in the 2008 Kasatochi eruption - AGU Publications

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, D00L21, doi:10.1029/2009JD013634, 2010

Volcanic ash and SO2 in the 2008 Kasatochi eruption: Retrievals comparison from different IR satellite sensors S. Corradini,1 L. Merucci,1 A. J. Prata,2 and A. Piscini1 Received 1 December 2009; revised 8 October 2010; accepted 14 October 2010; published 23 December 2010.

[1] The Kasatochi 2008 eruption was detected by several infrared satellite sensors including Moderate Resolution Imaging Spectroradiometer (MODIS), Advanced Very High Resolution Radiometer (AVHRR), and Atmospheric Infrared Sounder (AIRS). In this work a comparison between the volcanic cloud SO2 and ash retrievals derived from these instruments has been undertaken. The SO2 retrieval is carried out by using both the 7.3 and 8.7 mm absorption features while ash retrieval exploits the 10–12 mm atmospheric window. A radiative transfer scheme is also used to correct the volcanic ash effect on the 8.7 mm SO2 signature. As test cases, three near‐contemporary images for each sensor, collected during the first days of the eruption, have been analyzed. The results show that the volcanic SO2 and ash are simultaneously present and generally collocated. The MODIS and AVHRR total ash mass loadings are in good agreement and estimated to be about 0.5 Tg, while the AIRS retrievals are slightly lower and equal to about 0.3 Tg. The AIRS and MODIS 7.3 mm SO2 mass loadings are also in good agreement and vary between 0.3 and 1.2 Tg, while the MODIS ash corrected 8.7 mm SO2 masses vary between 0.4 and 2.7 Tg. The mass increase with time confirms the continuous SO2 injection in the atmosphere after the main explosive episodes. Moreover the difference between the 7.3 and 8.7 mm retrievals suggests a vertical stratification of the volcanic cloud. The results also confirm the importance of the ash correction; the corrected 8.7 mm SO2 total masses are less than 30–40% of the uncorrected values. Citation: Corradini, S., L. Merucci, A. J. Prata, and A. Piscini (2010), Volcanic ash and SO2 in the 2008 Kasatochi eruption: Retrievals comparison from different IR satellite sensors, J. Geophys. Res., 115, D00L21, doi:10.1029/2009JD013634.

1. Introduction [2] Satellite observations of global volcanic activity represent the main source of valuable data when eruptions occur in remote and uninhabited places, as in the cases studied here for the August 2008 Kasatochi eruption. Volcanic plumes injected into the atmosphere can be detected, and their ash and gas composition (principally H2O, CO2, SO2 and HCl) characterized, and tracked during their evolution. These properties are increasingly important in the studies of the effects of volcanic eruptions on climate [Robock, 2000; Grainger and Highwood, 2003], on the environment [Thordarson and Self, 2003], and on public health [Horwell and Baxter, 2006]. In particular large amounts (Tg) of volcanic ash and SO2 gas erupted from volcanoes can enter the atmosphere to reach heights that make them hazardous to aircraft [Casadevall, 1994]. Several satellite‐borne instruments are capable of detecting both ash and gas. SO2 is more readily identified and quantified than 1 Istituto Nazionale di Geofisica e Vulcanologia, Centro Nazionale Terremoti, Rome, Italy. 2 Climate and Atmosphere Department, Norwegian Institute for Air Research, Kjeller, Norway.

Copyright 2010 by the American Geophysical Union. 0148‐0227/10/2009JD013634

ash, but it is less hazardous to jet aircraft. In the likely case that ash and SO2 are collocated and travel together, a sensible strategy is to identify SO2 and use it as a surrogate to locate airspace that may be affected by ash. In the thermal infrared, it has been demonstrated that volcanic ash and meteorological clouds can be discriminated by using two channels centered at 11 and 12 mm [Prata, 1989a, 1989b]. Microphysical properties of silicate‐bearing ash and radiation transfer models were used in later work to retrieve effective particle radius, aerosol optical thickness and mass concentrations of fine ash particles [Wen and Rose, 1994; Prata and Grant, 2001; Yu et al., 2002]. The ash retrieval sensitivity has recently been improved by introducing water vapor corrections in the radiation transfer [Prata and Grant, 2001; Yu et al., 2002; Corradini et al., 2008]. [3] SO2 retrieval in the Thermal Infrared (TIR) multispectral data was first carried out using the 8.7 mm centered SO2 absorption feature for low‐level tropospheric plumes [Realmuto et al., 1994, 1997; Teggi et al., 1999; Watson et al., 2004; Pugnaghi et al., 2006], and then for high‐ level stratospheric plumes using the 7.3 mm centered SO2 absorption feature [Prata et al., 2003]. A recent improvement has been proposed by introducing an ash correction in the SO2 retrieval: silicate‐bearing ash absorbs radiation in the whole TIR range, and more strongly around 8.7 mm than

D00L21

1 of 10

D00L21

CORRADINI ET AL.: KASATOCHI VOLCANIC ASH AND SO2 RETRIEVALS

Table 1. Satellite Instrument Characteristics Comparison

Spectral Range (mm) Number of Channels Nadir Spatial Res. (km2) Scan angle (deg) Swath Width (km)

MODIS

AVHRR

AIRS

0.4–14.4 36 1a 55 2330

0.6–12.5 6 1.21 55.3 2900

3.7–15.4 2378 225 49 1650

a

The MODIS spatial resolution refers to the TIR channels.

7.3 mm, thus affecting SO2 retrievals when both ash and SO2 are present in the volcanic clouds. In multispectral sensors where 11 and 12 mm bands are present together with the 8.7 mm SO2 band, ash can be retrieved independently and used to correct SO2 retrievals [Corradini et al., 2009]. [4] This work addresses the issue of comparing the ash and SO2 retrievals in the Kasatochi volcanic clouds by using different IR satellite measurements collected by the Moderate Resolution Imaging Spectroradiometer (MODIS), the Advanced Very High Resolution Radiometer (AVHRR) and the Atmospheric Infrared Sounder (AIRS). The results are also discussed to give information about the eruption mechanism, the volcanic cloud altitude stratification and the SO2 and ash collocation. The effect of the ash influence on 8.7 mm SO2 retrievals has also been taken into account by applying a novel ash correction procedure. Three images for each satellite instrument, collected during the first days of the eruption, have been analyzed and used as test cases. [5] The paper is organized as follows: in section 2 the Kasatochi eruption is briefly described from an IR satellite data perspective, while in section 3 an overview of the IR satellite instruments characteristics is provided. Section 4 describes the data sets considered for the retrieval comparison, while in sections 5 and 6 an overview of the ash and SO2 retrieval schemes applied to the different instruments is presented. Sections 7 and 8 report the discussion of the results and the conclusions.

2. The Kasatochi 2008 Eruption [6] Kasatochi volcano [52.18°N, 175.51°W] is a small (2.7 × 3.3 km wide, 314 m height a.s.l.) stratovolcano located in the Aleutian arc (Alaska) (Smithsonian Institution, Global Volcanism Program, http://www.volcano.si.edu; hereafter Smithsonian Institution Web site). Between August 7 and 8, 2008, three explosive eruptions occurred approximately at 22:01 UTC on August 7, 01:50 and 04:35 UTC on August 8 (C. F. Waythomas et al., Small volcano, big eruption, scientists rescued just in time, 2008, http://www.avo. alaska.edu/activity/Kasatochi08/Kasatochi2008PLW.php). A thick cloud of ash and sulfur dioxide drifted to great distances in the atmosphere at altitudes of about 9.1–13.7 km (Smithsonian Institution Web site). The volcanic sulfur dioxide mass loadings, estimated from 1.2 to 2.5 Tg (Alaska Volcano Observatory (http://www.avo.alaska.edu) and Rix et al. [2008], represent one of the largest volcanic sulfur dioxide mass loadings observed since Chile’s Hudson volcano erupted in August 1991. Forty‐four flights between the most northern American and Canadian destinations were canceled until August 11 (http://www.reuters.com/article/ domesticNews/idUSN1132219720080811) because of the

D00L21

volcanic ash released in the atmosphere during the eruption. Different satellite instruments were able to detect and track the eruption clouds for several days.

3. Satellite Instruments [7] MODIS is a multispectral instrument on board the EOS‐Terra and EOS‐Aqua polar satellites launched in 1999 and 2002, respectively, as part of the Earth Observing System (EOS) mission [Barnes et al., 1998] (see also MODIS Web site, http://modis.gsfc.nasa.gov/). MODIS has 36 spectral bands in the wavelength range from Visible (VIS) to TIR (together covering the spectral range, 0.4–14.4 mm). Bands 1 and 2 operate at a nominal spatial resolution at nadir of 250 m, while bands 3 to 7 are at 500 m and the remaining 29 bands at 1000 m. The sensor scans ±55° across‐track about the nadir from the EOS orbit altitude of 705 km, resulting in a 2330 km swath and full global coverage every 1–2 days. [8] AVHRR on board NOAA polar satellites is a multispectral sensor that acquires images in the wavelength range from VIS to TIR (0.6–12.5 mm). The first AVHRR, launched on TIROS‐N in October 1978, was a four‐channel radiometer. It was subsequently improved to a five‐channel instrument (AVHRR‐2), initially carried on NOAA‐7 launched in June 1981. The latest instrument version is AVHRR‐3, with six channels, first carried on NOAA‐15 launched in May 1998. With an orbit altitude of about 833 km and a scanning angle of ±55.3°, the total swath width is about 2900 km. The spatial resolution at nadir is 1.1 km and increases to about 5 km at the most extreme off‐nadir viewing angles [Gary, 2007] (see also AVHRR Web site, http://noaasis.noaa.gov/ NOAASIS/ml/avhrr.html). [9] AIRS, also on board the EOS‐Aqua satellite, is a hyperspectral spectrometer operating at IR wavelengths between 3.7 and 15.4 mm (2378 channels) with a scanning angle of ±49° from nadir and a swath width of 1650 km. The instantaneous field of view is 1.1° providing nadir pixels with dimensions 13.5 × 13.5 km2 [Chahine et al., 2006] (see also AIRS Web site, http://www.airs.nasa.gov) increasing to 18 × 40 km2 at the swath edge. [10] Table 1 summarizes the main characteristics of these satellite instruments.

4. Images Data Set [11] To allow a useful ash and SO2 retrieval comparison, satellite images have been selected to be as contemporary as possible. Three images collected during the first days of the Kasatochi eruption are considered as test cases. Because all of the MODIS images have been acquired by the sensor on board the Aqua satellite, they are coincident with the AIRS measurements. Images were collected on 8 August at 1340 UTC and on 9 August at 0050 and 2355 UTC. The AVHRR measurements were collected on the same days at 1304, 0042 and 2248 UTC by the sensor on board the NOAA‐18 satellite. Figure 1 shows the MODIS‐Aqua RGB images obtained by using the 28, 29 and 31 channels. Because of the strong absorption on channels 28 and 29 (see section 6), the intense blue identifies the volcanic cloud regions where SO2 is prevalent. The cloud areas have been estimated to vary between a latitude‐longitude grid of about 5° ×

2 of 10

D00L21


D00L21

Figure 1. RGB MODIS Kasatochi composite images using channels 28, 29, and 31 for (left) 8 August 2008 at 1340 UTC, (middle) 9 August 2008 at 0050 UTC, and (right) 9 August 2008 at 2355 UTC. The intense blue identifies the volcanic cloud.

10° (Figure 1, left and middle) to about 10° × 10° (Figure 1, right).

5. Volcanic Ash Retrieval Techniques [12] Volcanic ash particles are detected and retrieved in the 10–12 mm atmospheric window to exploit its different selective absorption compared to atmospheric water vapor. In sections 5.1 and 5.2, volcanic ash retrieval techniques applied to multispectral (MODIS and AVHRR) and hyperspectral (AIRS) TIR measurements are described and compared. 5.1. Estimating Ash Mass From MODIS and AVHRR [13] For multispectral instruments such as MODIS and AVHRR, ash detection is carried out by using the “reverse” absorption method [Prata, 1989a, 1989b] applied to channels centered around 11 and 12 mm (MODIS channels 31 and 32 and AVHRR channels 4 and 5). While the effective radius (re) and aerosol optical depth (t) are retrieved using a microphysical model and brightness temperatures at 11 mm and 12 mm [Wen and Rose, 1994; Prata and Grant, 2001; Yu et al., 2002; Ellrod et al., 2003]. For each pixel (p) the ash mass mp can be calculated from the simplified formula suggested by Wen and Rose [1994] using the retrieved t and re, 4 re mp ¼ Ap ; 3 Qext

ð1Þ

where Ap is the pixel area, r is the ash density and Qext is the extinction efficiency. The total ash mass is computed as the sum of the pixel masses. The simulated Top Of Atmosphere (TOA) radiances Look‐Up Table (LUT) needed for the retrieval is derived using the MODTRAN 4 [Berk et al., 1989; Anderson et al., 1995] Radiative Transfer Model (RTM). [14] To compute the TOA simulated radiances, atmospheric profiles (Pressure, Temperature and Humidity (PTH)), surface characteristics (temperature and emissivity), the volcanic plume geometry (plume altitude and thickness) and the volcanic ash optical properties are needed.

[15] Due to the large geographical area occupied by the volcanic clouds (see section 4), the atmospheric profiles (and in particular the relative humidity [see Corradini et al., 2008]) are among the most critical parameters to set. The PTH profiles were obtained from the ECMWF database. For each satellite image the volcanic cloud area has been divided into a 1° by 1° latitude‐longitude grid and for each node the atmospheric profiles have been extracted. The PTH profiles input to MODTRAN, are computed as the mean of all the node profiles (approximately 50 and 100 nodes for the 2 first and the last MODIS and AVHRR images, respectively). Figure 2 shows the 8 August at 1200 UTC ECMWF temperature and humidity atmospheric profiles together with the associated standard deviations. Although the standard deviation of the humidity is significant, for almost all the atmospheric levels it lies within 20%. During the eruption a dense cloud system was present under the volcanic cloud area. The meteorological cloud altitudes have been estimated by interpolating the Cloud Top Pressure level 2 MODIS products with the ECMWF atmospheric profiles. Figure 3 shows the level 2 MODIS cloud‐top pressure and the interpolated cloud altitude maps. A wide region just outside the volcanic cloud has been considered as representative of meteorological cloud system under the volcanic cloud. The meteorological cloud altitude is estimated as the mean altitude of the yellow area in Figure 3 (right), the red area represents the volcanic ash cloud estimated by using the retrieval procedure. The mean altitude is 2050 m with a standard deviation of 650 m. To compute the meteorological cloud surface temperature, the estimated altitude with its standard deviation have been reported in the ECMWF temperature plot (see the horizontal solid and dotted green lines in Figure 2). The meteorological surface temperature is found from the intersection between the green horizontal lines and the ECMWF atmospheric profile: the result is 278 ± 2 K. The procedure has also been applied to the other two MODIS images, obtaining a meteorological cloud system altitude of 1500 m and a cloud‐top temperature of 282 ± 2 K and 284 ± 2 K for 9 August images at 0050 and 2355 UTC, respectively. The Kasatochi volcanic cloud altitude was derived from the comparison between

3 of 10

D00L21


D00L21

Figure 2. The 8 August 1200 UTC ECMWF (left) mean atmospheric temperature and (right) relative humidity profiles computed in a 5° × 5° latitude‐longitude grid containing the volcanic clouds and used for the MODIS and AVHRR LUT generation. The horizontal bars represent the standard deviations. The green horizontal lines drawn in the ECMWF temperature plot indicate the meteorological cloud‐top altitude and are used for the meteorological cloud‐top temperature computation. The purple lines represent the volcanic cloud‐top temperature and are used for the volcanic cloud‐top altitude computation. the 11 mm brightness temperature of the MODIS 8 August most opaque pixels and the ECMWF atmospheric temperature profiles [see Prata and Grant, 2001; Corradini et al., 2008]. The temperatures of the coldest parts of the plume (statistically ∼ 5% of the total ash cloud pixels) have been estimated to be in the range –48°to –46°C. Because it is reasonable to expect the cloud not to be completely black [see Platt and Stephens, 1980], the air temperature at the top of the cloud is considered to be a few °C lower. The purple dotted vertical lines in Figure 2 indicate temperatures of –50° and –48°C, which cover the range of opaque cloud‐top temperatures observed in the MODIS image. These temperatures translate to heights of about 10 to 11 km. This result fits well with the volcanic cloud altitude information reported by the Smithsonian Institution (see section 2). Using these results, the volcanic cloud altitude was set to 11 km with an arbitrary thickness of 2 km for all of the images. The ash optical properties (single scattering

albedo, extinction coefficient and asymmetry parameter) are derived using a Mie code developed by the Earth Observation Data Group (EODG) of the Atmospheric Oceanic and Planetary Physics Department (Oxford University) using an andesite refractive index [Pollack et al., 1973]. The density of ash was set to 2.6 · 106 g/m3 [Neal et al., 1994]. [16] The final set of RTM simulations computed in a multiple scattering atmosphere uses 801 wavelengths (from 700 to 1500 cm−1, step 1 cm−1) × 4 angles (from 0 to 75°, step 25°) × 10 optical depths (from 0 to 10, constant step in a logarithmic scale) × 8 particle effective radii (from 0.4 to 10 mm, constant step in a logarithmic scale). [17] Every MODTRAN input parameter has an uncertainty that will cause errors in the ash retrievals. The total ash mass retrieval error was estimated by Corradini et al. [2008] in a sensitivity study considering the uncertainties of many parameters such as atmospheric profiles, plume geometry, surface characteristics (temperature and emissivity) and ash

Figure 3. The 8 August 1340 UTC MODIS image. (left) MODIS level 2 top cloud pressure. (middle and right) Cloud altitude. The red and yellow regions in the right image indicate the volcanic ash cloud and the selected area in which the altitude of the meteorological cloud system under the volcanic plume is estimated, respectively. 4 of 10

D00L21


type. A 40% error value is estimated, taking into account a 20% atmospheric humidity uncertainty found in the ECMWF profiles. 5.2. Estimating Ash Mass From AIRS [18] AIRS v5.0 level 1b geolocated radiances are obtained from the NASA GSFC DAAC using the Mirador data access tool (http://mirador.gsfc.nasa.gov/cgi‐bin/mirador/ presentNavigation.pl?project=AIRS&tree=project). The radiances are converted to brightness temperatures, quality controlled and three channel subset regions are selected for the retrieval: 800–960 cm−1, 910–980 cm−1, and 1070– 1130 cm−1. All three regions have sensitivity to the composition and size distribution of ash; the third region is also sensitive to the presence of SO2. The information contained in the AIRS channels is sufficient in many cases to determine the mass of fine ash through a retrieval scheme utilizing channels between 800 and 1000 cm−1, where fine (1–10 mm radius) silicate ash differentially absorbs and scatters radiation. The scheme used here (see Gangale et al. [2010] for a full description) is based on a minimization between simulated AIRS radiances and AIRS measurements. The retrieval scheme involves an off‐line radiative transfer calculation based on MODTRAN 4 RTM for absorption and emission of radiation by atmospheric gases in the vertical column between the surface and the ash cloud, which acts as a boundary condition for a multiple scattering calculation assuming a plane‐parallel ash cloud. A Mie scattering calculation is used to determine the scattering efficiencies for the ash cloud by considering an andesitic refractive index. As the refractive index data are not at the same resolution as the AIRS measurements, interpolation is necessary. This leads to potential errors and highlights the need for higher‐resolution refractive index data for a variety of ash compositions. The end result of the RTM calculation is a set of TOA spectral radiances corresponding to the AIRS channel radiances over the interval from about 669 cm−1 to about 1400 cm−1 (1215 channels). The RTM calculations are performed for a range of mean particle radii sizes ranging from radii of 0.25 mm up to 12 mm, infrared optical depths (0–10 at 1000 cm−1) and viewing angles. The final set of RT simulations occupies 17 (angles) × 100 (optical depths) × 1215 (channels) × 49 (particle effective radii bins). [19] Error sources in the retrieval arise from lack of knowledge of the surface leaving spectral radiances, which depend on the surface temperature (usually unknown) and the spectral emissivity (always unknown). To reduce this error, we arbitrarily normalize the brightness temperature spectra (simulations and measurements) by dividing by the brightness temperature at 1000 cm−1. The spectra, simulations and measurements are then compared and the simulation which shows the minimum squared difference with the measured (normalized) spectra is selected as the best estimate. The procedure is performed on a pixel by pixel basis and the simulation corresponding to the closest view angle (which is known) is used in the minimization, which reduces the comparison by a factor 17. Nevertheless, the scheme requires a large number of comparisons to be made and it is slow. The result of the minimization is an estimate of the particle effective radii, r (±0.25 mm) and optical depth, t (±0.1). The total mass in the cloud is the sum of the

D00L21

masses in each pixel calculated from equation (1). A first‐ order estimate of the accuracy of the retrieval can be made assuming errors of ±0.25 mm in radius, ±0.1 in optical depth, and 10% error on Qext. The RMS error for a single pixel is then ±0.001 Tg, or about 20–30%. [20] The use of normalized radiances represents the main difference between the MODIS‐AVHRR and AIRS ash retrieval procedures; in this latter case the uncertainty of the surface parameters is avoided and together with the greater spectral content, the AIRS retrieval may be considered generally more accurate. However, lack of sufficient AIRS spatial resolution compared with MODIS and AVHRR implies that only ash clouds that are large (>200 km2) can be properly characterized.

6. Volcanic SO2 Retrieval Techniques [21] SO2 has three main absorption features in the IR spectral range, around 4, 7.3 and 8.7 mm. The 4 mm feature lies in a transparent window but is very weak and is affected by solar scattered radiation during the day. The 7.3 mm signature is the strongest but because it is highly affected by the atmospheric water vapor, it is generally used when volcanic clouds rise to the upper troposphere/lower stratosphere. The 8.7 mm channel lies in a relatively transparent region, and is generally used to retrieve the lower tropospheric volcanic cloud, but is affected by volcanic ash absorption, and to a lesser degree by water vapor. The difference between the 7.3 and 8.7 mm retrievals is particularly important because it gives an indication about the altitude stratification of volcanic clouds. [22] In this work only the 7.3 and 8.7 mm SO2 absorption features will be considered for the MODIS and AIRS retrievals. 6.1. Estimating SO2 Mass From MODIS [23] The 8.7 mm SO2 retrieval scheme for multispectral IR measurements was first described by Realmuto et al. [1994] for the TIMS airborne spectrometer and was based on a weighted least squares fit procedure using instrumental measured radiances and simulated radiances obtained by varying the SO2 column abundance. Further work carried out by several authors allowed the extension of SO2 retrieval to different satellite sensors such as MODIS, ASTER and SEVIRI [Realmuto et al., 1997; Corradini et al., 2008; Watson et al., 2004; Pugnaghi et al., 2006; Corradini et al., 2009]. In this work we refer to the procedure scheme described by Corradini et al. [2009] in which the SO2 abundance is retrieved on a pixel‐by‐pixel weighted least squares fit procedure using satellite sensor measurements and TOA simulated radiances. A similar scheme was developed for use with the 7.3 mm radiances. MODIS channels 28 and 29 are used for the 7.3 and 8.7 mm SO2 retrievals, respectively. [24] The retrieval procedure described is limited by two main factors: the thermal contrast between the volcanic cloud‐top temperature and the surface below, and the presence of opaque pixels. In this latter case, when the TOA measured radiance is less than the minimum simulated radiance, the SO2 retrieval is not possible [see Corradini et al., 2009]. [25] In the same manner as for the ash retrievals, the simulated TOA radiances LUT needed for the SO2 retrievals

5 of 10

D00L21


D00L21

Figure 4. AIRS and MODIS SO2 retrievals from Kasatochi eruption. (a) The AIRS, (b) MODIS at 7.3 mm, and (c) MODIS at 8.7 mm retrievals of (top) the 8 August 2008 image at about 1340 UTC, (middle) the 9 August 2008 image at about 0050 UTC, and (bottom) the 9 August 2008 images at 2355 UTC. are obtained from MODTRAN 4. In this case 31 values of SO2 column abundance (from 0 to 15 g m−2, in steps of 0.5 g m−2) have been added to the LUT described before for the ash MODIS and AVHRR retrievals. [26] The SO2 total mass retrieval error was estimated by considering both the surface temperature and atmospheric water vapor uncertainties. As described in section 5.1 the meteorological cloud‐top altitude and the atmospheric water vapor profile uncertainties have been estimated to be ±2 K and ±20%, respectively. The simulated TOA radiances perturbed by the temperature and humidity uncertainties have been used for both the 7.3 and 8.7 mm SO2 retrievals. The variation from the unperturbed result has been assumed as the SO2 total mass retrieval uncertainty. The 7.3 and 8.7 mm retrieval errors are estimated to be 5 and 30% considering the temperature uncertainty and 15 and 5% considering the humidity uncertainty. The total mass retrieval errors, computed by the error propagation, are estimated to be 16 and 31% for 7.3 and 8.7 mm bands respectively. [27] During a volcanic eruption, both ash and SO2 are often emitted simultaneously. The volcanic cloud ash par-

ticles (from 1 to 10 mm) tend to reduce the TOA radiance in the entire TIR spectral range including the channels used for the SO2 retrievals. The net effect is a significant SO2 column abundance overestimation which is more significant in the 8.7 mm channel. To avoid the impact of ash on SO2 retrievals, a novel correction procedure developed by Corradini et al. [2009] has been applied. The procedure is based first on the simultaneous SO2 and ash properties (optical depth and effective radius) retrievals, then on the identification of the pixels containing both ash and SO2. For every identified pixel the SO2 retrieval is carried out again by taking into account its retrieved optical depth and effective radius. [28] In this work the ash correction has been applied only to the 8.7 mm SO2 retrieval. Note that according to Corradini et al. [2009], the procedure to identify and avoid the volcanic cloud opaque pixels is applied after the ash correction procedure. 6.2. Estimating SO2 Mass From AIRS [29] The retrieval of SO2 from AIRS uses channels across the 7.3 mm absorption band and follows the method pro-

6 of 10

D00L21


D00L21

Figure 5. (a) AIRS, (b) MODIS, and (c) AVHRR ash retrievals from Kasatochi eruption. Retrievals are for (top) 8 August 2008 image at about 1340 UTC, (middle) the 9 August 2008 image at about 0050 UTC, and (bottom) the 9 August 2008 images at 2355 UTC.

posed by Prata and Bernardo [2007]. The retrieval is a two‐ step process: in the first step a correlation method is used to identify the band by correlating the spectral shape with a library band shape. In the second step an optimal estimation technique [Rodgers, 2000] is used to retrieve the SO2 in slabs of 2 km thickness. These are then summed up to provide a partial column SO2 amount. Typical errors range from ±0.1 to ±0.2 g/m2, depending on the amount of water vapor present in the atmosphere and in the cloud. The retrieval is limited by several factors, including thermal contrast (the temperature difference between the SO2 cloud and the surface below), the amount of water vapor, which usually restricts the retrieval to SO2 that resides above 3 km or so, and saturation of the band. This last effect is evident in some of the retrievals for the early phase of Kasatochi, where the cloud is optically thick and the spectra appear flat. In all of these cases the AIRS retrievals underestimate the total column of SO2. Prata et al. [2010] show that by using the more transparent, but weaker 4 mm SO2 band, SO2 in the lower part of

the troposphere can be detected when the 7.3 mm retrieval detects no SO2.

7. Results and Discussion [30] Figures 4 and 5 display the SO2 and ash retrieval maps for the different satellite instruments and measurements. As Figures 4 and 5 show, the retrieved volcanic clouds have the same location and shape and the two components are well collocated. This latter consideration is very important for aviation safety issues, because it demonstrates that, at least for the first days of the eruption, the more readily identifiable SO2 could be used as a proxy for volcanic ash, which is extremely dangerous for jet aircraft engines. [31] All the MODIS 8.7 mm SO2 maps in Figure 4 are shown after the application of the ash correction procedure (see section 6.1). As an example, Figure 6 shows the MODIS SO2 retrieval at 8.7 mm before and after the ash correction together with the ash map, for the 9 August at

7 of 10

D00L21

D00L21


Figure 6. Ash correction procedure application on 8.7 mm SO2 retrieval from the Kasatochi 9 August 2008 MODIS image collected at 0050 UTC. (left) SO2 retrieved before ash correction, (middle) ash mass retrieved, and (right) SO2 retrieved after ash correction. 0050 UTC image. The SO2 total loadings are 2.42 and 1.57 Tg before and after the procedure application, which leads to a mass variation of approximately 35%. In particular, it is noted that on the SO2 map after the ash correction (Figure 6, right), the lower volcanic cloud branch disappears. This region was detected as SO2 (Figure 6, left), whereas our final results indicate that this was in fact ash. [32] Table 2 summarizes the results for the SO2 and ash total masses retrieved for all the cases considered; the MODIS 8.7 mm SO2 mass loadings are shown before and after the application of the ash correction procedure. Taking into account the differences in spatial and spectral resolution between the sensors, together with the assumptions needed for the retrievals, an agreement to within 30% can be considered good. [33] By considering that the AIRS 9 August image collected at 0050 UTC is not complete, a deeper analysis of Table 2 and Figure 4 reveals a generally slightly higher AIRS SO2 sensitivity compared to the MODIS 7.3 mm retrievals. This is confirmed by the presence of low values of SO2 detected exclusively by AIRS in the 8 August (see the upper east corner of the volcanic cloud) and 9 August at 2355 UTC (see the western volcanic cloud branch) images. Furthermore, the 8 August SO2 total mass retrievals indicate significantly lower values compared to the 9 August images. Such behavior cannot be explained by an effective low presence of volcanic SO2 in the atmosphere

during the first hours after the eruption. By looking at Figure 4 (top), it can be noted that for all the satellite measurements there are pixels for which the retrieval has not been carried out. In these regions the volcanic cloud is so optically thick that the AIRS spectra appear flat (see section 6.2) and the measured MODIS satellite radiances are less than the minimum simulated TOA radiances (see section 6.1). Note also that for this image the difference between the MODIS 8.7 mm SO2 total mass retrievals before and after the ash correction procedure is particularly high. This is only due to the elimination of the opaque pixels, carried out after the application of the ash correction procedure (see section 6.1.1). [34] Table 2 also shows that there are meaningful differences between the AIRS/MODIS SO2 7.3 mm and the MODIS 8.7 mm retrievals. Because the 8.7 mm channel is sensitive to all the atmospheric SO2 column while the 7.3 mm channel is sensitive only to its upper part (see section 6), such a difference can be explained assuming that the volcanic cloud is present both in the lower troposphere and the upper troposphere/stratosphere. This vertical stratification of the volcanic cloud has also been found by Kristiansen et al. [2010], using the FLEXPART dispersion model and by the Cloud‐Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) measurements. Also Prata et al. [2010] showed evidence of the tropospheric SO2 presence by using the AIRS 4 mm SO2 spectral signature. The MODIS 8.7 mm SO2 total mass retrievals have been also

Table 2. Total SO2 and Ash Masses (in Tg) Retrieved From AIRS, MODIS, and AVHRR During the First Few Days After Kasatochi Eruptiona Total SO2 Mass Retrievals (Tg) Date

Time

AIRS

MODIS (7.3 mm)

MODIS (8.7 mm)

8 August 9 August 9 August

1340 UTC 0050 UTC 2355 UTC

0.36 ± 0.05 1.03 ± 0.05 1.22 ± 0.05

0.30 ± 0.05 0.94 ± 0.15 0.87 ± 0.14

0.34 ± 0.11 (1.21 ± 0.37) 1.57 ± 0.49 (2.42 ± 0.74) 2.65 ± 0.82 (3.61 ± 1.10)

Date

Time

AIRS

MODIS

AVHRR

8 August 9 August 9 August

1340 UTC 0050 UTC 2355 UTC

0.21 ± 0.06 0.25 ± 0.08 0.35 ± 0.12

0.46 ± 0.18 0.44 ± 0.18 0.45 ± 0.18

0.41 ± 0.16 0.50 ± 0.20 0.46 ± 0.18

Total Ash Mass Retrievals (Tg)

The values between brackets are the 8.7 mm SO2 total masses retrieved before the ash correction.

a

8 of 10

D00L21


compared with GOME‐2 [Richter et al., 2009] (see also GOME‐2 Volcanic SO2 Alert, http://www.doas‐bremen.de/ gome2_so2_alert.htm) and IASI [Karagulian et al., 2010] estimations. The 8 August 2125 UTC and 9 August 2105 UTC GOME‐2 SO2 total mass amounts, are estimated to be 1.70 and 2.80 Tg, respectively, therefore in very good agreement with the results obtain from the MODIS 8.7 mm images (1.57 and 2.65 Tg for 9 August at 0050 and 2355 UTC, respectively; see Table 2). The IASI SO2 total mass retrievals indicate a good agreement for the 8 August pm image (1.7 Tg), while a meaningful difference exists for the 9 August pm image (1.6 Tg). The reasons for such a discrepancy are not completely known. On the one hand the MODIS higher value can be, at least partially, explained by the formation of the H2SO4 [Karagulian et al., 2010] that could affect the 8.7 mm channel. On the other hand, because the IASI retrieval is carried out by considering an SO2 profile ranging from 6 to 21 km, it is more sensitive to the upper tropospheric stratospheric SO2, and less to the SO2 lower tropospheric contribute particularly important in the 9 August pm image. In any case the results obtained represent a further and stronger evidence of the importance of the ash correction procedure application on the 8.7 mm retrieval when SO2 and ash are simultaneously present in a volcanic cloud. [35] Considering the two 9 August images, the SO2 total mass results indicate also that the 7.3 mm SO2 mass loading remains approximately constant (about 1 Tg) while the 8.7 mm SO2 loading continues significantly to increase (from 1.5 to 2.7 Tg). This latter result could indicate a continuous SO2 tropospheric injection after the last explosive event occurred on 8 August. Such SO2 lower troposphere mass increasing has been also retrieved by AIRS using the 4 mm spectral signature [Prata et al., 2010]. [36] Table 2 also emphasizes the very good agreement between the MODIS and AVHRR ash mass retrievals, estimated to be approximately constant and equal to about 0.5 Tg. Even if the AIRS ash mass retrievals are comparable with the other retrievals, they are generally lower and equal to about 0.3 Tg. In fact there are some ash features not retrieved by AIRS as for example the lower branch of the 9 August image at 0050 UT (see Figure 5, middle row) and the western branch of the 9 August image at 2355 UTC (see Figure 5, bottom row), detected instead by MODIS and AVHRR. The reason for this is definitely not clear. It could perhaps be due to the presence of the meteorological cloud system under the volcanic cloud, which could decrease the ash concavity shape, thus reducing the retrieved ash signal. The constant ash values for all the images indicates that, conversely to the SO2 tropospheric injection, no ash has been emitted after the main explosive episodes as reported by the Smithsonian Institution.

8. Conclusions [37] In this work, AIRS, MODIS and AVHRR SO2 and ash total mass retrievals in the IR spectral range from the August 2008 Kasatochi eruption have been compared. The 7.3 and 8.7 mm absorption features and the 10 and 12 mm atmospheric window have been used for SO2 and ash retrievals, respectively. Because the simultaneous presence of ash and SO2 in a volcanic cloud can lead to a significant 8.7 mm SO2 retrieval overestimation, an ash correction pro-

D00L21

cedure has also been applied. To allow a significant ash and SO2 retrieval comparison, the satellite images were selected to be as close in time as possible. Test cases involving three images for each instrument, acquired during the first days of the Kasatochi eruptions (8 and 9 August), were analyzed. [38] The results indicate simultaneous presence of SO2 and ash and a general agreement between the AIRS, MODIS and AVHRR retrievals. Except for the 8 August image, the SO2 mass loading is significantly greater than the ash mass loading. The importance of the ash correction on the SO2 retrieval at 8.7 mm, which is greater than 30–40%, is also confirmed. The ash and SO2 are generally collocated; that is, in this case SO2 could be used as surrogate of volcanic ash. This is an important issue for aviation safety being the SO2 detection easier than ash. [39] The 8 August image presents many saturated (opaque) pixels that cause an underestimate in the 7.3 and 8.7 mm AIRS and MODIS SO2 total mass retrievals, which amounts to about 0.3 Tg. For the 9 August images the AIRS and MODIS 7.3 mm SO2 mass loadings are in good agreement and equal to about 1 Tg, while the MODIS ash corrected 8.7 mm SO2 mass varies from 1.5 to 2.7 Tg. Because the 8.7 mm and 7.3 mm retrievals are sensitive to the SO2 present in the entire atmospheric column and to that present only in its upper part, respectively, the difference between the two retrievals gives an indication of the volcanic cloud stratification, i.e., of the volcanic cloud presence in both the troposphere and stratosphere. Such stratification has also been confirmed by the FLEXPART atmospheric transport model, CALIPSO measurements and AIRS 4 mm retrieval. The continuous increase of the 8.7 mm SO2 total mass with time suggest a continuous SO2 tropospheric injection after the last explosive event occurred on 8 August. The 9 August MODIS SO2 total mass retrievals at 8.7 mm has been also compared with GOME‐2 and IASI estimations. The results indicate a good agreement between MODIS and GOME‐2 and between MODIS and IASI for the 9 August image at 0050 UTC. A meaningful difference exists between MODIS and IASI for the 9 August 2355 UTC image. The results obtained confirm the need of the ash correction procedure application when the SO2 retrieval is carried out at 8.7 mm. [40] The MODIS and AVHRR total ash mass loadings are in a very good agreement and equal to about 0.5 Tg, while the AIRS retrievals are slightly lower and equal to about 0.3 Tg. The constant ash mass values with time suggests that no further ash mass emission occurred after the main explosions of 8 August. [41] Regarding future work, because one of the aim of this paper was to emphasize the potentiality of the simultaneous 7.3 and 8.7 mm SO2 retrievals to extract information about the volcanic cloud stratification, the CALIPSO measurements have been considered only as a posteriori knowledge to confirm the results obtained (see section 7). Moreover, in the frequent cases in which the information about the eruption and volcanic cloud characteristics are required as soon as possible (for the aviation safety issue for example), the procedure described will give important insights about the cloud stratification in near real time. Undoubtedly the SO2 and ash retrievals based on a one‐layer volcanic cloud scheme (see sections 5 and 6) could be improved by introducing a two‐layer scheme based on the information derived from the CALIPSO measurements. In this case the cloud

9 of 10

D00L21


stratification retrieved by CALIPSO could be used as a priori information to compute the LUT needed for the inversion procedure. A new retrieval scheme should be implemented considering at first the 7.3 mm SO2 retrieval which is sensitive the upper tropospheric stratospheric SO2 columnar content. Then, depending on the atmospheric water vapor amount, the 7.3 mm penetration depth (i.e., the altitude until which the channel is able to “see”) needs to be estimated [Prata and Bernardo, 2007; Prata et al., 2010]. Finally, by fixing the SO2 columnar content information obtained before, the lower tropospheric SO2 retrieval should be retrieved by using the 8.7 mm channel. [42] Acknowledgments. The authors would like to thank the European Space Agency and in particular Claus Zehner for the support given to carry out this work through the SAVAA (Support to Aviation Services for Volcanic Ash Avoidance) project. The authors would further like to thank John Peter Merryman Boncori for carefully reading the original manuscript and for providing language corrections. A special thanks go to the three anonymous referees whose comments significantly improved the paper.

References Anderson, G. P., et al. (1995), FASCODE/MODTRAN/LOWTRAN: Past/ present/future, paper presented at 18th Annual Review Conference on Atmospheric Transmission Models, Phillips Lab., Lexington, Mass. Barnes, W. L., T. S. Pagano, and V. V. Salomonson (1998), Prelaunch characteristics of the Moderate Resolution Imaging Spectroradiometer (MODIS) on EOS‐AM1, IEEE Trans. Geosci. Remote Sens., 36(4), 1088–1100. Berk, A., L. S. Bernstein, and D. C. Robertson (1989), MODTRAN: A moderate resolution model for LOWTRAN7, Rep. GL‐TR‐89‐0122, Air Force Geophys. Lab., Bedford, Mass. Casadevall, T. J. (1994), The 1989‐1990 eruption of Redoubt Volcano Alaska: Impacts on aircraft operations, J. Volcanol. Geotherm. Res., 62(30), 301–316. Chahine, M. T., et al. (2006), AIRS: Improving weather forecasting and providing new data on greenhouse gases, Bull. Am. Meteorol. Soc., 87(7), 911–926, doi:10.1175/BAMS-87-7-911. Corradini, S., C. Spinetti, E. Carboni, C. Tirelli, M. F. Buongiorno, S. Pugnaghi, and G. Gangale (2008), Mt. Etna tropospheric ash retrieval and sensitivity analysis using Moderate Resolution Imaging Spectroradiometer measurements, J. Appl. Remote Sens., 2, 023550, doi:10.1117/12.823215. Corradini, S., L. Merucci, and A. J. Prata (2009), Retrieval of SO2 from thermal infrared satellite measurements: Correction procedures for the effects of volcanic ash, Atmos. Meas. Tech., 2, 177–191. Ellrod, G. P., B. H. Connell, and D. W. Hillger (2003), Improved detection of airborne volcanic ash using multispectral infrared satellite data, J. Geophys. Res., 108(D12), 4356, doi:10.1029/2002JD002802. Gangale, G., A. J. Prata, and L. Clarisse (2010), The infrared spectral signature of volcanic ash determined from high‐spectral resolution satellite measurements, Remote Sens. Environ., 114(2), 414–425. Gary, D. (2007), History of the NOAA satellite program, J. Appl. Remote Sens., 1, 012504, doi:10.1117/1.2642347. Grainger, R. G., and E. J. Highwood (2003), Changes in stratospheric composition chemistry, radiation and climate caused by volcanic eruptions, Spec. Publ. Geol. Soc. London, 213, 329–347. Horwell, C. J., and P. J. Baxter (2006), The respiratory health hazards of volcanic ash: A review for volcanic risk mitigation, Bull. Volcanol., 69, 1–24, doi:10.1007/s00445-006-0052-y. Karagulian, F., L. Clarisse, P. F. Coheur, A. J. Prata, D. Hurtmans, and C. Clerbaux (2010), Detection of SO2, ash, and H2SO4 using the Infrared Atmospheric Sounding Interferometer (IASI), J. Geophys. Res., 115, D00L02, doi:10.1029/2009JD012786. Kristiansen, N. I., et al. (2010), Remote sensing and inverse transport modeling of the Kasatochi eruption sulfur dioxide cloud, J. Geophys. Res., 115, D00L16, doi:10.1029/2009JD013286. Neal, C. A., R. G. McGimsey, C. A. Gardner, M. L. Harbin, and C. J. Nye (1994), Tephra‐fall from 1992 eruptions of Crater Peak, Mount Spurr Volcano, AK: A preliminary report on distribution, stratigraphy and composition, U.S. Geol. Surv. Bull., 2139, 220 pp.

D00L21

Platt, C. M. R., and G. L. Stephens (1980), The interpretation of remotely sensed high cloud emittances, J. Atmos. Sci., 37, 2314–2322. Pollack, J. B., O. B. Toon, and B. N. Khare (1973), Optical properties of some terrestrial rocks and glasses, Proc. Icarus, 19, 372–389. Prata, A. J. (1989a), Observation of volcanic ash clouds using AVHRR‐2 radiances, Int. J. Remote Sens., 10(4), 751–761, doi:10.1080/ 01431168908903916. Prata, A. J. (1989b), Infrared radiative transfer calculations for volcanic ash clouds, Geophys. Res. Lett., 16(11), 1293–1296. Prata, A. J., and C. Bernardo (2007), Retrieval of volcanic SO2 column abundance from Atmospheric Infrared Sounder data, J. Geophys. Res., 112, D20204, doi:10.1029/2006JD007955. Prata, A. J., and I. F. Grant (2001), Retrieval of microphysical and morphological properties of volcanic ash plumes from satellite data: Application to Mt. Ruapehu, New Zealand, Q. J. R. Meteorol. Soc., 127(576), 2153–2179. 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 Atmosphere, Geophys. Monogr. Ser., vol. 139, edited by A. Robock and C. Oppenheimer, pp. 75–92, AGU, Washington, D. C. Prata, A. J., G. Gangale, L. Clarisse, and F. Karagulian (2010), Ash and sulfur dioxide in the 2008 eruptions of Okmok and Kasatochi: Insights from high spectral resolution satellite measurements, J. Geophys. Res., 115, D00L18, doi:10.1029/2009JD013556. Pugnaghi, S., G. Gangale, S. Corradini, and M. F. Buongiorno (2006), Mt. Etna sulfur dioxide flux monitoring using ASTER‐TIR data and atmospheric observations, J. Volcanol. Geotherm. Res., 152, 74–90, doi:10.1016/j.jvolgeores.2005.10.004. Realmuto, V. J., M. J. Abrams, M. F. Buongiorno, and D. C. Pieri (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), 481–488. Realmuto, V. J., A. J. Sutton, and T. Elias (1997), Multispectral thermal infrared mapping of sulfur dioxide plumes: A case study from the East Rift Zone of Kilauea Volcano, Hawaii, J. Geophys. Res., 102(B7), 15,057–15,072. Richter, A., F. Wittrock, A. Schönhardt, and J. P. Burrows (2009), Quantifying volcanic SO2 emissions using GOME‐2 measurements, poster EGU2009‐7679 AS3.15 XY247 presented at European Geosciences Union General Assembly, EGU, Vienna. Rix, M., P. Valks, N. Hao, T. Erbertseder, and J. van Geffen (2008), Monitoring of volcanic SO2 emissions using the GOME‐2 satellite instrument, paper presented at 2008 Second Workshop on Use of Remote Sensing Techniques for Monitoring Volcanoes and Seismogenic Areas, IEEE, Naples, Italy. Robock, A. (2000), Volcanic eruptions and climate, Rev. Geophys., 38(2), 191–219. Rodgers, C. D. (2000), Inverse Methods for Atmospheric Sounding: Theory and Practice, World Sci., Singapore. Teggi, S., M. P. Bogliolo, M. F. Buongiorno, S. Pugnaghi, and A. Sterni (1999), Evaluation of SO2 emission from Mount Etna using diurnal and nocturnal multispectral IR and visible imaging spectrometer thermal IR remote sensing images and radiative transfer models, J. Geophys. Res., 104(B9), 20,069–20,079. Thordarson, H., and S. Self (2003), Atmospheric and environmental effects of the 1783–1784 Laki eruption: A review and reassessment, J. Geophys. Res., 108(D1), 4011, doi:10.1029/2001JD002042. Watson, I. M., V. J. Realmuto, W. I. Rose, A. J. Prata, G. J. S. Bluth, Y. Gu, C. E. Bader, and T. Yu (2004), Thermal infrared remote sensing of volcanic emissions using the moderate resolution imaging spectroradiometer, J. Volcanol. Geotherm. Res., 135(1–2), 75–89, doi:10.1016/j. jvolgeores.2003.12.017. Wen, S., and W. I. Rose (1994), Retrieval of sizes and total masses of particles in volcanic clouds using AVHRR bands 4 and 5, J. Geophys. Res., 99(D3), 5421–5431, doi:10.1029/93JD03340. Yu, T., W. I. Rose, and A. J. Prata (2002), Atmospheric correction for satellite‐based volcanic ash mapping and retrievals using “split window” IR data from GOES and AVHRR, J. Geophys. Res., 107(D16), 4311, doi:10.1029/2001JD000706. S. Corradini, L. Merucci, and A. Piscini, Istituto Nazionale di Geofisica e Vulcanologia, Centro Nazionale Terremoti, Via di Vigna Murata 605, I‐00143 Rome, Italy. ([email protected]; [email protected]; [email protected]) A. J. Prata, Climate and Atmosphere Department, Norwegian Institute for Air Research, Instituttveien 18, N‐2027 Kjeller, Norway. ([email protected])

10 of 10