a trace gas detection scheme channels [cm - cnes · references [1]l. clarisse, p. f. coheur, a. j....

1
References [1] L. Clarisse, P. F. Coheur, A. J. Prata, D. Hurtmans, A. Razavi, T. Phulpin, J. Hadji-Lazaro, and C. Clerbaux. Tracking and quantifying volcanic SO 2 with IASI, the September 2007 eruption at Jebel at Tair. Atmospheric Chemistry and Physics, 8(24):7723–7734, 2008. [2] Air Crash Investigations,. http://www.natgeochannel.co.uk/Programmes/air-crash-investigation/photos/series-4. [3] A. Dudhia. Reference Forward Model (RFM) [Internet], 2008. http://www.atm.ox.ac.uk/RFM. [4] A. Dudhia, V. L Jay, and C. D Rodgers. Microwindow selection for high-spectral-resolution sounders. Appl. Opt, 41:3665–3673, 2002. [5] D. P. Edwards. GENLN2: A general line-by-line atmospheric transmittance and radiance model. Version 3.0: Description and users guide. Technical report, NCAR, January 1992. [6] Theme 3 on Atmospheric Composition of the National Centre for Earth Observations. http://www.nceo.ac.uk/documents/Theme_3_Atmospheric_ Composition__submitted_version.pdf. [7] J. J. Remedios, R. J. Leigh, A. M. Waterfall, D. P. Moore, H. Sembhi, I. Parkes, J. Greenhough, M. P. Chipperfield, and D. Hauglustaine. MIPAS reference atmospheres and comparisons to V4.61/V4.62 MIPAS level 2 geophysical data sets. Atmospheric Chemistry & Physics Discussions, 7:9973–10017, July 2007. [8] M. Rix, P. Valks, Nan Hao, J. van Geffen, C. Clerbaux, L. Clarisse, P.-F. Coheur, D.G. Loyola R, T. Erbertseder, W. Zimmer, and S. Emmadi. Satellite monitoring of volcanic sulfur dioxide emissions for early warning of volcanic hazards. Selected Topics in Applied Earth Observations and Remote Sensing, IEEE Journal of, 2(3):196–206, Sept. 2009. [9] M.Z. Salvador, R.G. Resmini, and R.B. Gomez. Detection of sulfur dioxide in airs data with the wavelet packet subspace. Geoscience and Remote Sensing Letters, IEEE, 6(1):137–141, Jan. 2009. [10] J. C. Walker and A. Dudhia. Use of the RFM to model IASI spectra. Technical report, AOPP, Clarendon Laboratory, University of Oxford, 2009. Conclusions and future work An automatic channel selection algorithm and detection scheme is implemented using IASI spectra which works well for SO 2 . The channels selected by this algorithm are almost identical to those selected by [1] differing by just one background channel. It should be possible to extend the method to other molecules. Detection of Kasatochi event First detected by GOME-2 on its overpass on 8th August 2008 with column amounts of around 150 DU [8]. The SO 2 plume from this eruption is picked out successfully using the new method. (a) This method (b) BT flag by Clarisse et al [1] Figure 5: Detection of the eruption of the Kasatochi volcano in the Aleutian Islands, Alaska, showing SO 2 on 8th August 2008, the day after the eruption began. Figure 4 shows the random and systematic errors associated with this detection scheme for background levels of SO 2 for the top four channels selected by this analysis and the channels selected by Clarisse et al. [1]. Systematic errors are well below the random noise error in both cases indicating that false detections should be completely avoided. -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 Error [DU] 1371.50 1371.75 1407.25 1408.75 1371.50 1371.75 1408.00 1407.75 Channels [cm -1 ] SFCTEM HNO3 CH4 H2O TEM Random Sytematic Total Figure 4: Error analysis for detection step calculated assuming climatological background levels of SO 2 . Results for channels selected by this analysis (1371.50, 1371.75, 1407.00, 1408.75 cm -1 ) shown in bottom row, and results for channels selected by [1] (1371.50, 1371.75, 1407.25, 1408.75 cm -1 ) shown in top row. The total systematic error N i=1 (σ i sys ) 2 is very low in both cases. Error expressed in terms of Dobson units. In the detection step, a set of weights are generated G =(K T K) -1 K T which separate the SO 2 signal from the background contribution when applied to the IASI spectra. The flag is then expressed in terms of estimated dobson units (DU est ) DU est = DU 0 (1 + G(y - y clim )) (2) Simulations The SO 2 flag was tested for a simulated plume between 12 and 22 km for events up to around 100 DU. Up to 10 DU, the estimated and actual DU are very similar. The neglect of non-linearity leads to an underestimated column for larger events. 0.01 0.10 1.00 10.00 100.00 1000.00 10000.00 SO2 VMR [ppbv] 0 10 20 30 Altitude [km] 0.1 1.0 10.0 100.0 1000.0 Actual DU 0.1 1.0 10.0 100.0 1000.0 Estimated DU Figure 3: Effects of non-linearity on estimated column amount. For small plumes, linear retrieval gives reasonable estimate of true quantity of SO 2 . Ball-park estimates of SO 2 amounts are possible for large events. Automatic channel selection Channels are selected in the in the vacinity of the SO 2 ν 3 band centred on 1362 cm -1 . Calculations are performed using the Reference Forward Model (RFM) which is a versatile line-by-line radiative transfer code developed at the University of Oxford [3] based on GENLN2 [5] with optimisations for use with IASI described in [10]. Channels are selected on the basis of maximum information content for SO 2 considering an optimal estimation type linear joint retrieval of SO 2 and brightness temperature. The first step is the selection of the ‘top-pair’, considering 100% column perturbations to a standard SO 2 profile derived from the IG2 climatology [7] and a 20 K brightness temperature (BT) offset K T = " ∂y i SO 2,col , ∂y j SO 2,col ∂y i ∂BT , ∂y j ∂BT # where the gain and random error are computed as G = S a K(S y + K T S a K) -1 S rnd =(I - GK)S a where the covariance S a contains appropriate a priori constraints on SO 2 and BT and the instrument noise covariance is given by S y . Systematic errors due to water vapour, temperature, and minor contaminants (CH 4 , HNO 3 ) are modelled using IG2 1σ column perturbations. Sensitivity to surface temperature uncertainty is modelled using a 20 K perturbation. For N sources of error the systematic error is given by δ x i = Gδ y i S sys = N X i=1 E {(δ x i ) T (δ x i )} where δ y i are the error spectra due to the perturbation of each error source. Top-pair selected assessing information content of SO 2 component of all 97020 combinations of channels in 1300–1410cm -1 range H = log 2 σ 2 rnd + σ 2 sys σ 2 a ! (1) Additional channels are then selected exploring remaining combinations using sequential estimation as described in [4] starting from top pair. 1300 1320 1340 1360 1380 1400 Wno [cm -1 ] 200 220 240 260 280 300 320 BT [K] selected channels 1300 1320 1340 1360 1380 1400 Wno [cm -1 ] -20 0 20 40 60 BT difference [K] sfc so2 h2o tem orig Figure 2: BT spectra for a simulated 100 DU SO 2 event with stratospheric plume (12–22 km) showing contribution of SO 2 and climatological 1 σ column perturbation in water vapour and atmospheric temperature, and a 20 K perturbation in surface temperature. Top four channels selected in order 1371.50, 1408.75, 1407.00, 1371.75 cm -1 . Figure 1: The finely ground particles of rock in volcanic plumes present a serious risk to aviation. Measurements of SO 2 are correlated with volcanic ash in the first few days after an eruption and a global SO 2 alert system is in operation using IASI data [8]. Reconstruction of the ‘Jackarta Inci- dent’ on 24th June 1982 when a BA Boeing 747-236B flew into a volcanic ash cloud from Mount Galunggung in West Java, Indonesia [2] (the plane eventually landed safely). Introduction Detection-only methods allow the high volume of global IASI data to be processed quickly, and are suitable for identifying the presence of species associated with special events like volcanic eruptions. The main difficulty is avoiding false detections caused by variations in the background due to param- eters that are not measured. A general detection scheme is introduced which automatically selects the best channels for the separation of the target gas from the background with an optimal set of weights. This method is an extension of brightness temperature filter methods such as that used by Clarisse et al. [1] who have devised a manually selected detection filter for SO 2 based on an equally weighted dif- ference of the brightness temperature in two carefully selected channels in the SO 2 ν 3 band (1371.50, 1371.75 cm -1 ) and two background channels (1407.25, 1408.75 cm -1 ). Abstract A general method for trace gas detection from IASI is demonstrated for volcanic SO 2 emissions. This work is performed within Theme 3: ‘Measurements of Atmospheric Composition’ of the National Centre for Earth Observation, UK [6]. [email protected] Joanne Walker and Anu Dudhia, AOPP, Clarendon Laboratory, University of Oxford, UK A trace gas detection scheme

Upload: doankhanh

Post on 18-Jan-2019

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: A trace gas detection scheme Channels [cm - CNES · References [1]L. Clarisse, P. F. Coheur, A. J. Prata, D. Hurtmans, A. Razavi, T. Phulpin, J. Hadji-Lazaro, and C. Clerbaux. Tracking

References[1] L. Clarisse, P. F. Coheur, A. J. Prata, D. Hurtmans, A. Razavi, T. Phulpin, J. Hadji-Lazaro, and C. Clerbaux. Tracking and quantifying volcanic SO2 with IASI, the

September 2007 eruption at Jebel at Tair. Atmospheric Chemistry and Physics, 8(24):7723–7734, 2008.

[2] Air Crash Investigations,. http://www.natgeochannel.co.uk/Programmes/air-crash-investigation/photos/series-4.

[3] A. Dudhia. Reference Forward Model (RFM) [Internet], 2008. http://www.atm.ox.ac.uk/RFM.

[4] A. Dudhia, V. L Jay, and C. D Rodgers. Microwindow selection for high-spectral-resolution sounders. Appl. Opt, 41:3665–3673, 2002.

[5] D. P. Edwards. GENLN2: A general line-by-line atmospheric transmittance and radiance model. Version 3.0: Description and users guide. Technical report, NCAR,January 1992.

[6] Theme 3 on Atmospheric Composition of the National Centre for Earth Observations. http://www.nceo.ac.uk/documents/Theme_3_Atmospheric_Composition__submitted_version.pdf.

[7] J. J. Remedios, R. J. Leigh, A. M. Waterfall, D. P. Moore, H. Sembhi, I. Parkes, J. Greenhough, M. P. Chipperfield, and D. Hauglustaine. MIPAS reference atmospheresand comparisons to V4.61/V4.62 MIPAS level 2 geophysical data sets. Atmospheric Chemistry & Physics Discussions, 7:9973–10017, July 2007.

[8] M. Rix, P. Valks, Nan Hao, J. van Geffen, C. Clerbaux, L. Clarisse, P.-F. Coheur, D.G. Loyola R, T. Erbertseder, W. Zimmer, and S. Emmadi. Satellite monitoring of volcanicsulfur dioxide emissions for early warning of volcanic hazards. Selected Topics in Applied Earth Observations and Remote Sensing, IEEE Journal of, 2(3):196–206, Sept.2009.

[9] M.Z. Salvador, R.G. Resmini, and R.B. Gomez. Detection of sulfur dioxide in airs data with the wavelet packet subspace. Geoscience and Remote Sensing Letters, IEEE,6(1):137–141, Jan. 2009.

[10] J. C. Walker and A. Dudhia. Use of the RFM to model IASI spectra. Technical report, AOPP, Clarendon Laboratory, University of Oxford, 2009.

Conclusions and future workAn automatic channel selection algorithm and detection scheme is implemented using IASI spectrawhich works well for SO2. The channels selected by this algorithm are almost identical to thoseselected by [1] differing by just one background channel. It should be possible to extend the methodto other molecules.

Detection of Kasatochi eventFirst detected by GOME-2 on its overpass on 8th August 2008 with column amounts of around 150DU [8]. The SO2 plume from this eruption is picked out successfully using the new method.

(a) This method (b) BT flag by Clarisse et al [1]

Figure 5: Detection of the eruption of the Kasatochi volcano in the Aleutian Islands, Alaska, showing SO2 on 8th August2008, the day after the eruption began.

Figure 4 shows the random and systematic errors associated with this detection scheme for backgroundlevels of SO2 for the top four channels selected by this analysis and the channels selected by Clarisseet al. [1]. Systematic errors are well below the random noise error in both cases indicating that falsedetections should be completely avoided.

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8Error [DU]

1371.501371.751407.251408.75

1371.501371.751408.001407.75

Cha

nnel

s [c

m-1

]

SFCTEM HNO3 CH4 H2O TEM

Random Sytematic Total

Figure 4: Error analysis for detection step calculated assuming climatological background levels of SO2. Resultsfor channels selected by this analysis (1371.50, 1371.75, 1407.00, 1408.75 cm−1) shown in bottom row, and resultsfor channels selected by [1] (1371.50, 1371.75, 1407.25, 1408.75 cm−1) shown in top row. The total systematic error(∑N

i=1(σisys)

2)

is very low in both cases. Error expressed in terms of Dobson units.

In the detection step, a set of weights are generated G = (KTK)−1KT which separate the SO2 signalfrom the background contribution when applied to the IASI spectra. The flag is then expressed interms of estimated dobson units (DUest)

DUest = DU0(1 + G(y − yclim)) (2)

Simulations

The SO2 flag was tested for a simulated plume between 12 and 22 km for events up to around 100DU. Up to 10 DU, the estimated and actual DU are very similar. The neglect of non-linearity leads toan underestimated column for larger events.

0.01 0.10 1.00 10.00 100.00 1000.00 10000.00SO2 VMR [ppbv]

0

10

20

30

Alti

tude

[km

]

0.1 1.0 10.0 100.0 1000.0Actual DU

0.1

1.0

10.0

100.0

1000.0

Est

imat

ed D

U

Figure 3: Effects of non-linearity on estimated column amount. For small plumes, linear retrieval gives reasonableestimate of true quantity of SO2. Ball-park estimates of SO2 amounts are possible for large events.

Automatic channel selection

Channels are selected in the in the vacinity of the SO2 ν3 band centred on 1362 cm−1. Calculationsare performed using the Reference Forward Model (RFM) which is a versatile line-by-line radiativetransfer code developed at the University of Oxford [3] based on GENLN2 [5] with optimisations foruse with IASI described in [10].

Channels are selected on the basis of maximum information content for SO2 considering an optimalestimation type linear joint retrieval of SO2 and brightness temperature. The first step is the selectionof the ‘top-pair’, considering 100% column perturbations to a standard SO2 profile derived from theIG2 climatology [7] and a 20 K brightness temperature (BT) offset

(KT =

[∂yi

∂SO2,col,

∂yj

∂SO2,col

∂yi

∂BT,

∂yj

∂BT

] )where the gain

and random error are computed as

G = SaK(Sy + KTSaK)−1

Srnd = (I−GK)Sa

where the covariance Sa contains appropriate a priori constraints on SO2 and BT and the instrumentnoise covariance is given by Sy. Systematic errors due to water vapour, temperature, and minorcontaminants (CH4, HNO3) are modelled using IG2 1σ column perturbations. Sensitivity to surfacetemperature uncertainty is modelled using a 20 K perturbation. For N sources of error the systematicerror is given by

δxi = Gδyi

Ssys =

N∑i=1

E{(δxi)T (δxi)}

where δyi are the error spectra due to the perturbation of each error source. Top-pair selected assessinginformation content of SO2 component of all 97020 combinations of channels in 1300–1410cm−1

range

H = log2

(σ2

rnd + σ2sys

σ2a

)(1)

Additional channels are then selected exploring remaining combinations using sequential estimationas described in [4] starting from top pair.

1300 1320 1340 1360 1380 1400Wno [cm-1]

200220

240

260

280

300320

BT

[K]

selected channels

1300 1320 1340 1360 1380 1400Wno [cm-1]

-20

0

20

40

60

BT

diff

eren

ce [K

]

sfc so2 h2o tem orig

Figure 2: BT spectra for a simulated 100 DU SO2 event with stratospheric plume (12–22 km) showing contribution ofSO2 and climatological 1 σ column perturbation in water vapour and atmospheric temperature, and a 20 K perturbation insurface temperature. Top four channels selected in order 1371.50, 1408.75, 1407.00, 1371.75 cm−1.

Figure 1: The finely ground particles of rock in volcanic plumes present a serious risk to aviation.Measurements of SO2 are correlated with volcanic ash in the first few days after an eruption and aglobal SO2 alert system is in operation using IASI data [8]. Reconstruction of the ‘Jackarta Inci-dent’ on 24th June 1982 when a BA Boeing 747-236B flew into a volcanic ash cloud from MountGalunggung in West Java, Indonesia [2] (the plane eventually landed safely).

Introduction

Detection-only methods allow the high volume of global IASI data to be processed quickly, and aresuitable for identifying the presence of species associated with special events like volcanic eruptions.The main difficulty is avoiding false detections caused by variations in the background due to param-eters that are not measured. A general detection scheme is introduced which automatically selects thebest channels for the separation of the target gas from the background with an optimal set of weights.This method is an extension of brightness temperature filter methods such as that used by Clarisse etal. [1] who have devised a manually selected detection filter for SO2 based on an equally weighted dif-ference of the brightness temperature in two carefully selected channels in the SO2 ν3 band (1371.50,1371.75 cm−1) and two background channels (1407.25, 1408.75 cm−1).

AbstractA general method for trace gas detection from IASI is demonstrated for volcanic SO2 emissions. This work is performedwithin Theme 3: ‘Measurements of Atmospheric Composition’ of the National Centre for Earth Observation, UK [6].

[email protected] Walker and Anu Dudhia, AOPP, Clarendon Laboratory, University of Oxford, UK

A trace gas detection scheme