Estimating the volcanic emission rate and atmospheric lifetime of SO&lt;sub&gt;2&lt;/sub&gt; from space: a case study for Kīlauea volcano, Hawai`i

S. Beirle; C. Hörmann; M. Penning de Vries; S. Dörner; C. Kern; T. Wagner

doi:10.5194/acp-14-8309-2014

Estimating the volcanic emission rate and atmospheric lifetime of SO<sub>2</sub> from space: a case study for Kīlauea volcano, Hawai`i

Beirle, S.;Hörmann, C.;Penning de Vries, M.;Dörner, S.;Kern, C.;Wagner, T.;

2014-08-19 00:00:00

Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/ doi:10.5194/acp-14-8309-2014 © Author(s) 2014. CC Attribution 3.0 License. Estimating the volcanic emission rate and atmospheric lifetime of SO from space: a case study for Kılauea volcano, Hawai‘i 1 1,2 1 1 3 1 S. Beirle , C. Hörmann , M. Penning de Vries , S. Dörner , C. Kern , and T. Wagner Max-Planck-Institut für Chemie, Mainz, Germany Institut für Umweltphysik, Universität Heidelberg, Heidelberg, Germany USGS Cascades Volcano Observatory, Vancouver, Washington, USA Correspondence to: S. Beirle ([email protected]) Received: 20 September 2013 – Published in Atmos. Chem. Phys. Discuss.: 4 November 2013 Revised: 1 July 2014 – Accepted: 9 July 2014 – Published: 19 August 2014 Abstract. We present an analysis of SO column densi- SO is removed from the atmosphere by dry and wet de- 2 2 ties derived from GOME-2 satellite measurements for the position (in the boundary layer) or by chemical conversion Kılauea volcano (Hawai‘i) for 2007–2012. During a period to sulfuric acid (H SO ). In the gas phase, this conversion is 2 4 of enhanced degassing activity in March–November 2008, initiated by the OH radical. The respective SO lifetime in monthly mean SO emission rates and effective SO life- the troposphere is about 2 weeks (von Glasow et al., 2009). 2 2 times are determined simultaneously from the observed However, heterogeneous reactions on cloud droplets convert downwind plume evolution and meteorological wind ﬁelds, SO into H SO on much shorter timescales of days or even 2 2 4 without further model input. Kılauea is particularly suited hours (von Glasow et al., 2009). Empirically derived SO de- for quantitative investigations from satellite observations ow- pletion rates thus differ by several orders of magnitude, de- ing to the absence of interfering sources, the clearly deﬁned pending on whether homogeneous or heterogeneous reac- downwind plumes caused by steady trade winds, and gener- tions are predominant (Oppenheimer et al., 1998). ally low cloud fractions. For March–November 2008, the ef- Kılauea volcano, located on Hawai‘i Island (19.4 N, fective SO lifetime is 1–2 days, and Kılauea SO emission 155.3 W; 1247 m a.s.l.), has shown persistent effusive 2 2 −1 rates are 9–21 kt day , which is about 3 times higher than SO degassing for over more than 3 decades. A period of initially reported from ground-based monitoring systems. particularly high gas emissions began in early 2008 with the lead-up and opening of a new vent within the Halema‘uma‘u summit crater. A detailed overview about the dates, loca- tions, speciﬁcation of volcanic activity, and ground-based 1 Introduction SO emission rate estimates is provided by Elias and Sut- ton (2012). As Kılauea is located remotely from other Sulfur dioxide (SO ) plays an important role in the Earth’s SO sources and within the trade wind zone, it can be con- climate, as it is an important precursor of aerosols, which af- sidered a “natural laboratory” (Yuan et al., 2011), well suited, fect the planet’s radiative budget both directly and indirectly, for instance, for the investigation of aerosol indirect effects e.g., by inﬂuencing the number and size of cloud droplets from satellite observations (Yuan et al., 2011; Eguchi et al., (Robock, 2000, and references therein). 2011). Volcanoes are a large natural source of SO with high tem- During recent decades, methods for the quantiﬁcation of poral and spatial ﬂuctuations, and total emissions are still volcanic SO emissions from spectroscopic measurements highly uncertain (Andres and Kasgnoc, 1998). Consequently, have been developed and reﬁned (e.g., Moffat and Millan, the impact of volcanic emissions on aerosol radiative forcing 1971; Galle et al., 2002; Mori and Burton, 2006), and several is one of the key uncertainties in climate models (Carslaw volcanoes are now continuously monitored by such instru- et al., 2013). ments (e.g., Galle et al., 2010; Elias and Sutton, 2012). Published by Copernicus Publications on behalf of the European Geosciences Union. 8310 S. Beirle et al.: Emissions and lifetime of SO from Kılauea In addition to such ground-based measurements, satellite A ﬁfth-order polynomial was ﬁtted to account for broadband instruments have become available over the last decades, structures. In order to minimize nonlinear effects caused by providing global measurements of atmospheric trace gases, the strong absorption of O in the UV, the approach by among them SO , in particular the TOMS series, starting Puk ¸ ıte et al. (2010) has been implemented. in 1978 (e.g., Carn et al., 2003); UV–vis spectrometers like SCDs are converted into vertical column densities GOME, SCIAMACHY or OMI; and IR interferometers like (VCDs), i.e., vertically integrated concentrations, via so- TES, IASI or AIRS; for details on the different satellite in- called air mass factors (AMFs). AMFs are calculated us- struments and the respective references, see Martin (2008). ing the Monte Carlo radiative transfer model (RTM) McAr- These measurements have revolutionized our knowledge of tim (Deutschmann et al., 2011) under cloud-free conditions abundance, sources and transport of various pollutants over at 315 nm, for an albedo of 0.05, and different a priori the last decades in general (e.g., Martin, 2008; Monks and aerosol optical depths (AOD) of 0, 0.4 and 1, assuming non- Beirle, 2011, and references therein), and provide new po- absorbing aerosols (single-scattering albedo: 1; asymmetry tential for monitoring volcanic activity in particular. Several parameter: 0.85). Final VCDs are derived by interpolation studies have estimated burdens and ﬂuxes of SO from differ- according to the actual AOD as measured by MODIS (see ent volcanoes, e.g. Carn et al. (2003, 2005, 2008); Khokhar Sect. 2.2). The plume altitude of both SO and aerosols was et al. (2005); Krotkov et al. (2010); Monks and Beirle (2011) set to 2.0± 0.5 km (see Sect. 2.3). (see Table 8.2 and references therein). Reviews of the differ- The SO detection limit for the SCDs of individual 16 −2 ent methods applied for estimating SO ﬂuxes from satellite GOME-2 ground pixels was about 1× 10 molec cm in 16 −2 observations are provided by Carn et al. (2013) and Theys 2007 and increased steadily to about 2× 10 molec cm in et al. (2013) (see Sect. 4.1). 2011 due to instrument degradation (Hörmann et al., 2013). Recently, it has been demonstrated that lifetimes of trace For the given AMFs, this corresponds to a VCD detection 16 −2 gases can also be quantiﬁed by analyzing the downwind de- limit of 1.3–2.7×10 molec cm , or 0.5 to 1 Dobson units cay of point source emissions as observed from satellites (DU). (e.g., Leue et al. (2001) and Beirle et al. (2004, 2011) for The individual satellite observations are gridded on a regu- nitrogen oxides or Krotkov et al. (2010) for SO ). lar lat–long grid with 0.1 resolution, i.e., much ﬁner than the In this study we present an analysis of the downwind original GOME-2 ground pixel size. Only ground pixels with evolution of the SO plume from Kılauea, as derived from an effective cloud fraction below 20 % are considered, using GOME-2 (Callies et al., 2000). By combining the satellite the GOME-2 cloud product based on the FRESCO algorithm measurements with wind ﬁelds provided by the European (Wang et al., 2008). Subsequently, monthly mean maps are Centre for Medium-Range Weather Forecasts (ECMWF), we calculated. With temporal averaging, spatial gaps (which are demonstrate that an effective SO lifetime can be determined immanent in daily maps due to the GOME-2 swath width from a relatively simple and robust mathematical analysis. In and cloud screening) are closed, and the noise of individual addition, the SO emission rate from Kılauea is quantiﬁed satellite pixels is reduced. Finally, we apply an empirical off- and compared to ground-based estimates. set correction by subtracting the mean upwind VCD east of Hawai‘i (at 150–153 W) as we are interested in the increase of SO due to emissions from Kılauea. Figure 1 shows the gridded monthly mean SO VCDs for 2 Method 2007–2012. Enhanced SO column densities can be observed southwest of Hawai‘i Island during several months, espe- 2.1 SO from GOME-2 cially in 2008. For August 2008, a zoomed-in map is shown GOME-2, the second Global Ozone Monitoring Experiment in Fig. 2, providing additional information on the location of (Callies et al., 2000) was launched in October 2006 onboard Kılauea on Hawai‘i, elevation contour lines, and mean wind the MetOp-A satellite. It is operated in a Sun-synchronous directions for different altitude levels. In the following, we orbit, crossing the Equator at about 09:30 local time. Nomi- focus on the period March–November 2008 for our quantita- nal ground pixel size is 80 km× 40 km, and global coverage tive analysis. is attained every 1.5 days. SO concentrations integrated along the mean light path, 2.2 Aerosol and cloud effects referred to as SO slant column densities (SCDs), are derived from spectral GOME-2 measurements in the UV by means of differential optical absorption spectroscopy (DOAS) (Platt Satellite measurements are affected by aerosols and clouds and Stutz, 2008), as described in Hörmann et al. (2013). due to their inﬂuence on radiative transfer (RT). Aerosols and A ﬁt range between 312.1 and 324.0 nm was used, includ- clouds generally shield the troposphere below, but they in- ing cross sections for SO at 273 K (Bogumil et al., 2003) crease the satellite’s sensitivity to trace gases within or above and O at 223 K (Gür et al., 2005), as well as pseudo- the aerosol/cloud layer due to multiple scattering and the in- absorbers accounting for Raman scattering and stray light. creased albedo (e.g., Beirle et al., 2009; Leitao et al., 2010). Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/ S. Beirle et al.: Emissions and lifetime of SO from Kılauea 8311 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2.5 500 km 1.5 0.5 Figure 1. Monthly mean SO VCD 2007–2012. The upper end of the color scale corresponds to 10 DU. Each panel covers 161–154 W longitude and 16–20.5 N latitude (compare Fig. 2). GOME−2 ground pixel size&orientation Wind vectors 3.5−4.5 km 2.5−3.5 km 17 1.5−2.5 km 0.5−1.5 km Plume direction −161 −160 −159 −158 −157 −156 −155 −154 lon Figure 2. Monthly mean SO VCD for August 2008. Color bar as in Fig. 1. Kılauea on Hawai‘i Island is indicated by a triangle. The grey lines show surface elevation contours in 1 km intervals. The main plume direction is indicated by a grey arrow, while the mean ECMWF ◦ −1 wind vector for different altitudes is plotted in shades of green (a vector length of 1 corresponds to a velocity of 5 m s ). As transport on Hawai‘i Island is strongly affected by topography, the southwestern tip of the island was chosen as the origin. The size and orientation of an individual GOME-2 groud pixel is indicated in the legend. Here we describe the detailed treatment of aerosols and the monthly mean MODIS TERRA AOD (with a local over- clouds speciﬁed for the conditions at Kılauea. pass time similar to that of GOME-2), multiplied by 2 (corre- Aerosols are formed within the volcanic plume by con- sponding to an Ångström coefﬁcient of 1.24) to account for version of SO to H SO during formation of, or uptake in, the AOD wavelength dependency. We estimate the remaining 2 2 4 aqueous droplets. Thus, we assume the same vertical pro- uncertainty due to aerosol effects to be negligible (< 10 %). ﬁle for SO and aerosols. During the volcano’s active phase To minimize cloud effects, only observations with cloud in 2008, the aerosol optical depth (AOD) was signiﬁcantly fractions below 20 % are considered. The remaining cloud enhanced above background in the plume region over ocean, effects could in principle be corrected by radiative trans- reaching monthly mean values of 0.35 (compare Beirle et al., fer calculations, as long as the vertical proﬁles of both 2012) as measured by MODIS TERRA (at 550 nm). SO and clouds are accurately known. However, this is not The AMFs depend almost linearly on AOD and increase the case: the SO plume altitude has some uncertainty (see by about 20 % for an increase of AOD from 0 to 1 (at Sect. 2.3), and the cloud altitudes derived from satellite ob- 315 nm). For the calculation of the actual monthly mean servations have high uncertainties for low cloud fractions SO VCD, the AMFs for AOD 0, 0.4 and 1 are interpo- (see Fig. 4 in Koelemeijer et al., 2001). Thus, we decided to lated according to the “real” AOD. The latter is taken from consider the observations with cloud fractions below 20 % as www.atmos-chem-phys.net/14/8309/2014/ Atmos. Chem. Phys., 14, 8309–8322, 2014 2012 2011 2010 2009 2008 2007 lat 17 2 SO VCD [10 molec/cm ] 2 8312 S. Beirle et al.: Emissions and lifetime of SO from Kılauea – Additional, independent plume height information can 3.5−4.5 km be derived from the comparison of the plume direction 2.5−3.5 km with wind directions at different altitudes (e.g., Bluth 1.5−2.5 km 0.5−1.5 km and Carn, 2008; Hughes et al., 2012). The monthly mean maps for March–November 2008 reveal a clear SO outﬂow with a well-deﬁned direction. We deter- mine the mean plume direction by ﬁtting a line to the lat–long coordinates of those grid pixels exceeding 16 −2 3× 10 molec cm . For August, the resulting plume direction is displayed as a grey arrow in Fig. 2, while ECMWF operational analysis wind vectors at differ- ent altitudes are shown in shades of green. From this Mar Apr May Jun Jul Aug Sep Oct Nov comparison, an upper bound of the SO plume alti- Figure 3. Absolute deviation between the outﬂow direction of the tude can be derived. In particular, the August plume SO plume and the mean ECMWF wind direction for different alti- is clearly below 3 km, as the plume direction reveals tudes for March–November 2008. Wind directions above 2.5 km do a small southward component (consistent with ECMWF not match the observed movement of the plume. winds below 2.5 km), while ECMWF wind ﬁelds above 2.5 km show a northward component instead. For May– November 2008, ECMWF winds at 1.5–2.5 km reveal “cloud-free”, without further corrections. We justify this by the best agreement with the observed plume direction performing our analysis for different a priori thresholds for (Fig. 3). the effective cloud fraction. The dependency of the result- ing VCDs on the cloud fraction threshold turned out to be After taking the above considerations into account, we es- negligibly small (see Sect. 4). This indicates that the remain- timate that an effective plume altitude of between 1.5 and ing cloud effects (shielding vs. multiple scattering/albedo in- 2.5 km most accurately describes the plume’s location in crease) at least partly cancel out. March–November 2008. For the conversion of SCDs into VCDs, we thus calculate AMFs for a priori box proﬁles from 2.3 Plume altitude 1.5 and 2.5 km altitude for both SO and aerosols. As the plume height information derived from different, indepen- The SO plume altitude has a large impact on our analysis dent data sets is consistent, we estimate the uncertainty of via two different effects. First, the sensitivity of the satellite the mean plume height to be less than 0.5 km; a mean plume measurements (i.e., the AMF) depends on the trace gas ver- height below 1.5 km is unlikely, as it would be even lower tical proﬁle and generally decreases towards the ground for than the plume altitude close to the vent, while the plume has low albedo such as over ocean. Second, the horizontal wind high buoyancy. CALIOP measurements also indicate rather speed, needed for the lifetime estimate as explained below, a rising than a sinking plume. A mean plume height above depends on altitude as well. Thus, for the emission rate esti- 2.5 km, on the other hand, can be excluded due to the dis- mate, an accurate a priori plume altitude is needed. crepancy between wind and plume directions. We estimate the SO plume altitude from three indepen- Note, however, that all the considerations above refer to dent data sets: plumes originating from Kılauea summit vent for the strong degassing period from March 2008 onwards. SO emissions – Kılauea’s summit vent is located at about 1.1 km a.s.l., from the East Rift, on the other hand, are emitted at only and Halema‘uma‘u plume heights close to the vent are 0.7 km altitude and tend to be less buoyant, thus typically about 1.4–2.0 km (see Fig. 7 in Elias and Sutton, 2012). staying in the mid-MBL (A. J. Sutton, personal communi- Although this altitude range is generally within the ma- cation, 2013). Therefore, the satellite observations are gen- rine boundary layer (MBL) around Hawai‘i of approxi- erally less sensitive to East Rift emissions, and the VCDs mately 2 km (Cao et al., 2007), the plume from the sum- (based on AMFs derived for the summit vent) are thus bi- mit is generally buoyant enough to stay at the upper ased low. edge of the MBL or even break through the inversion at times (Elias and Sutton, 2012). 2.4 Determining SO emission rates and lifetimes – Eguchi et al. (2011) determine the Kılauea plume height ◦ ◦ to 1.4–2.0 km at 160 W, and to 1.6–3.0 km at 180 W, We investigate the downwind evolution of SO from based on the increase in aerosol extinction in July Kılauea based on monthly mean VCD maps in order to es- and August 2008 compared to 2007 as measured by timate SO emission rates and lifetimes via the following CALIOP. steps: Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/ abs(winddir−plumedir) [°] S. Beirle et al.: Emissions and lifetime of SO from Kılauea 8313 2. By multiplying the LDs by the longitudinal wind speed Nov u from ECMWF averaged over 1.5–2.5 km altitude, a longitudinal SO ﬂux is derived as a function of time 0 2 t since emission from the volcano. Figure 4 displays the observed SO ﬂux for March–November 2008 in red. Oct 2 3. SO lifetime and emission rates are derived simultane- ously by ﬁtting the model function F (t) to the observed SO ﬂux with a nonlinear least-squares algorithm, as- Sep suming steady state: −t/τ F (t) = E × e , (1) Aug with the emission rate E and the lifetime τ as ﬁt pa- rameters. In addition, F (t) is smoothed by a Gaussian with a standard deviation of σ = σ /u, where u is the t x monthly mean longitudinal wind speed in the plume and Jul σ is 80 km in order to account for the GOME-2 across- track ground pixel size. A similar approach was used by Beirle et al. (2011) to Jun estimate NO lifetimes and emissions from megacities. In the case of SO from Kılauea, however, some simpliﬁca- tions/modiﬁcations were possible/necessary: May – As the lifetime of SO is considerably longer than that of NO , tthe considered spatial and temporal scales are much larger (≈ thousand km, or hundred hours). Apr – Due to the steady trade winds, at least during summer, 5 a sorting of the observations by wind direction is not 0 necessary here. – As wind direction is stable and there are no interfer- Mar ing sources of SO , the background can directly be es- timated from upwind measurements, while it had to be included as a free ﬁt parameter in Beirle et al. (2011). −20 0 20 40 60 80 100 – In Beirle et al. (2011), an e-folding distance x is ﬁtted t [h] to the line densities as function of x, and the lifetime is then derived from x by division by the mean wind Figure 4. Measured SO ﬂux (i.e., latitudinally integrated VCD speed. In the current study, the wind speed can change times wind speed u) as function of time for March–November signiﬁcantly with distance from the volcano (as larger 2008 (red) and the ﬁtted exponential downwind decay according distances have to be considered). Thus, the downwind to Eq. (1) (black). Light red (April, October) indicates months in ﬂux is ﬁrst transferred into a function of time by variable which ECMWF winds turned westerly for at least one 6 h time step. Error bars in x and y reﬂect the statistical error of the mean SO ﬂux transformation via the local wind speeds (t = x/u). The and the statistical error of t deduced from ECMWF wind variability, subsequent ﬁt directly yields the effective lifetime τ . respectively. Note that the downwind reduction of the SO ﬂux with time shown in Fig. 4 in fact reﬂects the chemical conver- sion or depletion of SO , and is not caused by dilution of 1. The background-corrected monthly mean SO VCDs the plume, as the concentrations are integrated vertically (by ◦ ◦ are integrated in the latitudinal direction (10–25 N), the column measurement) and latitudinally (10–25 N). The resulting in “line densities” (LD) as a function of lon- outﬂow out of this area can be neglected, as it can easily be gitude. Note that the small southward component of the checked by extending the latitude range over which integra- main ﬂux, as well as effects of dilution in across-wind tion occurs. This had only a small impact on the results (see direction, are eliminated by the latitudinal integration. Sect. 4.2.3 and Table 2). www.atmos-chem-phys.net/14/8309/2014/ Atmos. Chem. Phys., 14, 8309–8322, 2014 SO flux [kilotons/day] 2 8314 S. Beirle et al.: Emissions and lifetime of SO from Kılauea The simultaneous ﬁt of SO lifetime and emission rate 80 40 as described above requires a well-deﬁned SO plume and steady easterly winds. Thus, we apply it for 60 30 the months March–November 2008 with the highest ob- served SO VCDs far above the detection limit. During this 40 20 period, the mean u component of ECMWF wind is nega- tive (easterly) for all 6-hourly time steps, except for Octo- 20 10 ber (one time step with westerly wind) and in particular for April (eight time steps with westerly wind). Consequently, 0 0 the background determined east of Hawai‘i is biased high in April, resulting in negative VCDs and ﬂuxes (for plume ages > 40 h) (compare Fig. 4). In addition to the ﬁtted SO lifetime and emission rates for this particular period, we also provide a rough emission rate estimate based on the monthly mean VCDs for the complete time series 2007–2012 (see Sect. 3). 3 Results 3 4 5 6 7 8 9 10 11 Month in 2008 Figure 4 displays the observed (red) and ﬁtted (black) down- Figure 5. Fitted monthly mean SO lifetimes τ (top) and emission wind evolution of the longitudinal SO ﬂux. The processes 2 rates E (bottom) for March–November 2008. Error bars indicate responsible for SO removal from the atmosphere, i.e., gas- the conﬁdence intervals derived from the least-squares ﬁt. In the phase reactions with OH and heterogeneous reactions on upper panel, the monthly mean cloud fraction (from GOME-2) is cloud droplets, have signiﬁcantly different time constants, also included, revealing an anticorrelation to τ (R = −0.76). and observed instantaneous loss rates of volcanic SO cover several orders of magnitude (Oppenheimer et al., 1998). Nev- ertheless, the observed monthly mean downwind loss of SO can be described by a single ﬁrst-order time constant. evolution of mean SO column densities is quantiﬁed over 2 2 Figure 5 shows the resulting monthly mean SO lifetimes time as opposed to simply taking a single snapshot. Never- and emission rates. theless, it is also possible that heterogeneous reactions on The derived SO lifetimes range from 16 to 57 h. They volcanic aerosols reduce the SO lifetime within the ﬁrst few 2 2 show a seasonal cycle and are anticorrelated to the monthly minutes after emission, and the lifetime for SO from East mean cloud fraction: lifetimes are highest in summer when Rift emissions might be generally shorter due to the lower cloud cover is smallest, and shorter for higher cloud fractions plume height (see Sect. 2.3). in spring and autumn. This anticorrelation is in accordance The ﬁtted monthly mean SO emission rates range from −1 with the impact of heterogeneous reactions on cloud droplets. 9 to 21 kt day . Integrated emissions from March to Octo- On average, we ﬁnd a mean SO lifetime of 1.56 days, which ber 2008 are 3.5 Tg (with an uncertainty of about 40 %; see is consistent with previous studies. For instance, Lelieveld Sect. 4), which is higher by a factor of 2 than the estimate of et al. (1997) give an average SO lifetime of 2 days, based 1.8± 1.2 Tg given by Eguchi et al. (2011), based on a com- on the general circulation model ECHAM. Lee et al. (2011) parison of SCIAMACHY observations to model simulations. derived mean lifetimes of 19± 7 h from in situ measurements This discrepancy is already visible in the monthly mean VCD over the eastern US in summer. For other degassing volca- (compare Fig. 2 with Fig. S1 in the supplement of Eguchi noes, mean lifetimes on the order of 1 day have been ob- et al., 2011) and probably caused by different retrieval set- served as well, e.g., Bluth and Carn (2008) (15–26 h for Nya- tings, most likely the assumed plume altitude. muragira) and McCormick et al. (2014) (19.6 h for Tungu- Figure 6 displays the derived emission rates in compari- rahua). son to the monthly mean SO VCD downwind of the vol- ◦ ◦ Our derived SO lifetimes are signiﬁcantly longer than the cano (averaged over 17–20 N, 155–160 W). A clear corre- 6 h (half-life) estimated by Porter et al. (2002) for the East lation can be seen (R = 0.92), which is expected based on Rift plume for 1 day of measurements. However, our values mass balance, as long as monthly mean lifetimes are com- are arguably more robust because (1) SO is measured di- parable. By assuming that the ﬁtted linear relation between rectly instead of indirectly deriving an aerosol mass from an emission rates and spatiotemporal mean column densities AOD, (2) the monthly mean composite reﬂects the average also holds for other months with lower SO column densi- plume over hundreds of kilometers after emission into the at- ties, emission rates can be estimated for the complete time mosphere instead of only the ﬁrst 9 km, and (3) the actual series 2007–2012 (see Fig. 7, where emission rates reported Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/ Emission rate [kt/day] Lifetime [h] Mean Cloud fraction S. Beirle et al.: Emissions and lifetime of SO from Kılauea 8315 vidual months. The assumption of a constant AMF, how- ever, can only be considered a rough estimate. During the period 2007–2010, the ratio of emission rates from the sum- 20 ¯ mit and from Pu‘u‘O‘o ¯ (East Rift) varied with the opening of the summit Overlook Vent and episodic unrest related to the ongoing East Rift eruption (Elias and Sutton, 2012). Due to the different emission altitudes of these two sources (which cannot be differentiated at the spatial resolution of the GOME-2 measurements), this variability is expected to in- ﬂuence the effective plume altitude, thereby modulating the sensitivity of the satellite measurements. This effect is partic- ularly apparent in the data collected prior to the 2008 summit vent opening, as is discussed in Sect. 4.4. 0 2 4 6 8 VCD [molec/cm ] x 10 Figure 6. Fitted monthly mean SO emission rates vs. the respec- 4 Discussion ◦ ◦ tive mean SO VCD (averaged over 17–20 N, 155–160 W) for March–November 2008. The correlation coefﬁcient is R = 0.92. 4.1 Estimating SO emission rates and lifetimes from The black line represents a linear ﬁt forced through origin. satellite observations Satellite measurements of SO provide valuable information 25 2 Fitted emission rate on volcanic emissions, and have been used to investigate vol- (downwind decay) canic activity since the late 1970s from TOMS. Strong explo- Upscaled emission rate sive volcanic eruptions can generally be well observed from (mean VCD) Summit emission rate space, particularly if the SO plume reaches the upper tropo- (ES12) sphere or even the stratosphere, where the satellites’ sensitiv- Rift emission rate ity is high and the plume is not shielded by clouds. The quan- (ES12) tiﬁcation of emission rates from volcanoes degassing into the lower troposphere, however, is often more difﬁcult. Different algorithms have been applied to estimate SO emission rates, and partly also SO lifetimes, from satel- 2 2 0 lite observations of various volcanoes. Reviews of such stud- Jan Jul Jan Jul Jan Jul Jan Jul Jan Jul Jan Jul ies are provided by Theys et al. (2013) and Carn et al. (2013). 2007 2008 2009 2010 2011 2012 Generally, one can discriminate four different approaches: Figure 7. Time series of Kılauea’s SO emission rate. Black: results 1. Methods based on mass balance from the monthly ﬁts (as in Fig. 5). Orange: emission rate estimates based on the monthly mean SO VCD, using the linear relation de- The SO emission rate is derived from the ratio of the 2 2 rived in Fig. 6. Values prior to the opening of the Kılauea sum- total amount of SO and the mean SO lifetime τ . For 2 2 mit vent in March 2008 are marked in light orange, as the up- this method, the complete volcanic plume has to be cap- scaled emission rates are very likely biased low (see text). Green tured, and a priori knowledge on τ is needed. and magenta: emission rate estimates by Elias and Sutton (2012) (labeled as ES12), derived from ground-based measurements, for 2. Methods based on mass ﬂuxes the Kılauea summit and East Rift, respectively. The emission rate is determined from the mass ﬂux through a deﬁned surface. This method requires a pri- ori information on the mean horizontal wind speed w, by Elias and Sutton (2012) are shown for comparison; see and might require a correction for SO decay as well if Sect. 4.4). the considered distances are comparable to w × τ . Note that this assumption of a linear relation between emission rates and mean column densities implies that monthly mean conditions like the SO AMFs, lifetime and 3. Methods based on temporal evolution wind speeds are comparable. According to mass balance In the case of explosive events, which release SO high (i.e., the spatially integrated mean VCD equals E × τ ), the in the stratosphere, the temporal evolution of the total ﬁtted slope in Fig. 6 corresponds to an effective SO life- integrated SO mass can be investigated directly, allow- 2 2 time of 2 days. This is in good agreement with the life- ing for a lifetime estimate and a subsequent emission times obtained from the plume evolution during the indi- rate estimate based on the plume age and τ . www.atmos-chem-phys.net/14/8309/2014/ Atmos. Chem. Phys., 14, 8309–8322, 2014 Emission rate [kt/day] Emission rate [kt/day] 8316 S. Beirle et al.: Emissions and lifetime of SO from Kılauea 4. Furthermore, more elaborate techniques like inverse Table 1. Baseline and alternative a priori settings for the calculation of AMFs and their impact on monthly mean VCD V for March– modeling might be required for complex scenarios November 2008. (multi-layered plumes, variable wind patterns). In this study, we apply a simultaneous lifetime and emis- A priori Baseline Alternatives 1V /V sion rate estimate optimized for Kılauea conditions, i.e., a a. Albedo 5 % continuously degassing volcano in a stable wind regime. Our 3 % 8 % method is related to previous methods in that it also assumes 7 % −8 % mass balance and that the SO loss over time can be de- b. Wavelength 315 nm scribed by a simple ﬁrst-order time constant. 313 nm 12 % As the Kılauea plume is located in the lower troposphere, c. Plume altitude 1.5–2.5 km daily satellite observations are not sufﬁcient, as they contain 1–2 km 19 % 2–3 km −14 % gaps due to clouds. This is overcome by calculating monthly means for almost cloud-free conditions. The proposed method has the following advantages: Table 2. Baseline and alternative a priori settings for the ﬁtting pro- cedure and their impact on emission rate and lifetime estimates for – No a priori SO lifetime is needed, as τ is directly ﬁtted March–November 2008. together with the emission rate E, and this τ actually re- ﬂects the appropriate effective mean lifetime according A priori Baseline Alternatives 1E/E 1τ/τ to mass balance (see Sect. 4.3). a. Cloud fraction 0.2 threshold 0.1 −4 % −11 % – The effects of dilution are accounted for by spatial inte- 0.3 −2 % 7 % gration of VCDs in across-wind direction. b. Latitude range 10–25 N 15–20 N 2 % −17 % – The full information on the SO ﬂux as a function of 2 ◦ 5–30 N −9 % 1 % time is used for the nonlinear least-squares ﬁt of the c. Time interval [−20,100] h [−10,70] h 2 % −1 % model function F (t). This is more robust than just tak- [−30,130] h 4 % −4 % ing the peak value at the volcano or determining the ﬂux [20,100] h 10 % 5 % at one selected distance. d. Background upwind correction B from ﬁt 5 % 5 % – Spatial smoothing effects caused by the satellite ground e. Plume altitude 1.5–2.5 km pixel size are explicitly accounted for. 1–2 km 29 % −8 % 2–3 km −25 % 32 % The implicit assumptions of our approach are that the observed monthly mean of daily cloud-free snapshots can be considered to represent steady state, i.e., that the source bars in Fig. 4), and the derived monthly means are also is steadily degassing, and a clear downwind transport pat- meaningful beyond 2009, despite GOME-2 degradation. tern is established by steady wind ﬁelds. The degassing of The impact of a potential additive bias in SCDs is mostly Kılauea from March 2008 on turned out to be an ideal occa- compensated for by the applied background correction. We sion for applying this method. Further possible applications thus conclude that uncertainties of SCDs are negligible for might include other steadily degassing volcanoes in stable our study. wind regimes, like Nyamuragira/Nyiragongo in Congo, or anthropogenic sources like metal smelters. 4.2.2 VCDs 4.2 Uncertainties The AMFs, used for the conversion of SCDs to VCDs, are based on the settings listed in Table 1. In this section, we discuss the uncertainties of the different steps of our procedure. Tables 1 and 2 list the applied a priori a. An albedo of 5 % was used for the calculation of AMFs. settings for the calculation of VCDs and the ﬁt of the model A change in this a priori by ±2 % changes the AMF by function F (t), respectively, and quantify the effects of alter- ±8 %, and the VCD by ∓8 %, respectively. native settings. b. The RTM calculations are performed at the lower end 4.2.1 SCDs of the DOAS ﬁt range (315 nm), where SO absorption The detection limit of individual SCDs is 1– bands are strongest. A change in the considered wave- 16 −2 2×10 molec cm . Due to averaging (monthly means) and length to 313 nm (at the absorption peak within the ﬁt spatial integration (line densities), however, the remaining window, where nonlinear effects are highest) increases statistical errors are rather small (compare the vertical error the VCD by 12 %. Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/ S. Beirle et al.: Emissions and lifetime of SO from Kılauea 8317 c. As the sensitivity of satellite measurements in the UV lifetime represents chemical loss rather than dilution ef- decreases with altitude over dark surfaces, the a priori fects, as the full plume is covered. plume altitude has a strong impact on the AMF and c. We ﬁt τ and E over a time interval from −20 (upwind) VCD. A shift of 0.5 km up or down changes the VCD to 100 h (downwind), i.e., the range shown in Fig. 4. by about 20 %. Variations in the time interval, i.e., choosing a longer or shorter time interval, have hardly any effect on the mean ﬁt results. We also tested a time interval starting at 20 h, 4.2.3 Fit of SO emission rates and lifetimes in order to avoid inhomogeneities due to terrain effects (ground albedo, wind speeds) over Hawai‘i Island. Even Emission rates and lifetimes of SO are derived from though the largest SO ﬂuxes are skipped for this ﬁt in- monthly mean GOME-2 VCDs, involving cloud masking, 2 terval, the ﬁt results hardly changed. spatial integration and the ﬁtting of a simple model function. The ﬁt provides conﬁdence intervals for τ and E, which are 1 d. The treatment of the SO background is also not critical. in the range of 15–30 and 10–20 %, respectively . In the baseline algorithm, an a priori background correc- Additional uncertainties may arise from the a priori set- tion is applied based on upwind measurements at 150– tings. Table 2 lists the baseline settings used in our analysis, 153 W. If, instead, the background B is added as a free and compares the resulting mean emission rates and lifetimes ﬁtting parameter in Eq. (1), as in Beirle et al. (2011), for March–November 2008 to the respective results for vari- both E and τ change by only 5 %. ous alternative a priori settings. e. The plume altitude has a strong impact on the resulting a. We calculate monthly mean SO VCDs for observations emission rates and lifetimes; a relatively small shift of with cloud fractions below 20 %, but do not correct for the a priori plume altitude of ±0.5 km changes the total remaining cloud effects in our analysis, as operational emissions by ≈ ∓30 %. This relatively large response is cloud heights are uncertain for low cloud fractions. We caused by two different effects: ﬁrst, the AMFs are al- justify this procedure by investigating the impact of the titude dependent, directly affecting VCDs, and thereby cloud fraction threshold on our results. The variation of emissions, by up to 20 % (see previous section). Sec- the cloud fraction threshold to either 10 or 30 % has only ond, horizontal wind speeds also change with altitude, a minor impact on the estimated lifetime (≈ 10 %), and affecting the calculation of ﬂuxes, and thereby the ﬁt- hardly any effect on the emission rate (≈ 3 %). Thus, ted lifetimes and emission rates. For the Kılauea case the actual choice of the cloud fraction threshold is not study, both effects have the same sign (lower a priori critical, and cloud effects on AMFs are negligible (at plume altitude leads to higher emission rate estimates) least up to a cloud fraction threshold of 30 %). This is and amplify instead of cancel each other. Note that the probably the result of cloud heights scattering around strong effect of the a priori plume altitude is also the the SO plume height: apparently, the effects of shield- main reason for the different results of this study com- ing (clouds above SO plume) vs. multiple scattering pared to those reported previously (Beirle et al., 2012), (clouds at or under plume) cancel out, at least partly which were gained via a similar approach, but based on (compare Sect. 2.2). a mean plume altitude of 2.5 km. Consequently, the re- ported emission rates in Beirle et al. (2012) are lower b. The SO VCDs are integrated in the latitudinal direction by about 25 %, and lifetimes are higher by about 30 % from 10 to 25 N in order to determine LDs. If a smaller (compare Table 2). integration interval (15–20 N) is chosen instead, this interval no longer covers the entire SO plume, espe- 2 As argued in Sect. 2.3, we can constrain the mean plume cially for larger distances from the vent. Consequently, altitude by measurements close to the vent, space-based the ﬁtted lifetime is biased low (−17 %), as the lack lidar measurements by CALIOP, and the comparison of plume coverage in the downwind regions acts as an of the plume outﬂow direction with the wind direc- additional virtual sink. On the other hand, an increase tion at different altitudes. In view of the consistency of in the latitude range over which integration takes place height information deduced from different and indepen- does not signiﬁcantly change the derived SO lifetime. 2 dent data sets, we conclude that it represents the average Hence, the chosen range is appropriate, and the ﬁtted plume altitude for the period of high SO VCDs in 2008 well, with an uncertainty of about 0.5 km (on average). In April, westerly winds occurred on some days, which pushed We use a constant plume height for our analysis. A the plume towards the region used for the determination of the back- possible uplift or descent of the SO plume during ad- ground. Consequently, the estimated background for April is biased vection would increase/decrease the satellite’s sensitiv- high, resulting in negative LDs at t >40 h, which cannot be repro- ity, and would gradually bias the derived line densities duced by the model function F (t). The respective ﬁt uncertainties are thus considerably higher (60 and 45 %) for April. high/low. This would lead to a virtual longer/shorter www.atmos-chem-phys.net/14/8309/2014/ Atmos. Chem. Phys., 14, 8309–8322, 2014 8318 S. Beirle et al.: Emissions and lifetime of SO from Kılauea lifetime, as the SO decay is suppressed/raised. The ef- Table 3. Effect of temporal variations of E and τ . For details see text. fect on E, however, would be rather low, as the maxi- mum ﬂux close to the source would not be affected. T τ /τ k /k E /E Fit Fit Fit Nevertheless, the strong effect of the a priori plume alti- 1 h 0.74 0.99 1.04 tude makes it the dominant source of uncertainty for the 10 h 0.81 0.91 1.04 derived emission rates. 100 h 1.02 0.71 0.96 Overall, we estimate the total uncertainty of emission rates and lifetimes to about 40 % for the mean of March– November 2008 by adding the dominant error terms quadrat- ically. For individual months, uncertainties might be higher, T= 1 h particularly for emission rates. 4.3 Temporal variability and nonlinear effects The ﬁtted model function assumes one value for both E and T= 10 h τ . In reality, however, emissions as well as instantaneous life- times are highly variable: – Volcanic emission rates can show high temporal ﬂuctua- tions. This could lead to sampling biases due to the lim- T= 100 h ited temporal coverage of the satellite observations. In the monthly mean, however, such effects at least partly cancel out, and the ﬁtted emission rates actually repre- hi sent the mean of the time-dependent emission rates 0 20 40 60 80 100 E , as the mean of exponential decays of varying emis- Time [h] sion rates equals the exponential decay of the respective mean emission rate E: Figure 8. Simulated (red, mean of 10 000 simulations) and ﬁtted (black) downwind decay for varying input emission rates and in- t t t stantaneous lifetimes, for different time intervals T over which the E exp − = hE i exp − = E exp − . (2) i i τ τ τ instantaneous lifetime is kept constant. – For variations of the lifetime of SO , which are intrin- sic due to the different removal processes, the situation For high values of T (i.e., each decay has only one in- is different insofar as τ in Eq. (1) is an argument of stantaneous lifetime), the ﬁtted lifetimes match the mean of a nonlinear function. Thus, the mean of different func- a priori lifetimes. However, for this scenario, the simulated tions exp − cannot, in a strict mathematical sense, downwind decay does not perfectly match the model func- tion of a single time constant as a consequence of the non- be transformed into a single function exp − . Nev- eff linearities discussed above. In addition, if the instantaneous ertheless, as shown in Fig. 4, the observed downwind rate constants k = 1/τ are averaged instead of the lifetimes, i i plumes are well described by a single, effective time they deviate from the mean a priori k by 30 %. constant. For short values of T , on the other hand, the ﬁt works well: the switch of time constants in the simulation results in a de- In order to investigate the effect of temporal variations of cay pattern which can well be described by a single time con- E and τ and the nonlinearity in τ , we calculated synthetic stant. The lifetime returned by the ﬁt is signiﬁcantly shorter downwind decays F = E exp − for E being a nor- i i i i than the mean of the lifetimes used for the simulations, while mally distributed random number with σ = E/2, and τ the respective average of rate constants matches the actual E i,j being a random number uniformly distributed between 1 and mean within 1 %. As the ﬁtted emissions are in very good 60 h. The additional index j indicates that for each i, τ is as- agreement with the a priori (±4 %), we conclude that the signed a new random number after a time step of length T . lifetime found by the ﬁt, though different from the mathe- We calculated 10 000 individual downwind plumes in a La- matical mean, is the appropriate quantity to link the mean grangian framework (i.e., the temporal evolution of each air SO VCDs to the respective emission rates by mass balance parcel, containing the initial SO amount 1E emitted in 1t , (E = (VCD)/τ ), and, in this sense, represents the effective is considered independently, and dilution is not accounted lifetime. for) and performed the ﬁt of E and τ to the average down- wind plume. Figure 8 and Table 3 summarize the results: Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/ Flux [A.U.] S. Beirle et al.: Emissions and lifetime of SO from Kılauea 8319 −1 The situation may be different if variations of instanta- Table 4. Monthly mean emission rates (in kt day ) from GOME-2 neous lifetimes are not purely random but instead systematic. compared to Elias and Sutton (2012) (ES12) for the Junes of 2007– For instance, if instantaneous lifetimes were always short 2010. close to the vent, e.g., due to heterogeneous reactions within GOME-2 GOME-2 ES12 ES12 ES12 the plume, a considerable fraction of SO might be lost be- Year (from ﬁt) (from VCD) (total) (rift) (summit) fore it could be detected from space. For such a scenario, the initial emission rates might be considerably higher than those 2007 0.7 1.3 1.1 0.2 2008 12.5 11.2 2.8 1.9 0.9 deduced from space. 2009 6.4 2.2 1.4 0.9 2010 5.3 1.3 0.5 0.8 4.4 Comparison to results from ground-based monitoring The US Geological Survey (USGS) has been monitoring nonlinear RT effects occurring in and around the optically SO emission rates from Kılauea for several decades. Elias thick volcanic plume emanating from Kılauea’s summit. and Sutton (2012) report emission rates for 2007–2010 de- Note that satellite measurements might be affected by such rived from stationary as well as vehicle-based spectroscopic nonlinear RT effects as well (Hörmann et al., 2013). Due to measurements with the FLYSPEC system (Horton et al., the large ground pixels and the view from above, however, 2006) downwind of the Kılauea summit and East Rift (about the effects are by far smaller than for slant observations from 2–4 measurements per week). The estimated emission rates the ground in very close proximity to the source. are included in Fig. 7. However, since the re-analyzed data are only available for SO emission rates from Kılauea’s summit were mainly a select few case studies, we compare the monthly mean measured by traversing beneath the plume on Crater Rim emissions derived from GOME-2 data to the time series ob- Drive. The plume was transected at distances between 0.3 tained with the standard FLYSPEC evaluation provided in and 1.3 km downwind of the active vent depending on its Figs. 29 and 30 in Elias and Sutton (2012). Overall, our propagation direction on a given day. The emission rates emission rate estimates based on GOME-2 are far higher from the East Rift were measured by traversing the plume on than those reported in Elias and Sutton (2012). For total Chain of Craters Road, approximately 9 km from the active Kılauea emissions in 2008, we ﬁnd 3.9 Tg, in comparison Pu‘u‘O’o ¯ vent from which it was originating. Wind speeds to the 1.1 Tg in Elias and Sutton (2012) (Fig. 29 therein). were determined from meteorology stations installed at the The ﬁndings of Kern et al. (2012) are already discussed in Hawaiian Volcano Observatory and Holei ¯ Pali (see Elias and Elias and Sutton (2012), and a ﬁrst, preliminary analysis ac- Sutton (2012) for details). counting for the nonlinear RT effects is presented. For this re- The ground-based FLYSPEC measurements were initially vised analysis, higher SO emission rates are derived (annual evaluated using standard FLYSPEC routines, i.e., by ﬁtting emissions sum up to 1.5 Tg instead of 1.1 Tg for 2008), but the spectrum of scattered solar radiation passing through cali- are still signiﬁcantly lower than those derived from GOME- bration cells containing known SO amounts to the measured 2 in this study. Note, however, that Kern et al. (2012) did spectrum of radiation passing through the plume (for details, not re-analyze data from 2008, the period during which our see Elias and Sutton, 2012, and references therein). How- measurements indicate the largest emissions from Kılauea’s 19 −2 ever, SO column densities were high (> 10 molec cm ) summit. Since the magnitude of the applied RT corrections and the plume was oftentimes highly opaque due to aerosols scales with the encountered SO column density, it is likely and condensed water droplets at the point of measurement in that the assessment of the 2008 emissions by Elias and Sutton close proximity to the vent. Thus, light is often prevented (2012) remains skewed to low values. Further detailed inves- from penetrating the plume core. Instead, light paths that tigation is necessary to explore this issue, but that is beyond only penetrate the plume edge before being scattered towards the scope of this study. the instrument become more likely, therefore skewing mea- On the other hand, in 2007, when summit emissions were surements towards lower column densities. These effects are negligible, our emission rates are actually lower than the discussed in Kern et al. (2012), and accounted for by using ground-based values. During this time period, Elias and Sut- an approach they called simulated radiative transfer DOAS ton (2012) report signiﬁcantly enhanced East Rift emissions (SRT-DOAS), which basically compares the measured spec- coinciding with eruptive activity. Since the SO emission tra to simulated spectra using a three-dimensional RTM. rates from the East Rift were measured on Chain of Craters Kern et al. (2012) applied these more elaborate retrievals on Road 9 km from the Pu‘u‘O’o ¯ vent, the plume was sig- several examples of the FLYSPEC data from 2007, 2010 and niﬁcantly diluted, i.e., optically thinner, such that nonlin- 2011. The resulting SO emission rates were between −20 ear RT effects were probably negligible, especially before and +90 % of the initial FLYSPEC results, depending on the 2008 (Elias and Sutton, 2012). Therefore, the ground-based measurement conditions (Kern et al., 2012; Elias and Sutton, SO emission rates from the East Rift in 2007 are likely 2012). The authors attributed these large discrepancies to the fairly accurate. For this time period, the discrepancy between www.atmos-chem-phys.net/14/8309/2014/ Atmos. Chem. Phys., 14, 8309–8322, 2014 8320 S. Beirle et al.: Emissions and lifetime of SO from Kılauea satellite and ground-based emission rates instead appears to this approach only works as long as SO lifetime and plume be caused by the lower plume altitude in 2007, which affects altitude do not change signiﬁcantly. the satellite AMF. As mentioned before, the difference in al- An accurate a priori vertical trace gas and aerosol con- titude between the Kılauea summit and East Rift emission centration proﬁle is needed for the quantitative interpreta- plumes leads to a different sensitivity of the satellite instru- tion of SO absorption measured by satellites. For an isolated ment towards SO emitted at each of the two locations. Prior “point source” (in terms of the satellite’s spatial resolution) to the opening of the Overlook Vent at Kılauea’s summit in like Kılauea, however, the plume altitude can be constrained 2008, the entire plume originated from the East Rift and was from its downwind propagation direction (at least after the on average located at a lower altitude. Although it is difﬁcult opening of the Overlook Vent at the summit in 2008). to accurately quantify the exact impact on the AMF without Kılauea turned out to be particularly suited for our ap- an accurate measurement of the plume altitude prior to 2008, proach, as (a) it is a singular, remote source of SO , (b) the it is clear that a lower plume would systematically reduce the outﬂow patterns are clear and constrained due to the steady sensitivity of the satellite towards SO , and this behavior is trade winds, and (c) clouds turned out to be not critical. deemed the main cause for the discrepancy of satellite and A similar analysis may, in principle, be performed on other ground-based results in 2007. In addition, the SO lifetime strong point sources of SO , like other degassing volcanoes 2 2 might be shorter for the East Rift emissions (Porter et al., or strong industrial sources, but will probably be more chal- 2002), which would also cause a low bias of emission rate es- lenging due to higher cloud fractions, higher wind variability timates based on mean VCDs. In Fig. 7, the respective emis- or interfering SO sources. sion rate estimates for 2007 are displayed in light orange. Satellite instruments like SCIAMACHY, OMI and GOME-2 will provide a continuous time series cover- ing more than 2 decades. Future satellite instruments like TROPOMI (Veefkind et al., 2012) will also provide better 5 Conclusions spatial coverage (once per day) and have smaller ground Satellite measurements provide new potential to investigate pixels, resulting in a higher fraction of cloud-free observa- and quantify sources and transformations of atmospheric tions. In addition, the better spatial resolution could poten- trace gases. By analyzing the downwind plume of SO from tially reveal a possibly different plume chemistry, i.e., effec- the Kılauea volcano, SO lifetimes and emission rates can tive SO lifetime, close to the source. Furthermore, the ex- 2 2 be derived from GOME-2 observations and ECMWF wind pected better detection limit might facilitate the investigation ﬁelds, but without the need for a chemical model. of weaker sources of SO . For the period of most active summit degassing in March– November 2008, we ﬁnd monthly mean effective SO life- Acknowledgements. We thank A. J. Sutton (USGS) and two times of 1–2 days, with highest lifetimes in summer, when anonymous reviewers for valuable comments on this manuscript. cloud cover is small, and lowest in spring and autumn. −1 We acknowledge EUMETSAT for providing GOME-2 spectra, Emission rates in 2008 are estimated to be 9–21 kt day , ECMWF for providing wind ﬁelds, and NASA for providing which is signiﬁcantly higher (about 3 times) than initially aerosol optical depth from MODIS Terra. Contour lines are derived reported from ground-based measurements. Further investi- from SRTM topographic data provided by the US Geological gation into the nonlinear RT affecting the ground-based mea- Survey (http://dds.cr.usgs.gov/srtm/version2_1/SRTM3). Oliver surements is needed to explore this disagreement. It should Woodford is acknowledged for providing the MATLAB routine be noted that the Kılauea data set is unique in that the export_fig, which signiﬁcantly simpliﬁes the processing of ground-based measurements were typically made less than MATLAB ﬁgures. 1 km from an extremely prodigiously degassing vent. Thus, 19 −2 the encountered SO column densities (> 10 molec cm ) The service charges for this open access publication have been covered by the Max Planck Society. and aerosol optical depths (> 10, values from Kern et al., 2012) were signiﬁcantly higher than is usually the case. Edited by: M. Van Roozendael Therefore, in this particular case, SO emission rates de- rived from ground-based traverses using standard retrieval schemes signiﬁcantly underestimate the true emissions (Kern References et al., 2012), while satellite instruments, which are able to measure the plume farther downwind where the plume is Andres, R. J. and Kasgnoc, A. D.: A time-averaged inventory of more dilute, experience more favorable measurement condi- subaerial volcanic sulfur emissions, J. Geophys. Res.-Atmos., tions. 103, 25251–25261, doi:10.1029/98JD02091, 1998. From the observed linear relation between emission rates Beirle, S., Platt, U., Von Glasow, R., Wenig, M., and Wag- and monthly mean column densities, emission rates can also ner, T.: Estimate of nitrogen oxide emissions from shipping be roughly estimated for months with lower SO VCDs, by satellite remote sensing, Geophys. Res. Lett., 31, L18102, where no clear downwind plume could be ﬁtted. However, doi:10.1029/2004GL020312, 2004. Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/ S. Beirle et al.: Emissions and lifetime of SO from Kılauea 8321 Beirle, S., Salzmann, M., Lawrence, M. G., and Wagner, T.: Sen- Jacobians and 3-D features, J. Quant. Spectrosc. Ra., 112, 1119– sitivity of satellite observations for freshly produced lightning 1137, doi:10.1016/j.jqsrt.2010.12.009, 2011. NO , Atmos. Chem. Phys., 9, 1077–1094, doi:10.5194/acp-9- Eguchi, K., Uno, I., Yumimoto, K., Takemura, T., Nakajima, T. Y., 1077-2009, 2009. Uematsu, M., and Liu, Z.: Modulation of cloud droplets and ra- Beirle, S., Boersma, K. F., Platt, U., Lawrence, M. G., and diation over the North Paciﬁc by sulfate aerosol erupted from Wagner, T.: Megacity emissions and lifetimes of nitro- Mount Kılauea, SOLA, 7, 77–80, doi:10.2151/sola.2011-020, gen oxides probed from space, Science, 333, 1737–1739, 2011. doi:10.1126/science.1207824, 2011. Elias, T. and Sutton, A. J.: Sulfur dioxide emission rates from Beirle, S., Hörmann, C., Penning de Vries, M., and Wagner, T.: Es- Kılauea Volcano, Hawai‘i, 2007–2010, US Geological Survey timating the Lifetime of SO and Aerosols from Space: a Case Open-File Report 2012–1107, available at: http://pubs.usgs.gov/ Study for the Kılauea Volcano, in: Proceedings of the ESA con- of/2012/1107/ (last access: 26 June 2012), 25 pp., 2012. ference “ATMOS – Advances in Atmospheric Science and Ap- Galle, B., Oppenheimer, C., Geyer, A., McGonigle, A. J. S., Ed- plications”, 18–22 June 2012, SP-708 (CD-ROM), Bruges, Bel- monds, M., and Horrocks, L.: A miniaturised ultraviolet spec- gium, 2012. trometer for remote sensing of SO ﬂuxes: a new tool for volcano Bluth, G. J. S. and Carn, S. A.: Exceptional sulfur degassing surveillance, J. Volcanol. Geoth. Res., 119, 241–254, 2002. from Nyamuragira volcano, 1979-2005, Int. J. Remote Sens., 29, Galle, B., Johansson, M., Rivera, C., Zhang, Y., Kihlman, M., Kern, 6667–6685, doi:10.1080/01431160802168434, 2008. C., Lehmann, T., Platt, U., Arellano, S., and Hidalgo, S.: Net- Bogumil, K., Orphal, J., Homann, T., Voigt, S., Spietz, P., Fleis- work for Observation of Volcanic and Atmospheric Change (NO- chmann, O., Vogel, A., Hartmann, M., Kromminga, H., Bovens- VAC) – A global network for volcanic gas monitoring: Network mann, H., Frerick, J., and Burrows, J.: Measurements of molec- layout and instrument description, J. Geophys. Res.Atmos., 115, ular absorption spectra with the SCIAMACHY pre-ﬂight model: D05304, doi:10.1029/2009JD011823, 2010. instrument characterization and reference data for atmospheric Gür, B., Spietz, P., Orphal, J., and Burrows, J. P.: Absorption Spectra remote sensing in the 230-2380 nm region, J. Photoch. Photobio. Measurements with the GOME-2 FMs using the IUP/IFE-UBs A, 157, 167–184, doi:10.1016/S1010-6030(03)00062-5, 2003. Calibration Apparatus for Trace Gas Absorption Spectroscopy Callies, J., Corpaccioli, E., Eisinger, M., Hahne, A., and Lefeb- VATGAS, Final Report, University of Bremen, October 2005, vre, A.: GOME-2-MetOp’s second-generation sensor for op- 2005. erational ozone monitoring, ESA Bull., 102, 28–36, avail- Hughes, E. J., Sparling, L. C., Carn, S. A., and Krueger, A. J.: able at: http://www.esa.int/esapub/bulletin/bullet102/Callies102. Using horizontal transport characteristics to infer an emission pdf (last access: 20 March 2013), 2000. height time series of volcanic SO , J. Geophys. Res.-Atmos., Cao, G., Giambelluca, T. W., Stevens, D., and Schroeder, T.: Inver- 117, D18307, doi:10.1029/2012JD017957, 2012. sion variability in the Hawaiian trade wind regime, J. Climate, Hörmann, C., Sihler, H., Bobrowski, N., Beirle, S., Penning de 20, 1145–1160, 2007. Vries, M., Platt, U., and Wagner, T.: Systematic investigation Carn, S. A., Krueger, A. J., Bluth, G. J. S., Schaefer, S. J., Krotkov, of bromine monoxide in volcanic plumes from space by using N. A., Watson, I. M., and Datta, S.: Volcanic eruption detec- the GOME-2 instrument, Atmos. Chem. Phys., 13, 4749–4781, tion by the Total Ozone Mapping Spectrometer (TOMS) instru- doi:10.5194/acp-13-4749-2013, 2013. ments: a 22-year record of sulphur dioxide and ash emissions, Horton, K. A., Williams-Jones, G., Garbeil, H., Elias, T., Sut- Geological Society, London, Special Publications, 213, 177–202, ton, A. J., Mouginis-Mark, P., Porter, J. N., and Clegg, S.: doi:10.1144/GSL.SP.2003.213.01.11, 2003. Real-time measurement of volcanic SO emissions: validation Carn, S. A., Strow, L. L., De Souza-Machado, S., Edmonds, Y., and of a new UV correlation spectrometer (FLYSPEC), B. Volcanol., Hannon, S.: Quantifying tropospheric volcanic emissions with 68, 323–327, doi:10.1007/s00445-005-0014-9, 2006. AIRS: the 2002 eruption of Mt. Etna (Italy), Geophys. Res. Lett., Kern, C., Deutschmann, T., Werner, C., Sutton, A. J., Elias, T., 32, L02301, doi:10.1029/2004GL021034, 2005. and Kelly, P. J.: Improving the accuracy of SO column den- Carn, S. A., Krueger, A. J., Arellano, S., Krotkov, N. A., sities and emission rates obtained from upward-looking UV- and Yang, K.: Daily monitoring of Ecuadorian volcanic de- spectroscopic measurements of volcanic plumes by taking re- gassing from space, J. Volcanol. Geoth. Res., 176, 141–150, alistic radiative transfer into account, J. Geophys. Res., 117, doi:10.1016/j.jvolgeores.2008.01.029, 2008. D20302, doi:10.1029/2012JD017936, 2012. Carn, S. A., Krotkov, N. A., Yang, K., and Krueger, A. J.: Measuring Khokhar, M. F., Frankenberg, C., Van Roozendael, M., Beirle, S., global volcanic degassing with the Ozone Monitoring Instrument Kühl, S., Richter, A., Platt, U., and Wagner, T.: Satellite obser- (OMI), Geological Society, London, Special Publications, 380, vations of atmospheric SO from volcanic eruptions during the 229–257, doi:10.1144/SP380.12, 2013. time-period of 1996–2002, Adv. Space Res., 36, 879–887, 2005. Carslaw, K. S., Lee, L. A., Reddington, C. L., Pringle, K. J., Rap, Koelemeijer, R. B. A., Stammes, P., Hovenier, J. W., and de Haan, A., Forster, P. M., Mann, G. W., Spracklen, D. V., Woodhouse, J. F.: A fast method for retrieval of cloud parameters using M. T., Regayre, L. A., and Pierce, J. R.: Large contribution of oxygen A band measurements from the Global Ozone Moni- natural aerosols to uncertainty in indirect forcing, Nature, 503, toring Experiment, J. Geophys. Res.-Atmos., 106, 3475–3490, 67–71, doi:10.1038/nature12674, 2013. doi:10.1029/2000JD900657, 2001. Deutschmann, T., Beirle, S., Frieß, U., Grzegorski, M., Kern, C., Krotkov, N. A., Schoeberl, M. R., Morris, G. A., Carn, S., and Kritten, L., Platt, U., Prados-Roman, C., Puk ¸ ıte, J., Wagner, T., Yang, K.: Dispersion and lifetime of the SO cloud from the Werner, B., and Pfeilsticker, K.: The Monte Carlo atmospheric August 2008 Kasatochi eruption, J. Geophys. Res.-Atmos., 115, radiative transfer model McArtim: introduction and validation of D00L20, doi:10.1029/2010JD013984, 2010. www.atmos-chem-phys.net/14/8309/2014/ Atmos. Chem. Phys., 14, 8309–8322, 2014 8322 S. Beirle et al.: Emissions and lifetime of SO from Kılauea Lee, C., Martin, R. V., Van Donkelaar, A., Lee, H., Dicker- Platt, U. and Stutz, J.: Differential Optical Absorption Spec- son, R. R., Hains, J. C., Krotkov, N., Richter, A., Vinnikov, K., troscopy, Springer-Verlag, Berlin, Heidelberg, 2008. and Schwab, J. J.: SO emissions and lifetimes: estimates from Porter, J. N., Horton, K. A., Mouginis-Mark, P. J., Lienert, B., inverse modeling using in situ and global, space-based (SCIA- Sharma, S. K., Lau, E., Sutton, A. J., Elias, T., and Oppen- MACHY and OMI) observations, J. Geophys. Res.-Atmos., 116, heimer, C.: Sun photometer and lidar measurements of the D06304, doi:10.1029/2010JD014758, 2011. plume from the Hawai‘i Kılauea Volcano Pu‘u ‘O‘ov ¯ ent: aerosol Leitão, J., Richter, A., Vrekoussis, M., Kokhanovsky, A., Zhang, Q. ﬂux and SO lifetime, Geophys. Res. Lett., 29, 30-1–30-4, J., Beekmann, M., and Burrows, J. P.: On the improvement of doi:10.1029/2002GL014744, 2002. NO satellite retrievals – aerosol impact on the airmass factors, Puk ¸ ıte, J., Kühl, S., Deutschmann, T., Platt, U., and Wagner, T.: Atmos. Meas. Tech., 3, 475–493, doi:10.5194/amt-3-475-2010, Extending differential optical absorption spectroscopy for limb 2010. measurements in the UV, Atmos. Meas. Tech., 3, 631–653, Lelieveld, J., Roelofs, G., Feichter, J., and Rodhe, H.: Terrestrial doi:10.5194/amt-3-631-2010, 2010. sources and distribution of atmospheric sulphur, Philos. T. Roy. Robock, A.: Volcanic eruptions and climate, Rev. Geophys., 38, Soc. B, 352, 149–157, 1997. 191–219, doi:10.1029/1998RG000054, 2000. Leue, C., Wenig, M., Wagner, T., Klimm, O., Platt, U., and Theys, N., Campion, R., Clarisse, L., Brenot, H., van Gent, J., Dils, Jähne, B.: Quantitative analysis of NO emissions from Global B., Corradini, S., Merucci, L., Coheur, P.-F., Van Roozendael, Ozone Monitoring Experiment satellite image sequences, J. Geo- M., Hurtmans, D., Clerbaux, C., Tait, S., and Ferrucci, F.: Vol- phys. Res., 106, 5493–5505, 2001. canic SO ﬂuxes derived from satellite data: a survey using OMI, Martin, R.: Satellite remote sensing of surface air quality, Atmos. GOME-2, IASI and MODIS, Atmos. Chem. Phys., 13, 5945– Environ., 42, 7823–7843, doi:10.1016/j.atmosenv.2008.07.018, 5968, doi:10.5194/acp-13-5945-2013, 2013. 2008. Veefkind, J. P., Aben, I., McMullan, K., Förster, H., de Vries, J., McCormick, B. T., Herzog, M., Yang, J., Edmonds, M., Mather, T. Otter, G., Claas, J., Eskes, H. J., de Haan, J. F., Kleipool, Q., A., Carn, S. A., Hidalgo, S., and Langmann, B.: A comparison van Weele, M., Hasekamp, O., Hoogeveen, R., Landgraf, J., of satellite- and ground-based measurements of SO emissions Snel, R., Tol, P., Ingmann, P., Voors, R., Kruizinga, B., from Tungurahua volcano, Ecuador, J. Geophys. Res.-Atmos., Vink, R., Visser, H., and Levelt, P. F.: TROPOMI on the ESA 119, 4264–4285, doi:10.1002/2013JD019771, 2014. Sentinel-5 Precursor: a GMES mission for global observations Moffat, A. J. and Millan, M. M.: The applications of optical corre- of the atmospheric composition for climate, air quality and lation techniques to the remote sensing of SO plumes using sky ozone layer applications, Remote Sens. Environ., 120, 70–83, light, Atmos. Environ., 5, 677–690, 1971. doi:10.1016/j.rse.2011.09.027, 2012. Monks, P. S. and Beirle, S.: Applications of satellite observations Von Glasow, R., Bobrowski, N., and Kern, C.: The effects of vol- of tropospheric composition, in: The Remote Sensing of Tropo- canic eruptions on atmospheric chemistry, Chem. Geol., 263, spheric Composition from Space, Springer-Verlag, Berlin, Hei- 131–142, doi:10.1016/j.chemgeo.2008.08.020, 2009. delberg, 365–449, 2011. Wang, P., Stammes, P., van der A, R., Pinardi, G., and van Roozen- Mori, T. and Burton, M.: The SO camera: a simple, fast and cheap dael, M.: FRESCO+: an improved O A-band cloud retrieval 2 2 method for ground-based imaging of SO in volcanic plumes, algorithm for tropospheric trace gas retrievals, Atmos. Chem. Geophys. Res. Lett., 33, L24804, doi:10.1029/2006GL027916, Phys., 8, 6565–6576, doi:10.5194/acp-8-6565-2008, 2008. 2006. Yuan, T., Remer, L. A., and Yu, H.: Microphysical, macrophysical Oppenheimer, C., Francis, P., and Stix, J.: Depletion rates of sulfur and radiative signatures of volcanic aerosols in trade wind cu- dioxide in tropospheric volcanic plumes, Geophys. Res. Lett., 25, mulus observed by the A-Train, Atmos. Chem. Phys., 11, 7119– 2671–2674, doi:10.1029/98GL01988, 1998. 7132, doi:10.5194/acp-11-7119-2011, 2011. Atmos. Chem. Phys., 14, 8309–8322, 2014 www.atmos-chem-phys.net/14/8309/2014/

http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

Atmospheric Chemistry and Physics Unpaywall

http://www.deepdyve.com/lp/unpaywall/estimating-the-volcanic-emission-rate-and-atmospheric-lifetime-of-so-TMgmOXbrFH

Estimating the volcanic emission rate and atmospheric lifetime of SO<sub>2</sub> from space: a case study for Kīlauea volcano, Hawai`i

Loading next page...

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher: Unpaywall
ISSN: 1680-7316
DOI: 10.5194/acp-14-8309-2014
Publisher site: See Article on Publisher Site

Abstract

Journal

Atmospheric Chemistry and Physics – Unpaywall

Published: Aug 19, 2014

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Estimating the volcanic emission rate and atmospheric lifetime of SO<sub>2</sub> from space: a case study for Kīlauea volcano, Hawai`i

Estimating the volcanic emission rate and atmospheric lifetime of SO<sub>2</sub> from space: a case study for Kīlauea volcano, Hawai`i

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Estimating the volcanic emission rate and atmospheric lifetime of SO&lt;sub&gt;2&lt;/sub&gt; from space: a case study for Kīlauea volcano, Hawai`i

Estimating the volcanic emission rate and atmospheric lifetime of SO&lt;sub&gt;2&lt;/sub&gt; from space: a case study for Kīlauea volcano, Hawai`i

References

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies

Estimating the volcanic emission rate and atmospheric lifetime of SO<sub>2</sub> from space: a case study for Kīlauea volcano, Hawai`i

Estimating the volcanic emission rate and atmospheric lifetime of SO<sub>2</sub> from space: a case study for Kīlauea volcano, Hawai`i