Intercomparison of Metop-A SO2 measure- ments during the 2010- 2011 Icelandic eruptions

The European Space Agency project Satellite Monitoring of Ash and Sulphur Dioxide for the mitigation of Avi­ ation Hazards, was introduced after the eruption of the Icelandic volcano Eyjafjallajökull in the spring of 2010 to facilitate the development of an optimal End­to­End System for Volcanic Ash Plume Monitoring and Predic­ tion. The Eyjafjallajökull plume drifted towards Europe and caused major disruptions of European air traffic for several weeks affecting the everyday life of millions of people. The limitations in volcanic plume monitoring and prediction capabilities gave birth to this observational system which is based on comprehensive satellite­derived ash plume and sulphur dioxide [SO2] level estimates, as well as a widespread validation using supplementary satellite, aircraft and ground­based measurements. Inter­comparison of the volcanic total SO2 column and plume height observed by GOME­2/Metop­A and IASI/Metop­A are shown before, during and after the Eyjaf­ jallajökull 2010 eruptions as well as for the 2011 Grímsvötn eruption. Co­located ground­based Brewer Spectro­ photometer data extracted from the World Ozone and Ultraviolet Radiation Data Centre for de Bilt, the Nether­ lands, are also compared to the different satellite estimates. Promising agreement is found for the two different types of instrument for the SO2 columns with linear regression coefficients ranging around from 0.64 when comparing the different instruments and 0.85 when comparing the two different IASI algorithms. The agree­ ment for the plume height is lower, possibly due to the major differences between the height retrieval part of the GOME2 and IASI algorithms. The comparisons with the Brewer ground­based station in de Bilt, The Nether­ lands show good qualitative agreement for the peak of the event however stronger eruptive signals are required for a longer quantitative comparison. doi: 10.4401/ag-6613 1 ANNALS OF GEOPHYSICS, Fast Track 2, 2014


I. INTRODUCTION
he main objective of the European Spa ce Agency [ESA] Satellite Monitoring of Ash and Sulphur Dioxide, SO 2 , for the mitigation of Aviation Hazards, SACS 2/SMASH, project is the development, impro vement and validation of satelitederived ash & SO 2 products.This system is based on im proved and dedicated satellitederived ash plume and sulphur dioxide level assessments following the Support to Aviation Control Ser vice (SACS, http://sacs.aeronomie.be),which is a free online service initiated by ESA for the nearrealtime satellite monitoring of volcanic plumes of SO 2 and ash [Brenot et al., 2014].The SACS2/SMASH system has been compre hensively validated using auxiliary satellite, aircraft and groundbased measurements [Koukouli et al., 2014;Spinetti et al., 2014.].This paper provides intercomparisons between three different SO 2 column and plume height assessments from two different satellite instruments for the 2011 Grímsvötn eruption.The satellite SO 2 products are further being compared to groundbased SO 2 column obser vations for the Eyjafjallajökull 2010 eruptions.

II. SO 2 OBSERVATIONS AND METHODOLOGIES
I.

IASI SULPHUR DIOXIDE PRODUCTS
The hyperspectral Infrared Atmospheric Sounding Interferometer (IASI) was launched in October 2006 onboard the Meteorological Operational satelliteA (MetOpA) [Clerbaux et al., 2009] for meteorological and scientific applications.Global nadir measurements are acquired twice a day at 09:30 and 21:30 mean local equatorial time with a 12 km radius foot print in the nadir.In the following analysis two IASI SO 2 products are used: one devel oped by the Atmospheric Spectroscopy group of the Université Libre de Bruxelles, hereafter ULB, and one developed by the Earth Observa tion Data Group at the University of Oxford, hereafter Oxford.Both products use similar al gorithms for the detection [Clarisse et al., 2014] and retrieval of total columns [Carboni et al., 2012] however the Oxford algorithm retrieves the SO 2 plume height simultaneously with the SO 2 column amount, which has the advantage of yielding realistic error estimates.The ULB algorithm retrieves the height prior to the col umn and has the added benefit of not relying on any apriori information.II.

GOME2 SULPHUR DIOXIDE PRODUCTS
The second Global Ozone Monitoring Experi ment (GOME2) is a UV/visible nadirviewing spectrometer covering the 240790 nm wave length interval with a spectral resolution of 0.2 -0.5 nm [Munro et al., 2006] also on board the MetOpA satellite.Its ground pixel size is nearly constant along the orbit, around 80 x 40 km².The GOME2 SO 2 products shown in the analysis have been developed at the Belgian Institute of Space Aeronomy, (hereafter BIRA IASB) and were derived with an extended ver sion of the GODFIT trace gas column retrieval algorithm [Lerot et al., 2010].The algorithm derives the SO 2 total column amount and effec tive plume height simultaneously by applying an iterative direct fitting method with an opti mal estimation inversion scheme.A priori in formation is only used for the SO 2 plume height. III.

BREWER SULPHUR DIOXIDE COLUMNS
Measuring total SO 2 column using Brewer in struments is discussed in Kerr et al., [2007], and references therein.Brewer spectropho tometers measure the UV absorption spectrum T at five wavelengths.Once the ozone column is calculated, the total SO 2 column can be deter mined using semiempirical equations based on the measured light intensity at each of the wavelengths, equivalent extraterrestrial light intensities, the air mass factor, the precalcu lated column amount of ozone and the absorp tion coefficients of the two interfering species.Since the signaltonoise level for the SO 2 ab sorption is usually quite low, very well calib rated instruments are required to monitor nominal SO 2 levels.To the best of our know ledge, no comprehensive error assessments of the Brewer total SO 2 columns have been re ported as yet.After imposing a rigorous qual ity control on the stations reporting Brewer SO 2 columns available through the World Ozone and Ultraviolet Radiation Data Centre (WOUDC), [Koukouli et al., 2014], a total of 16 European stations were accepted as possibly being able to show signs of increased SO 2 levels due to the volcanic eruptions.In extreme cases, such as volcanic eruptions [Fioletov et al., 1998], SO2 levels typically rise well above instrumental noise permitting their comprehensive study using the Brewer instrument.Flentje et al., [2010], showed enhanced atmospheric levels during the Eyjafjallajökull eruptive period us ing the Brewer observations at Hohenpeis senberg, Germany, and Rix et al., [2012], report good comparisons between GOME2/Metop A measurements and groundbased findings at Valentia, Ireland for the May 2010 events.Here, we show comparisons over de Bilt, The Netherlands, one of the few European loca tions that where substantial plume drifted over.

III. RESULTS
In Figure 1 the SO 2 column amount of the three algorithms is compared for May 23 rd 2011; the SO 2 column amounts are gridded on a regular grid of 0.25 x 0.25 deg latitude/longitude to ac count for the differences in footprint.Note that this is the first time the different retrieval schemes and findings are compared as such.The ULB/IASI results are shown on the left, the Oxford/IASI in the middle and the BIRA IASB/GOME2 in the right column.No cut offs/flags were used as part of a quality screening of the data since such a screening eliminates pixels with high SO 2 amounts in the central denser part of the plume.For aviation applications, such a strict quality control is not needed, as long as SO 2 is reliably detected.The results shown are very promising with all three algorithms capturing quite well both the magnitude of the loading as well as the central location of the plume to the NW of Iceland and its dispersion over the North Atlantic Sea.For the 190 common grids between the Oxford and ULB/IASI results, the mean SO 2 loading esti mated by the Oxford algorithm is 14.32±25.14D.U. and by the ULB algorithm at 11.76±16.60D.U. with a correlation coefficient of 0.85 for the two algorithms.For the 232 common grids between the BIRAIASB/GOME2 and the ULB/IASI results, the mean SO 2 loading esti mated by GOME2 is 9.06±14.33D.U. and by IASI 10.96±15.15D.U. with a correlation of 0.64.For the 173 common grids between the BIRAIASB/GOME2 and the Oxford/IASI re sults, the mean GOME2 loading is 10.97±15.68D.U. and the mean IASI loading is 16.52±28.60D.U. with a correlation of 0.64.In Figure 2 the SO 2 plume height estimated for the 23 rd of May 2011 from the IASI and GOME2 observations is shown.All three retrievals show a wellde fined plume, setting off in a westerly direction over Iceland before rising North and East wards, resembling an arm over the Arctic Sea.Since for small amounts of SO 2 , there is not enough information inherent in the measured spectra in order to retrieve an independent al titude value, the resultant height values will tend towards the a priori values.For Oxford/IASI the retrieval is performed in a pressure grid and the plume height a priori values is set to 400mbars (~6km) and the a pri ori error to 1000 mbars.In the BIRA IASB/GOME2 algorithm the a priori height er ror was set to 5km and so any a posteriori height error value that approaches the a priori error levels signifies nondependable detec tion.No screening was applied to either IASI data whereas a minimum amount of four de grees of freedom was permitted for the BIRA IASB/GOME2results in order to signify mean ingful plume height detection.As can be noted in the Figure, significant differences ex ist among the IASI and the GOME2 assess ments of the plume height.In numbers, for the 173 common grids, the mean GOME2 plume height is found at 7.15±0.84km and the mean Oxford/IASI plume height at 8.82±2.0km; for the 232 common points, the mean GOME2 plume height is at 7.12±0.95km and the mean ULB/IASI height at 9.40±1.44km.For the com parison between the two IASI algorithms, the 190 points provide a mean ULB/IASI height of 9.00±2.00km and the Oxford/IASI of 8.17±2.64km with a good correlation of Rsquared 0.62.The agreement between the two algorithms is valid for the entire European domain and not solely on the location of the maximum plume load and height.In Figure 3, the comparisons between satellite and groundbased SO 2 amounts in the days before, during and after the Eyjafjallajökull 2010 events [grey shaded areas] are shown for the De Bilt Brewer spec trophotometer in The Netherlands.The Ox ford/IASI product is shown in the left plot and the ULB/IASI product in the right.The groundbased daily mean Brewer data are shown in red, the instantaneous daytime [nighttime] satellite observations in light blue [light green] and the satellite daily mean in black.Since only four colocations were found for the BIRAIASB/GOME2 within a 200km radius and no time imposition, only the two IASI comparisons are presented.The compari son is quite good for the two IASI products with the levels rising equivalently between satellite and ground data for the highly loaded days after the eruption, reaching daily maxima around 23 D.U., well above instrumental noise.Taking into consideration substantial differ ences in instrumentation, algorithm, wavelength range, spatiotemporal resolution, and so on, the three algorithms are providing consistent results depicting well the central plume location as well as its dispersive pat terns.A highly satisfactory agreement, with correlations ranging between 0.64 and 0.85, is found for the comparisons of the SO 2 loading during the chosen Grímsvötn day; the plume height concordance provides a far greater chal lenge with the two IASI algorithms agreeing adequately both in location and in height, however the GOME2 results show a more uniform plume height across the domain and also seem not to be able to depict the plume height near the source.The IASI/MetopA comparisons with a groundbased station in de Bilt, The Netherlands, are also promising but would benefit greatly from a far larger statist ical sample during a stronger eruptive occa sion.We hence conclude that all three Metop A SO 2 products are mature enough to be used in a satellite volcanic monitoring system with possible improvements.As a side note, it is equally important to assess whether SO 2 may act as proxy for volcanic ash during volcanic eruptions for an early warning system to pre vent planes from potential encounters with volcanic plumes, sometimes not the case sev eral hours after an eruption [Thomas and Prata, 2011.]

IV. DISCUSSION
In this paper, a few representative examples of the validation activities performed within the SACS2/SMASH, European Space Agency project have been presented.SO 2 column loa ding and plume height extracted using three different algorithms analysing MetopA obser vations by the IASI and GOME2 instruments during the 15 to 26 April 2010 and 4 to 20 May 2010, Eyjafjallajökull, Iceland, eruptions and the 23 rd of May 2011, Grímsvötn, Iceland, erup tion are shown.

Figure 1 :
Figure 1: Estimated total atmospheric SO 2 load over Europe on May 23rd 2011 as seen by the ULB/IASI algorithm [left], the Oxford/IASI algorithm [middle] and the BIRAIASB/GOME2 algorithm [right].

Figure 2 :
Figure 2: Estimated total SO 2 plume height over Europe on May 23rd 2011 as deduced by the ULB/IASI algorithm [left], the Oxford/IASI algorithm [middle] and the BIRAIASB/GOME2 algorithm [right].