Comparison of Volcanic So2 Flux Measurements From Satellite and From the Novac Network

The main objective of the NOVAC project (Network for Observation of Volcanic and Atmospheric Change) is to establish a global network of stations for the quantitative measurement of volcanic gas emissions (in particular SO2 and BrO) by UV absorption spectroscopy. The data from the network (more than 21 volcanoes are currently monitored) are primarily used for risk assessment and volcanological research, but the data are also valuable for the study of tropospheric and stratospheric gas composition (SO2, NO2, CH2O, BrO and O3). Since volcanic SO2 is also monitored from satellite (e.g. the SACS service, http://sacs.aeronomie.be/) the NOVAC project provides an excellent opportunity to explore and inter-compare the different satellite SO2 data-sets under volcanic conditions. Furthermore the NOVAC ground-based data can be used to validate satellites estimates of gas flux emissions. In this work, we present an investigation focusing on GOME-2 and OMI SO2 data sets. Their mutual consistency is analysed and comparisons are performed with the NOVAC ground-based network measurements. A statistical study on the whole NOVAC dataset is performed, comparing mass estimation from satellites and flux measurements from the ground. A case study over Etna involving OMI (Ozone Monitoring Instrument) and ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) illustrates the impact of spatial inhomogeneities in the SO2 field inside the area covered by an OMI pixel. Moreover, this study illustrates the importance of external information (such as the height of the volcanic plume) to reduce the error on the SO2 estimation. 1. CONTEXT / VOLCANIC SO2 DATASETS Within the NOVAC project (Galle et al., 2010) a network of ground-based automated instruments for the measurement of volcanic SO2 emissions has been set up. An overview of the volcanoes monitored in Latin America is presented in figure 1. The possibility of monitoring volcanoes activity from space with UV-Vis sounders like OMI and GOME-2 has been shown, both for explosive eruptions and for degassing cases (e.g., Yang et al., 2007, Carn et al., 2008, Rix et al., 2009 and references therein). More information on the instruments, the main retrieval settings and references for the data used in this work are given in table 1. Figure 1: Location of the NOVAC volcanoes in Latin America. Instrument reference GOME-2 Global Ozone Monitoring Experiment (Munro et al., 2006) OMI Ozone Monitoring Instrument (Levelt et al., 2006) ASTER Advanced Spaceborne Thermal Emission and Reflection Radiometer (Pieri and Abrams, 2004) Platform and launch On MetOp-A, since 2006 On AURA, since 2004 On Terra, since 1999 Overpass time ~9h30 UT ~12h-13h30 UT 9h59 GMT Pixel resolution 80x40km2 (at nadir – swath of 1920km) 13x24km2 (at nadir -swath of 2600 km, CTP from 1 to 60) 0.09x0.09km2 (whole image 60x60km2) SO2 data origin Operational product from O3M-SAF. Columns given for SO2 content at 2km, 6km, 15km OMSO2 operational product from NASA. Columns given for SO2 content at 0.9km (PBL, pollution), 2.5km (TRL), 7.5km (TRM) , 17km (STL) Data obtained from Robin Campion (ULB), personal communication. SO2 detection limit 3σ noise: ~3e+16 molec/cm2 (1.15DU) 3σ noise: ~1.61e+16 molec/cm2 (0.6DU) 1σ noise: ~5e+17 molec/cm SO2 retrieval reference Valks et al., 2010 Rix et al., 2009 Yang et al., 2007 PBL: Krotkov 2006 Campion et al., 2010 Table 1: Details about the different satellite SO2 data used in this study. SO2 data from OMI and GOME-2 instruments are extracted in the vicinity of the NOVAC volcanoes, from 2007 onward, with the aim to (1) investigate the possibilities of using satellites for monitoring volcanic gas emissions and (2) assess the quality of the data compared to the NOVAC ground-based network. In this study, the consistency of the two satellite data sets is explored for a few volcanic eruption cases, and the SO2 columns are transformed into SO2 masses and compared with the NOVAC fluxes. A case study comparing OMI data to high resolution ASTER data (see table 1) is also presented here, in order to further explore the spatial gradients within an OMI pixel (case over Etna) and to calculate and compare fluxes from the 2 instruments. 2. OMI AND GOME-2 COMPARISONS SO2 data from OMI and GOME-2 are extracted and daily maps are created around every NOVAC station. An example is shown in figure 2. This type of maps allows a first visualisation of the coherence between the retrievals of the two satellites and of the possible transport of the volcano plume between the two overpass times (a few hours). The information on the daily maximum SO2 column value of OMI and GOME-2 is included in the map, as well as an estimation of the total SO2 mass in the area. The calculation of the SO2 mass is performed by selecting all the pixels that are above a threshold value (designated with the “fi” index), and by transforming the vertical columns [molec/cm2] into mass through equation (1): A SO fi notfi fi N M AirPix VCD mean VCD Mass 2 )) ( (     (1) [Kg] [molec/cm2] [cm2] [kg/molec] MSO2 and NA are respectively the molar mass of SO2 and the Avogadro number (MSO2= 64 g/mol and NA= 6.022e23 mol/molec). The threshold values for OMI and GOME-2 (respectively 0.6 DU and 1.15 DU) have been defined as the noise level of the SO2 VCD over a clean area in the Pacific Ocean (-30° lat, -120° long) over one month. The value found for OMI is coherent with previous literature studies (Carn et al., 2008). GOME-2 data are noisier than OMI data which, is partly due to instrumental characteristics and partly to differences in the SO2 algorithms (e.g., different fitting windows). Figure 3 presents two examples of the comparisons of OMI and GOME-2 SO2 data over Nevado del Huila and Tungurahua volcanoes, during 2008, showing a good agreement of columns and masses. However, the OMI data suffer from the so-called “row anomalies”, affecting L1B and L2 data (http://www.temis.nl/docs/omi_warning.html, http://www.knmi.nl/omi/research/product/rowanomalybackground.php, Claas et al. 2010), which results in a progressive degradation of the SO2 data product. An example of how this degradation affects the SO2 field is presented in figure 4 for a case over Nevado del Huila in November 2009. The main difficulty in handling these OMI-anomalies is that they are not stable in time, and several major changes occurred since the first appearance in June 2007. Efforts to flag and correct the affected data are currently on-going, but so-far not implemented in the OMSO2 product. The different evolution of the affected cross-track positions (CTP) over time do not allow a simple removal of pre-defined pixels (a manual check of the scene is needed) and the global comparisons can still be affected by an incorrect removal. Figure 2: Map of the SO2 column values (in DU) around Popocatepetl volcano (Mexico) on 22 November 2008, from OMI instrument (left panel) and GOME-2 instrument (right panel). Figure 3: Examples of the comparison of OMI and GOME-2 over Nevado del Huila (Colombia) and Tungurahua (Ecuador) in 2008. The upper panels present the daily maximum SO2 column values, and the lower panel the corresponding SO2 masses. Figure 4: as figure 2, but around Nevado del Huila volcano (Colombia) on 27 November 2009. Part of the OMI data are affected by the „cross-track row anomaly“ and should be excluded from the mass calculation (in this case, pixels with a cross-track position (CTP) from 27 to 40). 3. PRELIMINARY COMPARISONS WITH NOVAC The masses calculated from the satellite data are then compared with the SO2 flux measured from ground-based NOVAC stations. So far, only data from ScanDOAS instrument at Tungurahua during February 2008, MobileDOAS data from Nevado del Huila since 2007 and ScanDOAS data at Vulcano island since 2008 have been investigated. Preliminary results are presented in figure 5, showing e.g. good agreement in Tungurahua for OMI, GOME-2 and NOVAC data. Over Nevado del Huila the mobileDOAS fluxes are often much larger than the mass seen by the satellites. Very good agreement in conditions of very low SO2 emission (Vulcano Island) is obtained for OMI, but larger differences are obtained with GOME-2, which probably reflects the higher noise level of GOME-2 than OMI. Moreover, GOME-2 has larger pixels than OMI (40x80km2 vs 13x40km2 at best) and can therefore be contaminated by Etna emissions (which is at ~50km of Vulcano Island). Note also that both satellite datasets have been retrieved with the TRM SO2 columns, assuming a plume height at 6 or 7.5 km, while Vulcano is a degassing volcano and a lower emission height would probably be more realistic. Figure 5: Time series of OMI and GOME-2 masses comparisons vs ground-based SO2 fluxes over Tungurahua in Feb. 2008, Nevado del Huila since 2007 and Vulcano Island since 2008. From these preliminary comparisons it is difficult to conclude on the validity of the comparison of the satellite-based masses with the ground-based fluxes. More ground-based data are necessary to extend this comparison. Moreover, large uncertainties on both types of data exist: ground-based fluxes depend strongly on wind speed and plume altitude choices (Galle et al. 2010, Salerno et al. 2009), while for the satellites, clouds can mask part of the scene, and an assumption has to be made for the plume altitude (both OMI and GOME-2 datasets provide at least 3 types of SO2 VCD, with different assumed SO2 altitude depending on the assumed eruption type – see table 1). Uncertainties in the mass calculation also exist (related to data re-gridding, inclusion of a possible second OMI overpass orbit and choices of the CTP for the elimination of anomalies; Carn et al., 2008), and different values can be obtained for the same scene (e.g. http://so2.umbc.edu/omi). 4. OMI AND ASTER COMPARISONS A study focusing on the comparison of the OMI SO2 values with the high resolution ASTER data has been started, focusing on the different perception o