Journal of Geophysical Research: Atmospheres

Satellite remote sensing analysis of the 2010 Eyjafjallajökull volcanic ash cloud over the North Sea during 4–18 May 2010

Authors


Abstract

[1] Using the Moderate Resolution Imaging Spectroradiometer (MODIS), Spinning Enhanced Visible and Infrared Imager (SEVIRI), Clouds and the Earth's Radiant Energy System (CERES) instrument, and BaE146 aircraft data sets, we provide an overview of volcanic ash spatial distribution for six days (4–18 May 2010) and assess their properties and radiative impacts for 17 May primarily over the North Sea. We describe spectral signatures of volcanic ash, compare the MODIS-retrieved 550 nm aerosol optical thickness (AOT) and effective radii with the aircraft data, and then assess the change in radiative fluxes at the top of atmosphere using CERES. Our results indicate that the MODIS and SEVIRI thermal channels are adept at identifying volcanic ash near the source. However, the volcanic ash far from the volcanic source, especially over land, is contaminated by surface/atmospheric features. We assess the 17 May case in detail and show that MODIS AOTs (0.23–0.86) are higher than the aircraft values (0.07–0.54), probably due to different aerosol models used in the retrieval process. The MODIS effective radii values are between 0.4 and 0.9 μm with fine mode fraction values between 0.4 and 0.7. The aircraft-derived effective radii values are between 0.82 and 1.2 μm. The TOA shortwave radiative forcing for unit AOT of volcanic ash aerosols at the time of the satellite overpass is −77 ± 4.0 W m−2 and is larger than the longwave forcing per unit optical depth (11 ± 1.2 W m−2) by seven times indicating that ash could significantly impact radiative energy fluxes.

1. Introduction

[2] In April 2010, the Eyjafjallajökull volcano, located on the southern coast of Iceland (19.6°W, 63.6°N), spewed more than 250 million cubic meters of ash and debris into the atmosphere [Sigmundsson et al., 2010]. The ash was transported southeasterly by the polar jet stream into Great Britain, Ireland, and Northern Europe where its progress was tracked and monitored by various surface based lidars [e.g., Ansmann et al., 2011]. The arrival of ash clouds over Europe prompted the shutdown of large swaths of commercial airspace, bringing air travel to a virtual halt. This event prompted the closing of much of the airspace over Europe for nearly a week, displacing millions of passengers and resulting in monetary losses to airlines in excess of 1 billion U.S. dollars. Conditions improved by late April, but Eyjafjallajökull remained in an active phase through the month of May, since it had the ability to produce further large ash clouds with little to no warning.

[3] Satellite remote sensing is an important tool for monitoring and assessing the spatial distribution of volcanic ash. Satellite information can be used as a verification tool for models that forecast the spatial distribution and concentrations of ash at various altitudes [Millington et al., 2012]. Volcanic Ash Advisory Centers (VAACs) predominantly utilize high resolution atmospheric dispersion models to forecast height dependent ash concentrations. These forecasts are validated and tested using satellite [e.g., Millington et al., 2012], lidar remote sensing [e.g., Ansmann et al., 2011], and in situ [e.g., Johnson et al., 2012] measurements. For model forecasts to be effective, knowledge about the spatial and temporal distribution of aerosols and their properties such as size, shape, and composition is needed. While satellite data are adept at providing global view of volcanic ash, concentrated aircraft campaigns provide detailed information on ash properties [e.g., Johnson et al., 2012] that are needed to validate and improve satellite detection and retrievals. Therefore, a synergistic approach that combines satellite, in situ (aircraft/ground-based), and models to monitor and forecast volcanic ash is useful.

[4] In this study, we focus on satellite remote sensing of volcanic ash during six days in May 2010 over Great Britain and the North Sea when the Facility for Airborne Atmospheric Measurements (FAAM) BAe146 research aircraft measurements was available. We further provide a detailed analysis of satellite remote sensing and aircraft measurements for 17 May 2010 over the North Sea when several coincident spatiotemporal measurements were available. We primarily report 550 nm columnar aerosol optical depth (AOT) from the Moderate Resolution Imaging Spectroradiometer (MODIS) from the Terra and Aqua satellites along with spatial distribution from high temporal resolution Spinning Enhanced Visible and Infrared Imager (SEVIRI) data. Finally for the 17 May case, we compare the AOT and effective radii between MODIS and aircraft values and calculate the top of atmosphere net radiative forcing of volcanic ash clouds over the North Sea using MODIS and The Clouds and Earth's Energy System (CERES) data.

2. Data

[5] A variety of satellite data was available to assess the spatial distribution of aerosols including polar orbiting and geostationary satellites. The SEVIRI on the Meteosat Second Generation (MSG) and the GOES-E (Geostationary Operational Environmental Satellites–East) satellites were especially useful for monitoring volcanic ash on an hourly basis, although the SEVIRI has more infrared channels that are suitable for ash detection [Francis et al., 2012]. Polar orbiting satellites including MODIS from Terra and Aqua, MISR (Multiangle Scanning Spectroradiometer) from Terra, and OMI (Ozone Monitoring Instrument) from Aura were also useful for various purposes including assessment of ash spatial distribution and height (from MISR) and absorbing properties of ash and SO2 (from OMI). Each sensor has its own strengths and weaknesses for volcanic ash detection, although a common major issue is cloud cover. When ash is present below clouds, passive sensors cannot detect these aerosols. Also, we have limited capabilities for assessing volcanic ash above clouds from satellite data. However, the Cloud Aerosol Lidar with orthogonal Polarization (CALIOP) on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite is an active lidar system that can provide vertical distribution of aerosols, although the sampling is not useful for daily monitoring. Recent work also indicates that, by combining Ultraviolet (UV) measurements from OMI with the MODIS, aerosols above clouds may be assessed [e.g., Wilcox, 2011]. The Infrared Atmospheric Sounding Interferometer (IASI) on the Meteorological Polar Orbiting Satellite (METOP) also provided extremely useful high spectral infrared measurements of volcanic ash [Gangalea et al., 2010]. Several other satellites were also in orbit during the Eyjafjallajökull volcano eruption including the (Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), Polarization and Anisotropy of Reflectances for Atmospheric Sciences coupled with Observations from a Lidar (PARASOL), and Atmospheric Infrared Sounder (AIRS), but that is not the focus of this paper.

[6] We provide a brief description of the sensors and products that were used for studying volcanic ash including Terra and Aqua MODIS from polar orbiting platforms that provide routine retrievals of aerosol properties [King et al., 1992] and SEVIRI on the MSG-2 for examining the temporal variation of volcanic ash. Furthermore, we use the CERES top of atmosphere (TOA) fluxes to assess the radiative forcing of aerosols. The MODIS is a 36 channel Spectroradiometer that measures reflected and emitted radiation from the UV to the thermal part of the electromagnetic spectrum with a spatial resolution from 250 m to 1 km. It provides near daily global coverage with a morning and afternoon sun synchronous orbit with equatorial crossing time of 10:30 and 13:30 from Terra and Aqua, respectively. The MODIS also provides AOT and effective radii (re) of aerosols at a spatial resolution of 10 km based on a lookup table approach [Remer et al., 2005]. We use the MODIS-retrieved AOT to show the spatial distribution of volcanic ash and to validate those data against aircraft measurements. Note that this AOT does not indicate at which height the ash is concentrated, rather only the columnar values of the unitless AOT.

[7] The SEVIRI on MSG has four channels in the visible to the UV (0.4–1.6 μm) and eight infrared channels (3.9–13.4 μm) with a spatial resolution ranging from 1 (High Resolution Visible, HRV channel) to 3 km (for reflected solar and infrared channels). Data are available every 15 min and are most useful for tropical and midlatitude regions. The high temporal resolution makes it a valuable asset for monitoring volcanic ash, although routine retrievals of aerosol properties (e.g., AOT) are not available both over land and ocean from SEVIRI.

[8] The FAAM aircraft data used in this study includes information from the Leosphere 355 nm Lidar and in situ aerosol properties from the Passive Cavity Aerosol Spectrometer probe (PCASP) and Cloud and Aerosol Spectrometer (CAS). The Leosphere instrument has also been tested aboard an aircraft during AMMA [Chazette et al., 2007]. Aerosol extinction and AOT at 355 nm were retrieved from the lidar (for full details, see Marenco et al. [2011]) for the altitude range of 2 km up to 300 m below the aircraft altitude. The lidar data shown for the 17 May case is for the altitude range 2–7.5 km, which encompassed all observed aerosol layers identified as being volcanic ash. The AOTs were integrated over every minute, equivalent to an along-track distance of 8–10 km. Some of the AOT at 355 nm (∼20%) was attributed to fine-mode (radii < 0.3 μm) and assumed to be sulphuric acid and/or sulphate. The lidar AOTs do not include the contribution from boundary layer aerosol (aerosol below 2 km), although for the 17 May case that we analyze in detail (section 3), the AOT contribution below 2 km is less than 0.05. Ash mass concentrations were estimated from the lidar-derived aerosol extinction coefficient using specific extinction coefficients derived from the in situ measurements (0.72 m2 g−1 for the 17 May case). The ash mass concentration was also derived from the CAS, as described by Johnson et al. [2012], which corresponded to particle diameters from 0.6 to 35 μm. Effective diameters were calculated from the combined Passive Cavity Aerosol Spectrometer Probe (PCASP) and CAS aerosol size distribution that covered the diameter range 0.1–35 μm.

[9] The area of study is between 45° and 60°N and 5°E to 15°W. The period of study includes 6 days (4, 5, 14, 16, 17, and 18 May) that coincided with aircraft measurements. Table 1 provides a summary of the case study days and locations. Marenco et al. [2011] describe the lidar data and the associated uncertainties in detail for these 6 days where measurements of the ash layer height and layer optical depth were available. Table 1 also shows the “peak” lidar-derived AOT values and their locations. Turnbull et al. [2012] analyze the 17 May case in more detail primarily from aircraft measurements and retrievals.

Table 1. Summary of Case Study Days Analyzed From BAe146 Measurements
DateFlightObservation PeriodMajor LocationPeak Lidar AOT With Locations
4 MayB5261004–1551over Irish Sea0.27 (54°N, 5°W)
5 MayB5270911–1507Irish and North Sea0.36 (52°N, 5°W)
14 MayB5281007–1917Scotland and England0.32 (60°N, 7°W)
16 MayB5291255–1810Scotland and England0.84 (55°N, 4°W)
17 MayB5301126–1658Irish and North Sea0.66 (54°N, 2°E)
18 MayB5310944–1454North Sea0.20 (54°N, 1°E)

3. Methods and Analysis

[10] In this section, we investigate the volcanic ash cases from 4 to 18 May 2010 by combining satellite and airborne data sets. The spectral signatures of ash are first discussed and compared with clouds and other aerosol types. The satellite retrievals are then compared against aircraft data. A simple thresholding algorithm is designed and applied to SEVIRI data. Finally we provide a comprehensive evaluation of ash radiative impacts for 17 May primarily over the North Sea.

3.1. Spectral Signatures

[11] Detecting volcanic ash from multispectral satellite imagery is a non-trivial task. As an example, Figure 1 shows a true color composite of Aqua MODIS imagery from 11 May 2010. Near the source, the volcanic ash plume is clearly visible, but much of the ash away from the source region is above/beneath the cloud layers that cannot be detected by MODIS imagery alone. Probably the biggest challenge is to separate volcanic ash from ice clouds from reflected solar measurements. However, optically thick water clouds can readily be identified using thermal bands due to their low height and warmer cloud top temperatures. Moreover, dust, ash, and ice cannot be separated only from reflected solar measurements. The thermal channels are also useful in this regard [Prata, 1989]. The inset shows the spectral signatures of volcanic ash in the thermal infrared channels. The x axis is the 11 μm temperature, and the y axis is the 11–12 brightness temperature differences for various samples shown in the true color composite imagery. Often it is difficult to separate volcanic ash from ice clouds, but using the brightness temperature difference (BTD) between 11 and 12 μm (T11–T12) produces spectral separation due to the reverse absorption noted by Prata [1989], Prata and Grant [2001], and Pavolonis et al. [2006]. Ash is usually more absorbing at 11 μm compared to 12 μm; therefore, BTD is usually negative. This is in contrast to both water and ice clouds where BTD is usually positive. However, note that these are merely broad guidelines, and accurate detection of ash depends upon spatial resolution and spectral widths of satellite sensors, the atmosphere above the volcanic ash, ash concentration, viewing geometry, and surface features. The inset in Figure 1 clearly indicates that, closer to the source, the volcanic ash over the Atlantic Ocean has negative BTD values (<2.0) and, therefore, can be easily identified. However, the ash far from the source even over the ocean is difficult to detect as the BTD includes contributions from both clouds and ash. Detection is even more challenging over land at low optical depths since the spectral separation of ash from the underlying surface is ambiguous. Note the cluster of points for the pixels over ocean near the source is much tighter compared to ash over land with more scattered points due to the underlying background. Samples identified as cloud phase 1, 2, and 3 (C1, C2, and C3, respectively) shown in the inset have different cloud top temperatures indicating ice, possibly mixed-phase and water clouds. The BTD for these 3 cloud types are positive, but note these are idealistic situations where we have selected only a handful of samples to illustrate the point. As cloud/ash properties change over different surface and atmospheric conditions, these spectral signatures alone cannot be used to accurately identify these features.

Figure 1.

Aqua MODIS true color imagery for 11 May 2011 to assess thermal spectral signatures (inset) of volcanic ash and clouds. Various samples are shown in this figure whose spectral signatures are discussed in Figure 2.

[12] To examine this in further detail, we plotted (Figure 2) the spectral dependence (from the UV to near-IR) of various features for samples shown in Figure 1. Some well-known features include the water vapor absorption centered on the 0.94 μm channel and the 1.38 μm channel that is also used for cirrus detection [King et al., 1992]. Samples denoted as cloud phase 1 (C1), cloud phase 2 (C2), and cloud phase 3 (C3) have different reflectances at 550 nm primarily due to their optical depth. From the inset in Figure 1, it can be seen that C1 is ice cloud with cloud top temperatures (Tc) colder than 230 K; C2 is probably mixed phased with Tc around 255 K; and C3 is water cloud with Tc around 275 K. Ice cloud reflectivity is low at 1.6 μm since ice absorbs at this channel. Water cloud (C3) has higher reflectances at 1.6 μm because reflectivity at this wavelength is a function of particle size [King et al., 1992]. The spectral features of ice clouds are different enough to allow separation of various features. Not all aerosols, on the other hand, have clear separability in the UV to IR part of the spectrum. Here we point out selected key features of various aerosols. The volcanic ash plume close to the source (labeled as ash near source, in green, Figure 2) appears to be almost spectrally flat (i.e., reflectance does not vary much with wavelength) between 0.43 and 0.56 μm and then shows higher values until 0.87 μm with values quickly dropping to near zero at higher wavelengths. This spectral variation of reflectance is critical in the MODIS lookup up table method when comparing against simulated values (from a radiative transfer model) based on the four fine and five coarse models. Typically in this case, the retrieved Fine Mode Fraction (FMF, ratio of fine mode AOT to total AOT) will be lower indicating coarse mode particles with larger effective radii. Indeed the FMF for this volcanic ash sample was near zero and the effective radii were 1.2 μm. The same logic applies for dust aerosols that were sampled over Atlantic from the Sahara (shown in yellow), which has similar spectral dependence of ash near the source. Therefore, it is difficult to separate the various aerosols from reflected solar radiation alone. For smoke aerosols (black line) the reflectivity decreases with wavelength with higher UV values (due to Rayleigh scattering and aerosol absorption) with values near zero at near-IR wavelengths since smoke due to its small particle size is transparent to near-IR (and beyond) radiation. With this spectral dependence, the retrieved FMF will be higher since it denotes smaller sub-micron particles. Indeed the FMF for the smoke aerosol sample was 0.991 with an effective radii of 0.23 μm. Volcanic ash far from the source (blue) can be observed in Figure 1 (blue points of insets) and Figure 3 (in red color), both of which are over ocean but distant from Eyjafjallajökull. Spectral signatures of these pixels far from the source are quite different than the dust or the volcanic ash samples near the source. They appear to be similar to the smoke aerosol spectral variation. With respect to statistics, over six channels from 550 to 2130 nm used by MODIS in its AOT retrieval, the correlation coefficient between reflectance of volcanic ash and smoke aerosols is 0.73, which is higher than that between ash and dust aerosol (0.69). Although the aircraft measurements indicate a higher effective radii [Marenco et al., 2011], the MODIS retrievals indicate a smaller value since the spectral variation of ash pixels appear spectrally similar to that of smoke aerosols.

Figure 2.

Spectral variation from the UV to the near-IR for various samples shown in Figure 1. Also shown are smoke (near Florida) and dust (Sahara) over ocean spectral signatures for comparison. The spectral signatures of the 17 May 2010 ash (see Figure 3e) over the North Sea are shown in red.

Figure 3.

Terra-MODIS true color imagery showing clouds in white, overlaid with MODIS 550 nm aerosol optical depth retrievals in color for (a) 4 May, (b) 5 May, (c) 14 May, (d) 16 May, (e) 17 May, and (f) 18 May 2010. Also shown are flight tracks from the BAe146.

[13] After aerosols have been identified and surface conditions have been specified, the MODIS inversion procedure utilizes six reflectances from 550 to 2130 nm wavelengths and retrieves three pieces of information: the AOT, effective radii (re) and FMF [Remer et al., 2005]. The lookup table approach uses four fine modes (effective radii between 0.1 and 0.25 μm) and five coarse modes (effective radii between 0.98 and 2.50 μm) each with unique real and imaginary part of refractive index and, therefore, a unique set of extinction, single scattering albedo, and asymmetry parameters in a radiative transfer model. Volcanic ash is not included specifically in these models. The coarse mode aerosols in the lookup table correspond to either sea salt or dust like aerosols [Remer et al., 2005]. The effective radii for these nine models range from 0.10 to 2.50 μm. In the retrieval process, for each aerosol pixel, the algorithm looks for a combination of fine and coarse mode models that best fits the “measured” reflectances [Remer et al., 2005]. Therefore, the choice of the aerosol models that the algorithm picks based on the MODIS reflectances is crucial for proper estimation of AOT and aerosol properties. As in any satellite retrieval algorithm, there are several sources of uncertainties. Separating aerosols from clouds, specifying lower boundary conditions for retrievals, and selection of aerosol models are among the major uncertainties. Based on visual inspection of data for this case, Aqua MODIS appears to capture the spatial distribution of aerosols well for the 17 May case over the North Sea. Therefore, cloud contamination is probably not an issue. Since the AOT values are greater than 0.3 for a majority of these pixels, we assume that characterization of surface reflectance is not a major uncertainty.

3.2. Comparisons of Satellite Retrievals and Aircraft Data

[14] Figures 3 and 4 show the six case days over the area of study during the time of the Terra and Aqua overpass, respectively. Johnson et al. [2012] discuss the aircraft data and uncertainties in detail. Also shown on these figures are the flight tracks during these days. See Table 1 for further information on these flights along with the peak lidar-derived AOT values and their locations. A true color composite of the MODIS red (620 nm), green (550 nm), and blue (459 nm) bands shows clouds in white and ocean in black. Overlaid on these are the MODIS 550 nm AOT from blue to red color scales where red shows high AOT values of 1.0. These figures show the challenges in detecting volcanic ash from satellite imagery during the Terra and Aqua overpass times, which is around 10:30 and 13:30 LST. The ubiquitous cloud cover during 4–18 May (except 17 May over the North Sea) provides only selected opportunities for aircraft-satellite comparisons at large spatial scales. Also, the ash clouds were variable in space and time and were found between altitudes of 2–8 km and were often 0.5–3 km deep [Marenco et al., 2011]. The Terra and Aqua images for the same days also provide a sense of how quickly the clouds and volcanic ash moved through the area. On 4 May 2010, the ash features were similar (between Terra and Aqua) west of the United Kingdom with AOT values around 0.3–0.4. The aircraft derived mean ash concentrations were between 4.39 and 18.33 μg m−3 (Table 2). The major uncertainties on ash mass concentration are determined by several factors, such as the errors of the CAS instrument (i.e., sizing accuracy and counting accuracy), the uncertainty on the microphysical model used for ash (i.e., particle shape and refractive index), and the uncertainty on ash density [Johnson et al., 2012]. On 5 May there was persistent cloud cover in the images and aircraft-retrieved peak concentrations were between 200 and 600 μg m−3 between 3 and 4 km. On 14 May, high values of AOT (>0.4) were reported by Terra MODIS, but this observations from Terra did not provide AOT retrievals near the flight path. 16 May also provided limited opportunities for comparisons. 17 May provided the best opportunity to compare the Aqua MODIS and aircraft retrievals since, for most of the flight over the North Sea, AOT values ranged between 0.3 and 0.8 with a few pixels having higher values and with flight times very close to the Aqua overpass (see Table 1). Also, the pixel in Terra MODIS AOT (1°E, 55°N) just north of the flight track (Figure 3e) appears to be cloud contaminated and therefore appears as a hot spot.

Figure 4.

Same as Figure 3 but for Aqua-MODIS.

Table 2. Aerosols Observations From BAe146 Aircraft Measurements
DateAltitude Range of Ash Plume (km)Ash Mass Concentration (μg m−3)AOTPeak Extinction (m−1)
4 May1.99–5.4611.36 ± 6.970.17 ± 0.070.24 ± 0.11
5 May1.41–5.2523.11 ± 12.850.09 ± 0.050.17 ± 0.10
14 May2.52–8.11105.6 ± 60.680.31 ± 0.180.39 ± 0.16
16 May1.95–6.3639.18 ± 32.490.20 ± 0.160.27 ± 0.18
17 May2.26–5.6537.23 ± 31.420.18 ± 0.150.21 ± 0.14
18 May1.39–4.9919.27 ± 8.670.09 ± 0.040.13 ± 0.06

[15] Flight B530 on 17 May provided data to assess the height of the volcanic ash and to validate/improve satellite retrievals. The flight started at Nantes and flew over various airports including Gatwick, Heathrow, Stansted, Luton, Bristol, Cardiff, Birmingham, and Manchester to assess the ash concentrations and height. Often times, for safety reasons, the aircraft does not sample high ash concentrations. The main ash plume was located over the North Sea with layer heights between 4 and 6 km and with typical concentrations of 300–650 μg m−3, while reaching maximum values of 800–1900 μg m−3 in some relatively small high density patches [Marenco et al., 2011]. Typical values on the flight were around 300 μg m−3 and typical layer depths were reported to be around 1.3 km. Table 3 shows the collocated aircraft and MODIS collection 5.1 retrievals for the 17 May case. Note that the MODIS retrievals are column quantities whereas the aircraft integrates values over profiles. Also note that these comparisons are made for pixels over the North Sea only. Johnson et al. [2012, Table 5] indicate that, based on extinction coefficients, the 355 nm AOT is virtually the same as the 550 nm.

Table 3. Comparisons Between Collocated MODIS and Aircraft Aerosol Data
TimeLongitudeLatitudeBaE146 AOTMODIS AOTMODIS Effective Radius (μm)MODIS Fine Mode Fraction
355 (nm)550 (nm)550 (nm)
14:340.2854.030.190.190.2570.3850.666
14:370.6254.010.280.280.2280.5430.675
14:410.9654.000.350.350.3570.5990.509
14:491.8053.980.540.540.7290.5920.539
14:511.9753.980.390.390.4820.3360.549
14:552.2853.990.310.310.5980.6770.502
14:602.7554.000.460.460.7370.9010.474
14:682.7254.040.380.380.7690.9230.46
14:742.2854.030.300.300.5940.6780.527
15:650.1654.460.150.150.3180.2900.625
15:690.4954.470.120.120.3400.5350.516
15:720.8154.480.070.070.3300.3060.651
15:740.9754.490.150.150.7850.6640.496
15:791.4554.500.270.270.4830.4410.591
15:872.2654.510.220.220.8580.6160.518
15:912.5854.520.170.170.3820.3210.612
15:932.7354.520.200.200.8140.9540.439
16:032.7354.510.180.180.8140.9540.439

[16] There are several significant features to note in this comparison, while keeping in mind that there are uncertainties in both the satellite retrievals and the aircraft data sets. First the MODIS AOT values are much higher than the aircraft values especially for the larger AOT values from MODIS. To be specific, when MODIS reports lower AOT values (0.2 ∼ 0.4), they are in good agreement with aircraft retrievals, which varies from 0.07 to 0.35. When MODIS reports AOT values between 0.4 and 0.7, the aircraft retrievals are smaller and are between 0.27 and 0.39. Larger discrepancies between MODIS and aircraft data can be seen when MODIS reported larger AOT values (0.7 ∼ 0.9), where the corresponding aircraft retrieved AOT values are only from 0.15 to 0.54. Although the aircraft values are only from the ash and AOT contributions below the ash layer are not included, for the 17 May case the AOT contributions below the ash layer was less than 0.05.

[17] The MODIS FMF varies from 0.4 to 0.7. For larger particles such as dust, the FMF is usually around 0.5. For smoke particles, FMF values could be higher than 0.85 [Kaufman et al., 2005]. Volcanic ash particles usually have particle radii between 1 and 10 μm [Marenco et al., 2011] depending upon proximity to source and other factors and larger particles quickly settle out, but these ash clouds can be seen from IR imagery. We averaged data from 4 BaE146 profiles and the mean effective radii reported by the aircraft were between 0.8 and 1.2 μm. The MODIS reported values were between 0.4 and 0.9 μm. Therefore, it appears that the MODIS based on the spectral dependence of the six visible to near-IR channels is selecting models that have smaller effective radii. Without access to the actual lookup up tables (reflectance versus AOT values for various aerosol models) that went to retrieving AOT and effective radii, we can only surmise that the spectral signatures of volcanic ash in the visible and near-IR are similar to aerosols that have predominantly smaller particle radii.

3.3. Volcanic Ash Identification Based on SEVIRI

[18] Since polar orbiting data sets are limited in their ability to provide high temporal resolution information for events such as volcanic ash, we used the SEVIRI to assess the diurnal variation for volcanic ash clouds for 17 May 2010. Rather than providing the customary RGB composites using IR channels (red: 12.0–10.8 μm, green: 10.8–8.7 μm, blue: 10.8 μm), we developed a simple SEVIRI volcanic ash detection method by using a set of thresholds to identify ash. Figure 5 is the flow diagram for the SEVIRI volcanic ash detection scheme. The first test uses the brightness temperature difference between the 3.9 and 10.8 μm channels (BTD 3–11) and it includes the solar zenith angle (θ) since the 3.9 μm channel is affected by solar radiation during the daytime:

display math

This test can be applied over both land and water pixels where values larger than 3 K indicate cloud [Allen et al., 1990; Baum and Trepte, 1999]. If this test passes, then the U.S. Geological Survey (USGS) global land cover characteristics database version 2.0 is used to determine whether the pixel is land or water. If the pixel is water, then we use two spatial tests where the standard deviation of the surrounding 3X3 group of pixels is computed for the 1.6 and 12 μm test. The pixel is considered cloud contaminated if the standard deviation is larger than 1.2 for either test. These thresholds were set after analyzing dozens of scenes over the area of study. Then, if the pixel is still considered cloud free, it undergoes the final test which uses 11–12 brightness temperature differences (BTD 11–12). If this BTD is less than −0.2 K, then the pixel is classified as ash [Prata, 1989] while all other pixels are considered ash free. Note that for land pixels we do not use the spatial tests due to the inhomogeneity of the land surface. Figure 6 shows the results of applying these thresholds to identify volcanic ash for the 17 May case between 11:00 and 16:00 UTC. The simple algorithm appears to capture the ash pixels well, but more importantly, we see very few clouds misclassified as ash pixels.

Figure 5.

The flowchart description of the threshold test for volcanic ash detection.

Figure 6.

Hourly SEVIRI 0.6 μm images over England from 11:00 to 16:00 UTC for 17 May 2010. Clouds are displayed in shades of white along with darker background ocean and land surfaces. The track of the FAAM aircraft flight is shown in white. Dust is represented by its BTD11–12 temperature difference with the corresponding scale ranging from −5 to 0 K, and is mainly located over the North Sea.

3.4. Volcanic Ash Direct Radiative Effects at Top of Atmosphere

[19] Finally, Aqua CERES SSF data product which provides collocated data from CERES and MODIS instruments at CERES footprint resolution was used to assess the change in top of atmosphere (TOA) shortwave flux as a function of 550 nm MODIS-retrieved AOT [Geier et al., 2001]. The CERES measures broadband radiances that are converted to fluxes using angular models using AOT and FMF values [e.g., Zhang et al., 2005]. Figure 7 shows the relationship between the Aqua MODIS AOT at 550 nm and the CERES shortwave and longwave fluxes for the volcanic ash pixels over the North Sea (only over ocean). High correlation coefficients between CERES fluxes and MODIS AOT can be seen (0.884 in SW and −0.792 in LW, respectively). Extrapolating the regression line to the Y intercept indicates that for zero AOT, the SW flux is 77 W m−2, which is the clear sky value that is often reported in various studies for assessing the radiative forcing of aerosols [Christopher, 2011]. For a unit AOT, the shortwave flux is 153 W m−2 indicating that the shortwave forcing (clear – aerosol fluxes) due to these aerosols is about −76 W m−2. Newman et al. [2012] found a broadband aerosol shortwave (0.3–3 μm) radiative efficiency of approximately −130 W m−2 per unit AOT based on the comparison of aircraft mounted pyranometers with lidar AOT, and on two-stream radiative transfer calculations. Their results are not directly comparable with the CERES-MODIS result as the aircraft measured upwelling radiation at ∼7.5 km rather than the top of the atmosphere, the boundary layer aerosol (below 2 km) was not included in the lidar AOT retrieval, and differences in the timing of measurements lead to different solar zenith angles which lead to different instantaneous radiative forcing sensitivities [e.g., Osborne et al., 2011]. There is also variability in aerosol radiative efficiency between the aircraft overpasses shown by Newman et al. [2012], which is computed as the slope of the regression between SW fluxes and AOTs. The radiative efficiency has units if Watts per square meter per unit optical depth. The first of the two high altitude aircraft runs suggests a radiative effect of −100 W m−2 per AOT, closer to the CERES-MODIS value. The CERES longwave flux is related to MODIS 550 nm AOT through equation (2) below:

display math

indicating that for unit AOT the longwave forcing is about 11 W m−2 and therefore the shortwave forcing dominates the longwave forcing with a resultant net radiative forcing of −65 W m−2 for a unit AOT. For satellite observational-based methods, estimated uncertainties are mainly from CERES measurement uncertainties (calibration of CERES radiances, ∼0.4 W m−2 and conversion of CERES filtered radiances to radiances, ∼0.4 W m−2), radiance to flux conversion (∼0.4 W m−2), cloud biases (∼0.5 W m−2), and errors in clear sky fluxes calculations [Wielicki et al., 1996; Loeb et al., 2001, 2005; Zhang et al., 2005; Patadia et al., 2008]. The uncertainty of clear sky fluxes are estimated from the regression relationship between TOA SW fluxes and MODIS AOT, hence can be calculated by multiplying the maximum uncertainty in MODIS AOT of 0.05 [Levy et al., 2007] with instantaneous SW radiative efficiency of −76 W m−2 τ−1 and LW radiative efficiency of 11 W m−2 τ−1, which yields uncertainty in flux of 1.90 W m−2 for SW and 0.28 W m−2 for LW [Zhang et al., 2005]. Assuming all certainties are independent, the total uncertainty in fluxes estimations can be calculated based on equation (3) [Penner et al., 1994]:

display math

where Ui is the uncertainty factor from each individual source of uncertainty and Ut is the total uncertainty factor. According to equation (3), all sources of uncertainties are combined. The averaged uncertainties in the “instantaneous” clear sky aerosol radiative effects are 4.0 W m−2 for SW and 1.2 W m−2 for LW, respectively.

Figure 7.

Relationship between collocated Aqua-MODIS 550 nm AOT with CERES shortwave and longwave fluxes for 17 May 2010. The number of points and linear regression relationships for the shortwave (FSW) and longwave fluxes (FLW) and correlation coefficients are also shown in the inset.

4. Summary and Conclusions

[20] Satellite remote sensing is an important tool for assessing the spatial distribution of volcanic ash and its properties. However, most sensors are adept at providing columnar information in cloud-free conditions rather than height-dependent values. The polar orbiting and geostationary satellites were extremely useful in mapping spatial properties of volcanic ash. Aircraft data provided independent retrievals of volcanic ash that were compared against MODIS retrievals. The following are the key conclusions from our study:

[21] 1. The agreement between satellite and aircraft AOT values are better for lower AOT values but for higher AOTs (as reported by MODIS) the differences are larger. The differences are probably due to selection of aerosol models in the MODIS retrieval process.

[22] 2. Spectral signatures are useful for identifying ash near the source, but the infrared channels used to separate ash from other features are not robust enough in areas far from source, especially over land. Therefore, spatial/textural/angular methods combined with multitemporal approaches must be developed.

[23] 3. The net radiative forcing of volcanic ash at the time of the satellite overpass (instantaneous) is about −77 ± 4.0 W m−2, and the shortwave effect is larger than the longwave by nearly 7 times (11 ± 1.2 W m−2) and, therefore, could have a significant impact on the radiative energy budget.

[24] To assess ash above clouds, multisensor approaches need to be developed, and there is a critical need for future aerosol sensors to provide 3-D retrievals of aerosol properties that can aid Volcanic Ash Advisory Centres.

Acknowledgments

[25] This research is supported by NASA's Radiation Sciences Program. The data were obtained through the NASA Langley and Goddard Distributed Active Archive Systems.