Abstract

In this paper, accurate UV polarization measurements are performed on a volcanic ash cloud after long-range transport at Lyon, France (45.76°N, 4.83°E). The volcanic particles are released from the mid-April 2010 eruption of the Eyjafjallajökull Icelandic volcano (63.63°N, 19.62°W). The aerosol UV depolarization, which arises from nonspherical volcanic ash particles, serves as an independent means to discriminate ash from nonash particles in the volcanic cloud. This discrimination is only feasible if the intrinsic ash particle depolarization ration δash is accurately determined. In this paper, the δash value [δash = (40.5 ± 2.0)%] is derived from literature laboratory measurements on a scattering matrix to ensure ash particle specificity. It is shown that traditional approaches, based on direct lidar depolarization ratio δa measurements, are only valid very close to the source region, because δa may be very different from δash, when nonash particles are present. For the first time, observed lidar depolarization ratios, in the range from a few percent to 40%, are hence interpreted in comparison with δash taken as a reference. It is then shown how to use the sensitive and accurate UV polarization measurements to access to the size-averaged number concentration vertical profile of volcanic ash particles in the troposphere. Vertical profiles of the backscattering coefficient specific to volcanic ash particles, providing altitude-resolved ash particle number concentrations, are presented in the troposphere after optical scattering computation. This new methodology can be applied to other aerosols events and for other optical remote sensing experiments.

1. Introduction

Volcanic eruptions eject gas and volcanic ash particles in the atmosphere, thus affecting the earth’s climate and atmospheric chemistry (Ramaswamy et al. 2001). To quantify the impact of volcanic ash, the evaluation of ash mass, ash number, and even surface concentration is necessary (Ravishankara 1997). Ash mass concentrations may be determined from lidar measurements and are useful for airspace closures (the aviation safety limit is 2000 μg m−3). Ash number concentrations are often measured by filtration and sampling using either optical particle counters (Kaaden et al. 2009; Schumann et al. 2011) or inverse modeling of measured aerosol optical thicknesses (Tsanev and Mather 2007). Such measurements are, however, sensitive to an ensemble of particles and are not generally specific to volcanic ash particles, so that the retrieved number concentrations must be carefully analyzed. In addition, the volume particle size distribution (PSD) may be determined from complementary approaches relying on numerical simulations (Veselovskii et al. 2010), with algorithms being improved by additional input from lidar observations at several wavelengths. Lidar remote sensing provides fast and reliable access to the volcanic aerosols backscattering properties under real atmospheric conditions of temperature and relative humidity, as shown by Winker and Osborn (1992) for the Mount Pinatubo stratospheric eruption, or more recently for the 2002 Mount Etna tropospheric eruption (Wang et al. 2008). Moreover, as explained by Mishchenko et al. (2002), polarization-sensitive lidar systems can be used as shape indicators, as first done in the troposphere by Sassen et al. (2007) in the visible spectral range close to the particle source region.

We herein propose accurate UV polarization measurements performed on a volcanic cloud after long-range transport, when particles are highly dispersed and aged, which is the most frequently observed situation. The UV spectral range is chosen to be sensitive to both fine and coarse particles. Our calibrated polarization experiment, which is very seldom in the UV spectral range, serves to discriminate ash from nonash particles in the volcanic cloud, as first done by Shimizu et al. (2004) for dust and nondust particles in the visible spectral range. The UV polarization measurements are then used to develop a new methodology to remotely retrieve the volcanic ash particle number concentration Nash at altitude z MSL by applying scattering matrix formalism. This methodology is totally new: it has never been applied to any lidar data, especially in the UV spectral range. Moreover, we show that the UV polarization backscattering experiment provides ash particle specificity.

The novelty of this work is hence threefold. First, to ensure ash particle backscattering specificity, the intrinsic ash particle depolarization ratio δash = (40.5 ± 2.0)% is derived from laboratory measurements on volcanic ash particles performed by Muñoz et al. (2004) in the frame of the scattering matrix. It is shown that the traditional approach, where δash is directly retrieved from the lidar depolarization ratio δa measurements, is only valid if ash particles are the only detected particles, that is, they are very close to the source region. Second, the observed δa variations can then be interpreted by analyzing the contribution of nonash particles to the δa profile: δa equals δash only when there is no detectable nonash particle. Third, we show how our sensitive and accurate UV polarization measurements can be used to access to ash-specific number concentration vertical profile Nash in the troposphere, which is new.

The paper is organized as follows. Section 2 is devoted to UV polarization measurements and error bar analysis, performed at Lyon, France (45.76°N, 4.83°E), during the mid-April 2010 eruption of the Eyjafjallajökull volcano in Iceland (63.63°N, 19.62°W). Section 3 details the interpretation of the retrieved δa values as a function of δash. The determination of δash from laboratory scattering matrix measurements is achieved and then discussed by analyzing the possible influence of ash particle morphology, atmospheric aging, and sedimentation processes. Then, the retrieved δa values are interpreted as a function of δash because, after long-range transport, nonash particles are present in the volcanic cloud. It is shown that these nonash particles contribute to δa, which is hence found to always be below δash. This provides a way to discriminate between ash and nonash particles, and to further develop in section 4 the new methodology, giving remote access to the ash particle number concentration vertical profile. Hence, in section 4, from the δa interpretation in section 3, we provide vertical profiles of ash particles backscattering coefficient βash, leading to Nash after optical scattering computation, for two case studies observed at Lyon on 1200 UTC 18 April and 1800 UTC 19 April. The retrieved Nash vertical profiles, which are by construction ash particle specific, are then discussed to show the originality and the robustness of this new retrieval methodology, which is found to be independent of possible nonash depolarization, within our error bars.

2. Accurate UV polarization lidar measurements

a. Lidar experimental setup

Our lidar remote sensing experiment is ground based and polarization sensitive to specifically address the volcanic ash nonspherical particles. The lidar laser wavelength (λ = 355 nm) has been chosen to be sensitive to the fine and coarse modes of the PSD (Schumann et al. 2011). A detailed discussion on the PSD will be given (see Fig. 3). The laser source is a 10-Hz repetition rate neodymium-doped yttrium aluminum garnet (Nd:YAG) laser providing 10-mJ laser energy at λ = 355 nm wavelength after third harmonic generation, hence ensuring eye safety. A beam expander reduces the laser beam divergence down to 0.4 mrad while emitted UV laser pulses are almost perfectly linearly polarized (better than 1:10 000). Backscattered photons from the atmosphere [aerosols (a) and molecules (m)] are collected with a 200-mm-diameter f/3 Newtonian telescope, before entering our home-built UV polarization detector through a 3-mm-diameter pinhole, to reduce sky background contribution. The field of view of our lidar system is 5 mrad to reduce multiple scattering under clear-sky conditions. Each polarization channel (//, ⊥), defined with respect to the laser linear polarization, is composed of two successive UV polarizing beam-splitter cubes, a very selective interference filter (Δλ = 0.35 nm), and a Licel photomultiplier tube. These optical components have been specified separately at the laboratory to ensure that there is no leakage of the parallel polarization channel (//) into the perpendicular one (⊥). Photoelectrons are then sampled with a 40-MHz Licel analog-to-digital (A/D) converter. To reduce statistical noise, each lidar vertical profile is an average over 4000 laser shots. The lidar signals P// and P are then range averaged after high-frequency noise filtering, leading to a 75-m-altitude resolution.

b. Accurate aerosol backscattering and depolarization vertical profiles

Vertical profiles of parallel (perpendicular) aerosol backscattering coefficients βa,//a,⊥), and the corresponding particle depolarization ratios δa = βa,⊥a,// are displayed in Fig. 1 at 1200 UTC 18 April and 1800 UTC 19 April. The βa,// vertical profiles are evaluated from the parallel lidar channel P// by applying Klett’s algorithm (Klett 1985) to correct for the particle extinction. We computed the parallel scattering ratio R// = 1 + βa,//m,// by choosing a boundary value above the ash layer (Miffre et al. 2011). For the Klett inversion, the mean value for the aerosol extinction-to-backscatter ratio was S = (55 ± 5) sr, according to Raman measurements performed after long-range transport during the mid-April 2010 eruption of the Eyjafjallajökull volcano (Ansmann et al. 2010). Vertical profiles of βm,// and βm,⊥ are determined from molecular scattering computation by including our interference filter bandwidth, using reanalysis model from the European Centre for Medium-Range Weather Forecasts (ECWMF). In the absence of hydrometeors (under clear-sky conditions), P is exclusively linked to the presence of nonspherical particles, which provides depolarization of the backscattered laser light. Hence, βa,⊥ is nonspherical particle specific and is directly determined from the measured total (a, m) depolarization δ = β//,

 
formula

where δm = βm,⊥m,// = 0.37% is the retrieved molecular depolarization and δ is derived from the measured lidar depolarization δ* = P/P//, by performing accurate δ calibration as proposed by Alvarez et al. (2006); the calibration constant G = δ*/δ is known with a 2% uncertainty). Please note that Eq. (1) is deduced from Winker and Osborn’s (1992) expression of δa

 
formula

since δa = βa,⊥a,// and R// − 1 = βa,//m,//. Equation (2) is then used to retrieve the δa vertical profiles from R// and δ. Despite strong UV molecular scattering, error bars on δa, calculated from Eq. (2), remain below 30% for altitudes lower than 5 km MSL.

Fig. 1.

Vertical profiles of aerosol backscattering coefficients βa,// and βa,⊥ and aerosol depolarization ratio δa at (top) 1200 UTC 18 Apr and (bottom) 1800 UTC 19 Apr. For altitudes up to 5 km MSL, the uncertainty on βa,// values is below 15%, evaluated by applying the Klett (1985) algorithm for maximum and minimum S values. Error bars on βa,⊥a) are retrieved from Eq. (1) [(2)]. Because of the overlap function, the lidar signals start from 0.6 km MSL.

Fig. 1.

Vertical profiles of aerosol backscattering coefficients βa,// and βa,⊥ and aerosol depolarization ratio δa at (top) 1200 UTC 18 Apr and (bottom) 1800 UTC 19 Apr. For altitudes up to 5 km MSL, the uncertainty on βa,// values is below 15%, evaluated by applying the Klett (1985) algorithm for maximum and minimum S values. Error bars on βa,⊥a) are retrieved from Eq. (1) [(2)]. Because of the overlap function, the lidar signals start from 0.6 km MSL.

3. Interpretation of the UV polarization measurements

The presence of volcanic ash particles above the lidar station is confirmed by analyzing 7-day FLEXTRA airmass back trajectories (Stohl and Seibert 1998) and potential temperature vertical profiles, as shown in Fig. 2. Further evidence of volcanic ash presence above Lyon is given by the FLEXPART ash numerical dispersion model. Comparison with FLEXPART is beyond the scope of this paper and has been published elsewhere (Miffre et al. 2011). On 1200 UTC 18 April, back trajectories confirm the volcanic origin of the observed air masses at 4- and 5-km altitude, while below 3-km air masses originate from eastern Europe. The inversion temperature occurring at 3 km hence separates two air masses of different origins. The presence of a residual layer containing ash particles at altitudes below 3 km cannot be excluded, but it is not evidenced by airmass back trajectories either. Hence, we will consider that the volcanic cloud base is located above 3 km. At 1800 UTC 19 April, the inversion temperature observed at 1.7 km delimits the volcanic ash cloud (located at the upper altitudes) from the local residual layer (observed at lower altitudes). As a consequence, the Nash determination proposed in section 4 will be limited to altitudes above 3 km MSL at 1200 UTC 18 April (above 1.7 km MSL at 1800 UTC 19 April).

Fig. 2.

FLEXTRA 7-day airmass back trajectories arriving at Lyon at (top) 1200 UTC 18 Apr and (bottom) 1800 UTC 19 Apr. Corresponding potential temperature θ and relative humidity RH are derived from ECWMF.

Fig. 2.

FLEXTRA 7-day airmass back trajectories arriving at Lyon at (top) 1200 UTC 18 Apr and (bottom) 1800 UTC 19 Apr. Corresponding potential temperature θ and relative humidity RH are derived from ECWMF.

a. Volcanic ash particle depolarization ratio δash

As shown by Mishchenko et al. (2002), the δa ratio is an intrinsic property of particulate matter, which is mainly governed by the aerosol’s shape. Hence, volcanic ash particles themselves depolarize laser light at a rate determined by their intrinsic lidar depolarization ratio δa(ash) = δash = βash,⊥ash,//. The aim of this section is to determine δash, the intrinsic depolarization ratio of ash particles.

In the literature, intrinsic depolarization ratios have only been determined for desert dust particles as a consequence of several field campaigns (Shimizu et al. 2004; Tesche et al. 2009). Like volcanic ash particles, dust particles are highly irregularly shaped, and hence exhibit an intrinsic dust particles depolarization ratio. During the above-cited field campaigns, the intrinsic depolarization of dust particles was directly determined from lidar δa measurements. Such a methodology cannot be applied to our specific case, because we are far from the source region where depolarizing particles are the only particles present in the aerosol cloud, as a consequence of advection. Hence, δa differs from δash as will be shown in the next paragraph (see section 3b). To retrieve the ash particle intrinsic depolarization ratio δash, it is necessary to specifically address volcanic ash particles, and this is feasible by using laboratory measurements on randomly oriented volcanic ash particles performed by Muñoz et al. (2004) as recognized reference.

The formal link between lidar and δash laboratory measurements is given by the scattering matrix theory. By introducing the normalized scattering matrix is the scattering cross section per ash particle, averaged over the PSD, and is the scattering matrix, relating the Stokes parameters of the incident and backscattered light), δa can then be expressed as a function of and (Mishchenko et al. 2002) as

 
formula

As shown by scanning electron microscope images (Muñoz et al. 2004), the ash particle morphology is different for different volcanic eruptions. Hence, δash may be influenced by the variability in the ash particle morphology. However, a striking feature of Muñoz et al.’s laboratory measurements is that the measured scattering matrices for the distinct samples taken from different volcanoes are remarkably similar, despite observed variability in the ash particle morphology. This similarity in the measured scattering matrices elements justifies the construction of a synthetic scattering matrix, which corresponds to samples taken from different volcanoes. Hence, to include the variability in the ash particle morphology, we used Muñoz et al.’s synthetic scattering matrix optical measurements. Moreover, this synthetic scattering matrix extends up to the lidar backward scattering direction. We hence derived a1 = 0.283 ± 0.015 and a2/a1 = 0.423 ± 0.030 at λ = 633 nm, corresponding to δash = (40.5 ± 2.0)%. For λ = 355 nm, because of the increased size parameters, δash might be slightly lower (O. Muñoz 2011, personal communication). A quantitative estimation of this effect is, however, very difficult, because it has neither been measured nor numerically simulated for volcanic ash particles. Hence, this uncertainty is difficult to evaluate. However, it seems to be already included in our error bars, because δash mainly depends on the shape rather than on the laser wavelength. Moreover, section 3b shows that, within our error bars, the retrieved 40.5% δash laboratory value allows interpretation of our UV lidar measurements.

The δash value may be influenced by atmospheric aging, sedimentation processes, and possible water uptake. As can be seen in Fig. 3, where different PSDs are analyzed, Muñoz et al.’s (2004) PSD is representative of atmospheric volcanic ash particles after long-range transport. This is not surprising because their samples were mechanically sieved to remove the largest particles, a situation that resembles to volcanic ash particles after long-range transport, once larger particles (typically for sizes larger than 10 μm) have been removed from the volcanic cloud by gravitational settling. To quantitatively estimate the effect of sedimentation processes on δash, numerical simulations seem to be necessary, because after long-range transport, the measured PSDs do not provide ash particle specificity (as shown by Schumann et al. 2011). Such numerical simulations are, however, not available in the literature on volcanic ash particles, and to develop this specific point is beyond the scope of this paper. In addition, water uptake may also influence the retrieved δash value because water uptake can be high in a volcanic cloud (Carrico et al. 2003). However, specific ash particles practically do not contribute to water uptake, as first shown by Delmelle et al. (2005), and as recently confirmed by Lathem et al. (2011) for the specific case of the Eyjafjallajökull volcanic eruption. Quantitatively, it is found that for RH = 90%, the hygroscopic growth for ash particles is between 2% and 5% (ash particles are found to be 35 times less hygroscopic than sulfates). Because the a2/a1 ratio is not affected by such mixing (Mishchenko et al. 2004), low ash particle hygroscopicity will not affect, within our error bars, the retrieved δash value, which brings robustness to our δash retrieval methodology.

Fig. 3.

Normalized volcanic particles size distribution dn/dLog(r) as a function of r effective radius (μm) derived from Muñoz et al. (2004) (full line with circles), Schumann et al. 2011 (dashed lines with triangles), and Ilyinskaya et al. (2011) (dashed dotted lines with squares). Muñoz et al.’s ash particle PSD was derived from samples collected near the ground close to the volcano, mechanically sieved to remove the largest particles. Ilyinskaya et al.’s PSD was derived at 15 km from the Eyjafjallajökull’s volcano. Schumann et al.’s PSD was evaluated in the atmosphere above Leipzig (northeast Germany) after long-range transport from the Eyjafjallajökull Icelandic volcano.

Fig. 3.

Normalized volcanic particles size distribution dn/dLog(r) as a function of r effective radius (μm) derived from Muñoz et al. (2004) (full line with circles), Schumann et al. 2011 (dashed lines with triangles), and Ilyinskaya et al. (2011) (dashed dotted lines with squares). Muñoz et al.’s ash particle PSD was derived from samples collected near the ground close to the volcano, mechanically sieved to remove the largest particles. Ilyinskaya et al.’s PSD was derived at 15 km from the Eyjafjallajökull’s volcano. Schumann et al.’s PSD was evaluated in the atmosphere above Leipzig (northeast Germany) after long-range transport from the Eyjafjallajökull Icelandic volcano.

b. Interpretation of the UV depolarization measurements

Within our error bars, the Fig. 1-observed δa values exhibit strong variations, from a few percent to 40%. The reference δash value is observed on the δa vertical profile at 1200 UTC 18 April at 4.7 km MSL, where the atmosphere mainly contains ash particles. If the volcanic cloud was only composed of ash particles, then δa would be equal to δash everywhere the volcanic cloud is present. After long-range transport, nonash (nash) particles are present in the volcanic cloud and are likely to be hydrated sulfates (Mather et al. 2003; Schumann et al. 2011). Such small-sized particles scatter light efficiently with negligible depolarization (because of their almost spherical shape), thus increasing βa,// while preserving βa,⊥, which lowers δa. Because βa,// is sensitive to both ash and nash particles (βa,// = βash,// + βnash,//), while βa,⊥ = βash,⊥ (assuming negligible depolarization from nash particles), δa is

 
formula

Hence, δa is effectively below δash in the presence of nash particles, in agreement with our lidar observations. Consequently, far from the source region, δash cannot be directly retrieved from the lidar δa measurements without care. Moreover, the βa,⊥ vertical profile, which is specific to nonspherical volcanic ash particles, may serve as a tracer for volcanic ash particles better than δa.

4. Application to volcanic ash number concentration retrieval

In this section, the βa,⊥ volcanic ash particle tracer enhanced in section 3 is used to develop a new methodology to remotely access to the volcanic ash particles number concentration at altitude z MSL. Such discrimination between ash and nonash particles is new. In the literature, it has only been done for separating dust from nondust particles (Shimizu et al. 2004). Moreover, Eqs. (5) and (6) below are new, and the volcanic ash number concentration proposed methodology is new. In particular, by construction, it is ash particle specific.

a. Methodology

By combining our lidar depolarization measurement with Muñoz et al.’s (2004) laboratory measurements, we retrieve ash particle backscattering coefficient at altitude z

 
formula

To correct for a possible depolarization ratio δnash from small-sized nonash particles, we expressed βash,⊥ as a function of δnash and δash,

 
formula

so that βash,⊥ = βa,⊥ only when δnash = 0, as stated in section 3. Then, the volcanic ash mean number concentration Nash is derived at altitude z from the βash definition

 
formula

where βash is derived from Eqs. (5) and (6) and 〈(dσ/dΩ)ash〉 is the volcanic ash backscattering cross section averaged over the normalized PSD. Hence, for accurate Nash evaluation, accurate lidar measurements of the perpendicular backscattering coefficient βash,⊥ are necessary, in combination with precise determination of 〈(dσ/dΩ)ash〉, δash, and δnash.

Equation (6) describes the quantitative impact of the nonash particle depolarization ratio δnash on the βash backscattering coefficient: assuming zero nonash depolarization leads to a maximum overestimation of βash,⊥ equal to δnash βa,//. In the main volcanic layers (above 3 km), βa,// is low and background measurements performed before the studied volcano episode exhibited a 1% δnash value. Because the weather conditions remained stable with clear-sky conditions at Lyon during the volcanic episode, no supplementary depolarization, for example, resulting from clouds, can be assumed, so that δnash is still 1% during the volcano observation. Below 3 km, because of the particles hygroscopicity, δnash is also 1% (Sakai et al. 2010).

b. Volcanic ash backscattering cross section

As shown by Mishchenko et al. (2002), the size-weighted backscattering cross section per ash particle only depends on the normalized scattering phase function a1 and on 〈Csca〉 for polarized incident light,

 
formula

To compute 〈Csca〉, we assumed ash particles to behave like projected surface area–equivalent spheres. According to Mishchenko et al. (2002), p. 294, negligible error is done compared to the spheroid model for aspect ratios of 1.5, observed by Riley et al. (2003) during Mont Spurr’s volcanic eruption, which revealed similar chemical composition to that of Eyjafjallajökull (http://www.earthice.hi.is). The parameters used for the 〈Csca〉 computation are all detailed in Table 1. We used Winchester’s (1998) accurate determination of the ash complex 355-nm refractive index, as given by Muñoz et al. (2004). In addition, an ash PSD is needed. Ash particle PSD is changing during the explosive and ascending phases of the volcanic eruption, while complex physical and chemical processes are occurring, as pointed out by several authors (Schumann et al. 2011; Delmelle et al. 2005). During advection, the ash PSD undergoes several modifications with possible scavenging, sedimentation processes, and water adsorption on the ash particle surface (Lathem et al. 2011). Because these processes are complex, a quantitative in situ observation of the change in the volcanic ash PSD has never been reported in the literature from the source region down to the observation region. Figure 3 represents the normalized PSD dn/dLog(r) as a function of the effective size radius r in micrometers. Different volcanic particles size distributions are presented, derived from the up-to-date literature on Eyjafjallajökull, either close to the source region (Ilyinskaya et al. 2011) or after long-range transport (Schumann et al. 2011). We also added Muñoz et al.’s ash PSD as a recognized reference for volcanic ash particles. As shown in Fig. 3, all PSD are similar. The corresponding 〈Csca〉 values are then derived by numerical simulation, by assuming that ash particles behave like projected surface area–equivalent spheres,

 
formula
Table 1.

Volcanic ash particle complex 355-nm refractive index (Winchester 1998) and double lognormal size distribution (radius r, geometrical mean σ, number concentration N) derived from Mount Spurr’s eruption analysis by Muñoz et al. (2004). The m uncertainty is taken from Winchester (1998) (Re{m} varies between 1.52 and 1.56, while Im{m} varies between 0.0048 and 0.006 at 355 nm). Eyjafjallajökull’s PSD proposed by Ilyinskaya et al. (2011) or Schumann et al. (2011) are similar to Spurr’s PSD (Muñoz et al. 2004); the latter was chosen, to be ash specific.

Volcanic ash particle complex 355-nm refractive index (Winchester 1998) and double lognormal size distribution (radius r, geometrical mean σ, number concentration N) derived from Mount Spurr’s eruption analysis by Muñoz et al. (2004). The m uncertainty is taken from Winchester (1998) (Re{m} varies between 1.52 and 1.56, while Im{m} varies between 0.0048 and 0.006 at 355 nm). Eyjafjallajökull’s PSD proposed by Ilyinskaya et al. (2011) or Schumann et al. (2011) are similar to Spurr’s PSD (Muñoz et al. 2004); the latter was chosen, to be ash specific.
Volcanic ash particle complex 355-nm refractive index (Winchester 1998) and double lognormal size distribution (radius r, geometrical mean σ, number concentration N) derived from Mount Spurr’s eruption analysis by Muñoz et al. (2004). The m uncertainty is taken from Winchester (1998) (Re{m} varies between 1.52 and 1.56, while Im{m} varies between 0.0048 and 0.006 at 355 nm). Eyjafjallajökull’s PSD proposed by Ilyinskaya et al. (2011) or Schumann et al. (2011) are similar to Spurr’s PSD (Muñoz et al. 2004); the latter was chosen, to be ash specific.

For the 〈Csca〉 calculation, water uptake on ash particles should be considered only for relative humidity approaching condensation levels. In this case, water clouds will interfere, which does not occur under clear-sky conditions. The effect of atmospheric aging on the ash PSD is seen by comparing Ilyinskaya et al.’s (2011) and Schumann et al.’s (2011) PSD, under the assumption that Schumann et al.’s PSD is ash specific, as is Ilyinskaya et al.’s. For the Nash retrieval, Muñoz et al.’s (2004) PSD was hence chosen to be both ash particle specific (instead of Schumann et al.’s PSD) and representative of long-range transport (instead of Ilynskaya et al.’s PSD), as detailed at the end of section 3a. Using Eq. (8), we obtain 〈(dσ/dΩ)ash〉 = (3.69 ± 0.55) × 10−10 cm2 sr−1 per ash particle.

Sedimentation processes may influence this result. The quantitative role of sedimentation processes has been intensively studied by Schumann et al. (2011) who discussed how the ash particle size distribution has been modified by sedimentation processes during the Eyjafjallajökull eruption. It follows that, for a given plume age, sedimentation processes act as a low-pass size filter for the ash particle size distribution so that “the plumes of ages larger than 2 days should be free of particles for diameter larger than 15 μm due to sedimentation” (p. 2270). The cutoff radius is determined by the square root of the fall distance [see Schumann et al.’s Eq. (3)]. From this equation, we determined the modification of the ash particle size distribution at altitude z and computed 〈Csca〉 as a function of altitude z within the ash layer to include the sedimentation processes. Table 2 presents the retrieved cutoff radius and corresponding 〈Csca〉 values for our two case studies. Sedimentation processes induce less than 4% variations on retrieved 〈Csca〉 values, leading to the same relative uncertainty on Nash [see Eqs. (7) and (8)]. We included this 〈Csca〉 variation as a function of altitude z in our Nash retrieval methodology.

Table 2.

Influence of sedimentation processes on the ash particle size distribution (cutoff radius) and corresponding 〈Csca〉 values at different altitudes for our two case studies within the ash layer. The cutoff radius has been calculated by assuming a 10-km plume top height (Schumann et al. 2011).

Influence of sedimentation processes on the ash particle size distribution (cutoff radius) and corresponding 〈Csca〉 values at different altitudes for our two case studies within the ash layer. The cutoff radius has been calculated by assuming a 10-km plume top height (Schumann et al. 2011).
Influence of sedimentation processes on the ash particle size distribution (cutoff radius) and corresponding 〈Csca〉 values at different altitudes for our two case studies within the ash layer. The cutoff radius has been calculated by assuming a 10-km plume top height (Schumann et al. 2011).

c. Volcanic ash number concentration retrieval

From Eqs. (5)(7), we derived vertical profiles of βash and Nash, as displayed in Fig. 4, for δnash = 0 and δnash = 1% at altitudes where the volcanic cloud is present, as discussed at the beginning of section 3. The Nash vertical profile differs from the δa vertical profile, because the βa,⊥ lidar measurement is nonspherical particles specific. At 1200 UTC 18 April, several layers of volcanic ash particles are hence emphasized. This new Nash retrieval methodology reveals the dispersion behavior of the volcanic cloud. Quantitatively, at 1800 UTC 19 April, the ash number concentration reaches a few tens particles per cubic centimeter. Because in situ measurements specific to volcanic ash particles do not exist in the literature, it is difficult to provide a correlative measurement. During the Eyjafjallajökull eruption, the measurements performed by Schumann et al. (2011) after long-range transport appear to be the most appropriate reference, because the air masses that they studied in the northeastern part of Germany passed over Lyon a few hours later, as confirmed by Fig. 2 back trajectories. Schumann et al. (2011) distinguished three size classes (10–160 nm, 0.25–1.0 μm, and >1.5 μm). He noticed that the first size class (10–160 nm) did not contain ash particles but was composed of secondary particles, such as ammonium sulfates. Hence, to compare our retrieved Nash concentrations with those of Schumann et al., we only considered the following two size classes—0.25–1.0 μm and >1.5 μm—by adding their measured number concentration in each independent size class. In the volcanic ash layer, we hence retrieve a mean value of 10 parts per cubic centimeter, averaged over altitudes at which the volcanic cloud is present. On 19 April, in the same ash layer, our mean Nash value (9.0 ± 3.0) part cm−3 agrees with Schumann et al.’s mean value, within our error bar. Our Nash concentration is below that of Schumann et al.’s because our new Nash retrieval methodology is volcanic ash particle specific. Hence, comparison with optical particle counters (Schumann et al. 2011) should be done with care. To become quantitative, further specific in situ comparison measurements at the lidar location are needed. It would hence be interesting to simultaneously apply the traditional Nash retrieval methodology described by Mueller et al. (2001) together with ours at the lidar location, hence using multiwavelength Raman lidar measurements. However, this was not feasible at Lyon for the two studied cases, and will hence be the subject of the next paper.

Fig. 4.

Vertical profiles of volcanic ash particle backscattering coefficient βash and ash particle number concentration Nash on at (top) 1200 UTC 18 Apr and (bottom) 1800 UTC 19 Apr for δnash = 0 (dotted lines) and δnash = 1% (solid line).

Fig. 4.

Vertical profiles of volcanic ash particle backscattering coefficient βash and ash particle number concentration Nash on at (top) 1200 UTC 18 Apr and (bottom) 1800 UTC 19 Apr for δnash = 0 (dotted lines) and δnash = 1% (solid line).

At altitude z MSL, the uncertainty on the retrieved Nash concentration depends on the uncertainty on our lidar measurements, on the 〈Csca〉 computation uncertainty, and on the depolarization ratios δash and δnash. Thanks to our sensitive and accurate polarization measurement, the error bar on Nash slightly depends on the uncertainty on the βa,⊥ lidar measurement. For the 〈Csca〉 computation uncertainty, use of Schumann et al.’s (2011) PSD instead of Muñoz et al.’s (2004) PSD would lead to a 3% lower Nash value, and a 90% relative humidity will lead to a 4% decrease on Nash. Use of T-matrix computation instead of the assumed projected surface area–equivalent spheres approximation will further reduce the 〈Csca〉 computation uncertainty. The uncertainty on δashash = (40.5 ± 2.0)%] leads to a 3.5% relative error on Nash, which is included in Fig. 4. The effect of a possible bias and higher uncertainty in the δash laboratory value can be evaluated by using Eq. (5). The exact value of a possible bias is difficult to determine, but to fix the way the errors propagate, one can note that the relative error on Nash grows to 6.7% for δash = (50.0 ± 5.0)%. Such a 10% bias appears to be high under our operating conditions [the use of accurate Muñoz et al. (2004) laboratory scattering measurements are a very low influence of atmospheric aging]. This confirms, however, that the proposed methodology can only be applied if accurate laboratory scattering matrix measurements are available. In complement to this approach, numerical simulations using T-matrix computations might be interesting to develop for precise evaluation of δash. However, the complexity resulting from the numerical simulation of randomly oriented vesicular and nonvesicular ash particles (Lindqvist et al. 2011) may lead to higher bias and uncertainties in the retrieved Nash number concentration. Finally, within our error bars, Fig. 4 Nash vertical profiles are slightly dependent on the exact δnash value at altitudes where the volcanic cloud is present. Hence, our new Nash retrieval methodology is volcanic ash particle specific and very robust; it is practically not affected by possible nonash particle depolarization.

5. Conclusions and outlook

In this paper, we showed that from the interpretation of accurate UV depolarization measurements, it is possible to retrieve a new methodology, giving access to the volcanic ash backscattering coefficient and, as an application, to their mean number concentration in the troposphere. Our sensitive and accurate UV depolarization measurements enable discrimination between nonspherical volcanic ash particles and nonash particles in the volcanic cloud. Hence, as shown by careful analysis of the δnash value, our methodology is ash particle specific. Moreover, the volcanic ash specificity relies on laboratory measurements from Muñoz et al. (2004), which are used as a reference to retrieve the volcanic ash particle depolarization ratio δash = (40.5 ± 2.0)%. A key point of this contribution is that δa generally differs from δash, because nonash particles are present in the volcanic cloud, especially after long-range transport. It follows that δash cannot be directly determined from δa lidar measurements. Within our error bars, our measured depolarization ratios in the volcanic cloud compare well with Muñoz et al.’s laboratory measurements, which validate the use of their scattering matrix. In addition, numerical simulations (spheroid models) can represent another complementary reference for deriving δash. Finally, vertical profiles of ash mean number concentration have been provided by combining our βa,⊥ lidar measurements with laboratory measurements and the computed ash particle optical backscattering cross section. This methodology is new and provides Nash values that are remarkably in the same order of magnitude as that in the literature (Schumann et al. 2011), while the Nash uncertainty could be further reduced by applying the T matrix (Mishchenko 2002). Other approaches, based on multiwavelength measurements and algorithms, have proven efficiency in retrieving the volume PSD (Veselovskii et al. 2010). However, our single wavelength approach, based on the 355-nm wavelength, is now directly applicable to most of the lidar stations. Next, we plan to calibrate our methodology with another instrument. Moreover, we may apply this new methodology to dust particles, whose episodes should occur more and more, as a consequence of climate change.

Acknowledgments

The authors thank O. Muñoz for fruitful discussions, A. M. Fjaeraa, N. I. Kristiansen, and A. Stohl for providing the back-trajectories, and Région Rhône-Alpes for a scientific grant.

REFERENCES

REFERENCES
Alvarez
,
J.
, and
Coauthors
,
2006
:
Calibration technique for polarization-sensitive Lidars
.
J. Atmos. Oceanic Technol.
,
23
,
683
699
.
Ansmann
,
A.
, and
Coauthors
,
2010
:
The 16 April 2010 major volcanic ash plume over central Europe: EARLINET lidar and AERONET photometer observations at Leipzig and Munich, Germany
.
Geophys. Res. Lett.
,
37
,
L13810
,
doi:10.1029/2010GL043809
.
Carrico
,
C. M.
,
P.
Kus
,
M. J.
Rood
,
P. K.
Quinn
, and
T. S.
Bates
,
2003
:
Mixtures of pollution, dust, sea salt, and volcanic aerosol during ACE-Asia: Radiative properties as a function of relative humidity
.
J. Geophys. Res.
,
108
,
8650
,
doi:10.1029/2003JD003405
.
Delmelle
,
P.
,
F.
Villiéras
, and
M.
Pelletier
,
2005
:
Surface area, porosity and water adsorption properties of fine volcanic ash particles
.
Bull. Volcanol.
,
67
,
160
169
.
Ilyinskaya
,
E.
,
V. I.
Tsanev
,
R. S.
Martin
,
C.
Oppenheimer
,
J.
Le Blond
,
G. M.
Sawyer
, and
M. T.
Gudmundsson
,
2011
:
Near source observations of aerosol size-distribution in the eruptive plumes from Eyjafjallajökull volcano, March-April 2010
.
Atmos. Environ.
,
45
,
3210
3216
.
Kaaden
,
N.
, and
Coauthors
,
2009
:
State of mixing, shape factor, number size distribution, and hygroscopic growth of the Saharan anthropogenic and mineral dust aerosol at Tinfou, Marocco
.
Tellus
,
61B
,
51
63
.
Klett
,
J.
,
1985
:
Lidar inversion with variable backscatter/extinction ratios
.
Appl. Opt.
,
24
,
1638
1643
.
Lathem,
T. L.
,
P.
Kumar
,
A.
Nenes
,
J.
Dufek
,
I. N.
Sokolik
,
M.
Trail
, and
A.
Russell
,
2011
:
Hygroscopic properties of volcanic ash
.
Geophys. Res. Lett.
,
38
,
L11802
,
doi:10.1029/2011GL047298
.
Lindqvist
,
L.
,
T.
Nousiainen
,
E.
Zubko
, and
O.
Muñoz
,
2011
:
Optical modeling of vesicular ash particles
.
J. Quant. Spectrosc. Radiat. Transfer
,
112
,
1871
1880
.
Mather
,
T. A.
,
D. M.
Pyle
, and
C.
Oppenheimer
,
2003
:
Tropospheric volcanic aerosol. Volcanism and the Earth’s Atmosphere, Geophys. Monogr., Vol. 139, Amer. Geophys. Union, 189–212
.
Miffre
,
A.
,
G.
David
,
B.
Thomas
, and
P.
Rairoux
,
2011
:
Atmospheric non‐spherical particles optical properties from UV‐polarization lidar and scattering matrix. Geophys. Res. Lett., 38, L16804, doi:10.1029/2011GL048310
.
Mishchenko
,
M. I.
,
L. D.
Travis
, and
A. A.
Lacis
,
2002
:
Scattering, Absorption and Emission of Light by Small Particles. 3rd ed. Cambridge University Press, 426 pp
.
Mishchenko
,
M. I.
, and
Coauthors
,
2004
:
Scattering and radiative properties of semi-external versus external mixtures of different aerosols types
.
J. Quant. Spectrosc. Radiat. Transfer
,
88
,
139
147
.
Mueller
,
D.
,
U.
Wandinger
,
D.
Althausen
, and
M.
Fiebig
,
2001
:
Comprehensive particle characterization from three-wavelength Raman-lidar observations
.
Appl. Opt.
,
40
,
4863
4869
.
Muñoz
,
O.
,
H.
Volten
,
J. W.
Hovenier
,
B.
Veihelmann
,
W. J.
van der Zande
,
L. B. F. M.
Waters
, and
W. I.
Rose
,
2004
:
Scattering matrices of volcanic ash particles of Mount St. Helens, Redoubt, and Mount Spurr volcanoes
.
J. Geophys. Res.
,
109
,
D16201
,
doi:10.1029/2004JD004684
.
Ramaswamy
,
V.
, and
Coauthors
,
2001
:
Stratospheric temperature trends: Observations and model simulations
.
Rev. Geophys.
,
39
,
71
122
.
Ravishankara
,
A. R.
,
1997
:
Heterogeneous and multiphase chemistry in the troposphere
.
Science
,
278
,
1058
1065
.
Riley
,
C. M.
,
W. I.
Rose
, and
G. J. S.
Bluth
,
2003
:
Quantitative shape measurements of distal volcanic ash
.
J. Geophys. Res.
,
108
,
2504
,
doi:10.1029/2001JB000818
.
Sakai
,
T.
,
T.
Nagai
,
Y.
Zaizen
, and
Y.
Mano
,
2010
:
Backscattering linear depolarization ratio measurements of mineral, sea-salt, and ammonium sulfate particles simulated in a laboratory chamber
.
Appl. Opt.
,
49
,
4441
4449
.
Sassen
,
K.
,
J.
Zhu
,
P.
Webley
,
K.
Dean
, and
P.
Cobb
,
2007
:
Volcanic ash plume identification using polarization lidar: Augustine eruption, Alaska
.
Geophys. Res. Lett.
,
34
,
L08803
,
doi:10.1029/2006GL027237
.
Schumann
,
U.
, and
Coauthors
,
2011
:
Airborne observations of the Eyjafjalla volcano ash cloud over Europe during air space closure in April and May 2010
.
Atmos. Chem. Phys.
,
11
,
2245
2279
.
Shimizu
,
A.
, and
Coauthors
,
2004
:
Continuous observations of Asian dust and other aerosols by polarization lidars in China and Japan during ACE-Asia
.
J. Geophys. Res.
,
109
,
D19S17
,
doi:10.1029/2002JD003253
.
Stohl
,
A.
, and
P.
Seibert
,
1998
:
Accuracy of trajectories as determined from the conservation of meteorological tracers
.
Quart. J. Roy. Meteor. Soc.
,
124
,
1465
1480
.
Tesche
,
M.
,
A.
Ansmann
,
D.
Müller
,
D.
Althausen
,
R.
Engelmann
,
V.
Freudenthaler
, and
S.
Groß
,
2009
:
Vertically resolved separation of dust and smoke over Cape Verde using multiwavelength Raman and polarization lidars during Saharan Mineral Dust Experiment 2008
.
J. Geophys. Res.
,
114
,
D13202
,
doi:10.1029/2009JD011862
.
Tsanev
,
V. I.
,
T. A.
Mather
, cited
2007
:
MicrotopsII inverse. NERC Field Spectroscopy Facility. [Available online at http://fsf.nerc.ac.uk/resources/software/MicrotopsInverse.shtml.]
Veselovskii
,
I.
, and
Coauthors
,
2010
:
Application of randomly oriented spheroids for retrieval of dust particle parameters from multiwavelength lidar measurements
.
J. Geophys. Res.
,
115
,
D21203
,
doi:10.1029/2010JD014139
.
Wang
,
X.
, and
Coauthors
,
2008
:
Volcanic dust characterization by EARLINET during Etna’s eruptions in 2001-2002
.
Atmos. Environ.
,
42
,
893
905
.
Winchester
,
L. W.
,
1998
:
Determination of the complex refractive index of samples of volcanic ash
.
NASA Contract Rep.
S01394
G
.
Winker
,
D. M.
, and
M. T.
Osborn
,
1992
:
Preliminary analysis of observations of the Pinatubo volcanic cloud with a polarization-sensitive lidar
.
Geophys. Res. Lett.
,
19
,
171
174
.