on August 12. The positions of the two

peaks practically coincide, therefore, in Fig. 6 only their average wind speed and direction is shown.

The instantaneous wind speed shown in Fig. 6 (middle image) exhibits a minor trend. The wind speed increased from

about 10 to about 13 m s-1

from 10:00 to 12:00, it was almost constant from 12:00 to 14:30, but again increased to about

14 m s-1

from 14:30 to 17:00. The Trapani WMO upper air station measured a similar wind speed trend at plume

altitude (dashed line). The wind direction was also not perfectly constant. It was approximately constant at about 305

degrees from North until about 15:00, then slightly increased to 315 degrees at 17:00. The wind direction at Trapani

(dashed line) was about 270 degrees at midnight (11 August), 310 degrees at midday, and 330 degrees at midnight (12

August).


16/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

Fig. 6Evolution of the distance of the gravity center of the SO 2and ash/ice peaks (top). Mean wind speed (middle) and

direction (bottom) obtained from the procedure. The dashed lines represent the values from the Trapani upper-air

sounding station.


17/20


18/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

4 Conclusions

The VPR procedure formerly developed for the MODIS sensor was successfully applied to the MSG-SEVIRI

multispectral radiometer. All the parameters required by the VPR procedure to retrieve abundances in any Mt. Etna

volcanic cloud containing ash and SO2 were computed and presented here for all the currently available SEVIRI

sensors, namely MSG-1, MSG-2, and MSG-3. By following the scheme adopted for the MODIS sensor, the SEVIRIVPR ash and SO2 retrievals has been compared with the well established LUT approach. The results show good

agreement.

A method for reconstruction of the flux time series at the source of the volcanic emission, based on a SEVIRI image

sequence of an eruption event, was presented and applied to the VPR results. In principle, this method is also applicable

to the ash and SO2mass maps retrieved by other retrieval schemes, such as those based on the LUTs: starting from the

sulfur dioxide and ash maps, the SO2and ash fluxes at the volcano vents can be reconstructed using the wind speed at

plume altitude. If the wind speed is not known, it can be computed directly from the ash and SO2mass maps computed

from a sequence of SEVIRI images. The velocity retrieved by this procedure is inherently the speed of the wind

blowing on the volcanic plume at its specific altitude and time, and it is thus much more reliable than the wind speed

measured by an upper air station, which is recorded twice a day up to hundreds of kilometers away from the plume

region.

Therefore, in principle three input data are required for the complete VPR procedure (ash and SO 2mass map retrieval

plus flux reconstructions), all assumed to be constant for the whole plume: altitude, temperature, and velocity. However,

plume velocity can be obtained from the sequence of computed maps, and plume temperature can be estimated from the

images using the coldest pixel method. Therefore, plume altitude is the only real unknown data that requires

independent assessment.

The VPR procedure based on MSG SEVIRI data is extremely fast, and thus suitable for operational use during volcanic

crises to quickly provide quantitative estimations of the ash and sulfur dioxide contents of volcanic clouds. Ash particle

classification based on VPR retrieval gives additional information on the evolution of an ash cloud burden in relation to

particle size.

The ash and SO2fluxes of emitted mass reconstructed at source can give valuable insight into the dynamics of volcanic

events and extend the information available in a multiparametric approach. The reconstructed fluxes, together with the

plume speed and direction established from the retrievals, can be used as improved inputs for transport and deposition

models and so provide more reliable forecasting of volcanic ash cloud development.

Acknowledgments

This work was partially funded by the EC-FP7 APhoRISM project (Research, Technological Development and

Demonstration Activities, grant agreement n. 606738).

The authors would like to thank Gavin Taylor for carefully reading the original manuscript and for providing language

corrections.

References

Aiuppa A, Moretti R, Federico C, Giudice G, Gurrieri S, Liuzzo M, Papale P, Shinohara H, Valenza M (2007)Forecasting Etna eruptions by real-time observation of volcanic gas composition. Geology, 35(12): 11151118.

Berk A, Bernstein LS, Robertson DC (1989) MODTRAN: A Moderate Resolution Model for LOWTRAN 7. Rep. GL-TR-89-0122, Air Force Geophys. Lab., Bedford, MA.

Boichu M, Menut L, Khvorostyanov D, Clarisse L, Clerbaux C, Turquety S, Coheur PF (2013) Inverting for volcanic

SO2 flux at high temporal resolution using spaceborne plume imagery and chemistry-transport modelling: the 2010

Eyjafjallajkull eruption case study. Atmos. Chem. Phys., 13(17): 85698584.

Bourassa AE, Robock A, Randel WJ, Deshler T, Rieger LA, Lloyd ND, Llewellyn EJ, Degenstein DA (2012) LargeVolcanic Aerosol Load in the Stratosphere Linked to Asian Monsoon Transport . Science, 337(6090): 7881.

Casadevall TJ (1994) The 1989-1990 eruption of Redoubt Volcano, Alaska: impacts on aircraft operations. Journal of

Volcanology and Geothermal Research, 62(14): 301316.

Corradini S, Spinetti C, Carboni E, Tirelli C, Buongiorno MF, Pugnaghi S, Gangale G (2008) Mt. Etna tropospheric ash

retrieval and sensitivity analysis using Moderate Resolution Imaging Spectroradiometer measurements. J. Appl.

Remote Sens., 2, 023550, doi:10.1117/1.3046674.Corradini S, Merucci L, Prata AJ (2009) Retrieval of SO2 from thermal infrared satellite measurements: correction

procedures for the effects of volcanic ash. Atmospheric Measurement Techniques, 2(1): 177191.


19/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

Delmelle P, Stix J, Baxter P, Garcia-Alvarez J, Barquero J (2002) Atmospheric dispersion, environmental effects and

potential health hazard associated with the low-altitude gas plume of Masaya volcano, Nicaragua. Bull. Volcanol.,

64(6): 423434.

Duffell HJ, Oppenheimer C, Pyle DM, Galle B, McGonigle AJS, Burton MR (2003) Changes in gas composition prior

to a minor explosive eruption at Masaya volcano, Nicaragua. Journal of Volcanology and Geothermal Research,

126(34): 327339.

Horwell CJ, Baxter PJ (2006) The respiratory health hazards of volcanic ash: a review for volcanic risk mitigation.Bulletin of Volcanology, 69(1): 124.

Hufford G, Simpson JJ, Salinas L, Barske E, Pieri DC (2000) Operational considerations of volcanic ash for airlines.

Bull. Am. Meteorol. Soc., 22(6): 945967.

INGV (2011a) INGV weekly bulletin Etna of 16th August 2011. 33/2011,

http://www.ct.ingv.it/it/rapporti/multidisciplinari/cat_view/61-multidisciplinari.html?limitstart=0

INGV (2011b) INGV weekly bulletin Etna of 25th

October 2011. 43/2011,

http://www.ct.ingv.it/it/rapporti/multidisciplinari/cat_view/61-multidisciplinari.html?limitstart=0

Langmann B, Zakek K, Hort M, Duggen S(2010) Volcanic ash as fertiliser for the surface ocean. Atmos. Chem. Phys.,

10, 3891-3899, doi:10.5194/acp-10-3891-2010.

Mather TA, Pyle DM, Oppenheimer C (2003) Tropospheric volcanic aerosol. In: Robock A, Oppenheimer C (ed)Geophysical Monograph Series, 139, American Geophysical Union, Washington DC, pp 189212.

Merucci L, Burton M, Corradini S, Salerno GG (2011) Reconstruction of SO2 flux emission chronology from space-

based measurements. J. Volcanol. Geotherm. Res., 206(34): 8087.Miller TP, Casadevall TJ (2000) Volcanic ash hazards to aviation. Encyclopedia of volcanoes, pp 915930.

Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur Degassing From Volcanoes: Source Conditions, Surveillance,

Plume Chemistry and Earth System Impacts. Reviews in Mineralogy and Geochemistry, 73(1): 363421.

Picchiani M, Chini M, Corradini S, Merucci L, Sellitto P, Del Frate F, Stramondo S (2011) Volcanic ash detection and

retrievals using MODIS data by means of neural networks. Atmospheric Measurement Techniques, 4(12): 2619

2631.

Piscini A, Picchiani M, Chini M, Corradini S, Merucci L, Del Frate F, Stramondo S (2014) A Neural Network approach

for the simultaneous retrieval of volcanic ash parameters and SO2using MODIS data. Submitted to AMTD.

Prata AJ (1989) Infrared Radiative Transfer Calculations for Volcanic Ash Clouds. Geophys. Res. Lett., 16(11): 1293

1296.

Prata AJ (2009) Satellite detection of hazardous volcanic clouds and the risk to global air traffic. Nat. Hazards, 51(2):

303324.

Pugnaghi S, Teggi S, Corradini S, Buongiorno MF, Merucci L, Bogliolo MP (2002) Estimation of SO2 abundance inthe eruption plume of Mt. Etna using two MIVIS thermal infrared channels: a case study from the Sicily-1997Campaign. Bulletin of Volcanology, 64(5): 328337.

Pugnaghi S, Guerrieri L, Corradini S, Merucci L, Arvani B (2013) A new simplified approach for simultaneous retrieval

of SO2 and ash content of tropospheric volcanic clouds: an application to the Mt Etna volcano. Atmospheric

Measurement Techniques, 6(5): 13151327.

Realmuto VJ, Abrams MJ, Buongiorno MF, Pieri DC (1994) The use of multispectral thermal infrared image data to

estimate the sulfur dioxide flux from volcanoes: A case study from Mount Etna, Sicily, July 29, 1986. J. Geophys.

Res., 99(B1): 481488.

Robock A (2000) Volcanic Eruptions and Climate. Rev. Geophys., 38(2): 191219.

Rose WI, Durant AJ (2009) Fine ash content of explosive eruptions. Journal of Volcanology and Geothermal Research,

186(12): 3239.

Scollo S, Prestifilippo M, Pecora E, Corradini S, Merucci L, Spata G, Coltelli M (2014) Eruption Column Height

Estimation of the 2011-2013 Etna lava fountains. Annals of Geophysics, 57(2), S0214; doi:10.4401/ag-6396.Solomon S, Daniel JS, Neely RR, Vernier JP, Dutton EG, Thomason LW (2011) The Persistently Variable

Background Stratospheric Aerosol Layer and Global Climate Change. Science, 333(6044): 866870.

Sparks RSJ, Burski MI, Carey SN, Gilbert JS, Glaze LS, Sigurdsson H, Woods AW (1997) Volcanic Plumes. John

Wiley & Sons.

Sparks RSJ (2003) Forecasting volcanic eruptions. Earth and Planetary Science Letters, 210(12): 115.

Theys N, Campion R, Clarisse L, Brenot H, Van Gent J, Dils B, Corradini S, Merucci L, Coheur PF, Van RoozendaelM, Hurtmans D, Clerbaux C, Tait S, Ferrucci F (2013) Volcanic SO2 fluxes derived from satellite data: a survey

using OMI, GOME-2, IASI and MODIS. Atmos. Chem. Phys., 13(12): 59455968.

Thomas HE, Watson IM (2010) Observations of volcanic emissions from space: current and future perspectives. Nat

Hazards, 54(2): 323354.

Volz FE (1973) Infrared Optical Constants of Ammonium Sulfate, Sahara Dust, Volcanic Pumice, and Flyash. Appl.Opt., 12(3): 564568.

Wen S, Rose WI (1994) Retrieval of sizes and total masses of particles in volcanic clouds using AVHRR bands 4 and 5.J. Geophys. Res., 99(D3): 54215431.

Yu T, Rose WI, Prata AJ (2002) Atmospheric correction for satellite-based volcanic ash mapping and retrievals using

split window IR data from GOES and AVHRR. J. Geophys. Res.,107(4311), doi:10.1029/2001JD000706.


20/20

ACCEP

TED

MANUSCRIPT

ACCEPTED MANUSCRIPT

Highlights

We present the results of VPR procedure applied for the first time to SEVIRI images

VPR is able to obtain fast retrievals of ash and SO2abundances in volcanic clouds

Plume speed and direction are retrieved from the temporal series of ash and SO2maps

The fluxes of the volcanic emission applied to the VPR results are presented too

Two 2011 Etna lava fountain episodes were considered as test cases

Documents

Lorenzo Guerrieri