A new scheme for sulfur dioxide retrieval from IASI measurements: application to recent volcanic eruptions. Elisa Carboni 1, Roy Grainger 1, Joanne Walker 1, Anu Dudhia 1, Richard Siddans 2 (1) University of Oxford, Oxford, United Kingdom. (2) Rutherford Appleton Laboratory, Didcot, United Kingdom.

IASI measurements introduction: IR spectra - SO2 absorption band SO2 retrieval: Optimal Estimation and covariance matrix Fast forward model Sensitivity to ash and cloud Preliminary results: Evolutions of Eyjafjallajökull eruption (April-May 2010) Grimsvötn eruption (May 2011) Puyehue-Cordón Caulle eruption (June 2011) Summary and further work Outline

Thermal infrared spectra Wavenumber [cm-1] Brightness Temperature [k] SO2 absorption bands 1 3 ash

stratosphere troposphere BTD (std.atm. - plume)

- IASI is sensitive to both the amount of SO2 and the altitude of the plume - amount and altitude have different spectral signature => we attempt to retrieve both - Note that getting the altitude correct is important not just for itself, but also in order to get the correct amount of SO2, since the signal depends strongly on altitude. - often near real time algorithm make assumption about height of the plume, and this can lead to large error in the retrieved amount this applies to both UV and IR methods

y is the measurement vector, x the state vector, F forward model, S y measurement error vector Usually: y would contain the IASI measured radiances S y would represent noise on those measurements x would contain all atmospheric + surface parameters that affect these radiances and are imperfectly known - in this case including H2O,T, cloud and many other minor species, as well as SO2. However: - we're not interested in most of these potential state variables - SO2 is very rarely present in significant amounts in the IASI spectra (except during volcanic eruptions and degassing regions) - H2O,T are well predicted by Met data - Other gases have features which are spectrally uncorrelated with SO2 OPTIMAL ESTIMATION APPROACH physics, radiative transfer S y (i,j) = [(y mi -y si )-(y mi -y si )][(y mj -y sj )-(y mj -y sj )] we can greatly simplify the retrieval problem by including in the S y covariance matrix all the differences between the IASI observed radiances and radiances predicted by a forward model, based on ECMWF met data. and...

4 day - 14 orbit a day-> 5x10 6 pixels Global April May We compute the error covariance S y using similar conditions but when we know there is no SO2 in the atmosphere. This will give a measurement covariance which allows deviations due to cloud and trace- gases but still enable the SO2 signature to be detected. For Eyja eruption we consider the data, in the same geographic region, of the same month the year before (eg. April or May 2009) We also use a global covariance matrix, considering all the data of 4 days (1 every season) of 2009

Fast forward model State vector: - Total column amount of So2 - Altitude H - Thickness s - Surface temperature Ts H s + ECMWF profile (temperature, h2o, p, z) fast radiative transfer (RTTOV + SO2 RAL coefficients) FM IASI simulated spectra Ts IASI measurements OE retrieval best estimate of stare vector: SO2 amount, plume altitude, Ts

Ash Re=2 [  m], h=3 [km] Dust Re=2 [  m], h=3 [km] Water Cloud Re=10 [  m], h=3 [km] IASI forward model can include aerosol and cloud. Obtained with refractive index measured by Daniel Peters. Ash and cloud effects? To test if the retrieval is sensitive to ash/cloud etc...

Sensitivity of retrieval to presence of ash and cloud Retrieval using simulations with ASH at different optical depth and altitude SO2 underestimate for thick ash high cost !! SO2 signal ‘covered’ from cloud within or above the plume OK for cloud below the plume!!! grrr

IASI - SO2 [DU] Eyjafjallajökull eruption 14 April - 17 May 2010

From (Thomas and Prata, ACP 2011) SO2 plume mass [Tg] Total mass of So2 in the atmospheric plume obtained assuming an average area of 25 x 25 km for any IASI pixel The error bar is in the worse scenario of correlated errors: obtained as sum of all the pixels errors (overestimate in case of independent error)) IASI total mass of SO2

IASI

IASI - CALIPSO ~ 3h difference

Grimsvötn eruption May 2011

Puyehue-Cordón Caulle eruption June 2011

Minimum error estimate surface contrast = skin temperature - temperature of the first atm. layer std. atm. profiles assumption: we know the altitude of the plume. considering 60 overpass a month => error reduced of 1/sqrt(60) [g/m2][km] SO 2 monthly errors 1 DU = g/m 2

Ecuador monthly mean - degassing The monthly mean SO 2 column amount for January The black triangle indicates the location of the Tungurahua volcano. Error on monthly mean

Summary and further work New scheme for IASI developed to retrieve height and amount of SO2 Uses our new detection scheme applied to pixels for the full retrieval Initial results seem to compare reasonably to other observations. Though sometimes see much larger amounts then uv methods Thick ash can affect the retrieval, recognizable from cost >2 Cloud at the same altitude or above the plume mask the So2 signal Retrieving the ash and cloud (optical depth, altitude and effective radius) properties is possible and subject of parallel work at AOPP & RAL. Next steps: - SO2 comparison with ground and others satellite measurements - Altitude compare with Models and more comparison with CALIPSO - Combined retrieval IASI + G0ME-2 THANK YOU!

IASI 11:30 SO 2 [DU]H [hPa] From (Thomas and Prata, ACP 2011)

Error analysis ‘true state’ DXa=[100, 1000, 1, 20] Xa=[ 0.5, 400, 100, 290] NO thermal contrast +10 o thermal contrast SO2, H, s, Ts linear error on the ‘true’ state obtained as:

‘true state’ NO thermal contrast +10 o thermal contrast Error analysis retrieved state

y is the measurement vector, x the state vector, F forward model, S y measurement error vector Usually: y would contain the IASI measured radiances Sy would represent noise on those measurements x would contain all atmospheric + surface parameters that affect these radiances and are imperfectly known - in this case including H2O,T, cloud and many other minor species, as well as SO2. However: - we're not interested in most of these potential state variables - SO2 is very rarely present in significant amounts in the IASI spectra (except during volcanic eruptions and degassing regions) - H2O,T are well predicted by Met data - Other gases have features which are spectrally uncorrelated with SO2 OPTIMAL ESTIMATION APPROACH physics, radiative transfer S y (i,j) = [(y mi -y si )-(y mi -y si )][(y mj -y sj )-(y mj -y sj )] we can greatly simplify the retrieval problem by including in the Sy covariance matrix all the differences between the IASI observed radiances and radiances predicted by a forward model, based on ECMWF met data. and...

Sensitivity of retrieval to presence of ash

Sensitivity of retrieval to presence of clouds

Global Ozone Monitoring Experiment (GOME-2) ESA/EUMETSAT Scanning spectrometer, nm, resolution nm, 960 km or 1920 km swath, resolution 80 x 40 km. Global coverage can be achieved within one day. High-resolution Infra-Red Sounder (HIRS/4) NOAA 20-channel optical/IR filter-wheel radiometer, 2000 km swath, IFOV 17.4 km (nadir). 35 kg, 2.9 kbit/s, 21 W. Replaced on MetOp-C by IASI. Infrared Atmospheric Sounding Interferometer (IASI) CNES/EUMETSAT Fourier-transform spectrometer, µm in three bands. Four IFOVs of 20 km at nadir in a square 50 x 50 km, step-scanned across track (30 steps), synchronised with AMSU-A km swath. Resolution 0.35 cm-1. Radiometric accuracy K. Global coverage will be achieved in 12 hours. METeorological OPerational satellite programme (MetOp)‏ European polar-orbiting meteorological satellite. Operational in May 2007 First of tree polar satellite system (EPS) that will cover 14 years Equator crossing time 9:30 local time Advanced Very High Resolution Radiometer (AVHRR/3) NOAA 6-channel visible/IR ( µm) imager, 2000 km swath, 1 x 1 km resolution. Global imagery of clouds, ocean and land. 35 kg, 622/39.9 kbit/s (high/low rate), 27 W.

Eyjafjallajökull eruption - MIRS stereo height vs IASI 7 May 16 May

Sensitivity of retrieval to presence of ash

Sensitivity of retrieval to presence of clouds

Eyjafjallajökull eruption - SO2 amount IASI vs OMI

Night time AIRS SO2 retrieval at ~03:00 UT

The optimal estimate of x taking into account total measurement error may be computed as: Sytot is computed considering an appropriate ensemble of N measured spectra to construct an estimate of total measurement error variance-covariance Syobs [Rodger 2000] OE detection theory [ Walker, Dudhia, Carboni, Atmos. Meas. Tech. Discuss., 2010 ]

Brightness Temperature [k] Reference spectra - IASI spectra

SO2 detection Morning of 15th April 2010 as observed using OMI, GOME-2 and SCIAMACHY near-real-time total column SO2 retrievals as implemented in the Support to Aviation Control Service (SACS Web, 2010). 15 April - morning v1 SO2 fractional enhancementv3 SO2 fractional enhancement

Initial state estimate: x 0 A priori: x a Run forward model: f(x i ) Compare to J = [y - f(x i )]S e -1 [y - f(x i )] + measurements (y): [x i - x a ]S a -1 [x i - x a ] Update state: x i → x i+1 (Levenburg-Marquardt) Stop when:  J is small, or when i is large. OPTIMAL ESTIMATION[Rodgers 2000] NB Optimal estimation method provides quality control and error estimate RETRIEVAL METHOD

Total mass of SO2 SO2 plume mass [Tg] AprilMay Julian day Total mass of So2 in the atmospheric plume obtained assuming an average area of 25 x 25 km for any IASI pixel The error bar is in the worse scenario of correlated errors: obtained as sum of all the pixels errors (overestimate in case of independent error)