C b di id b i f IASCarbon dioxide observation from IASCarbon dioxide observation from IASS Del Bianco U Cortesi M GaiS. Del Bianco, U. Cortesi, M. Gai

Istituto di Fisica Applicata "Nello Carrara" del Consiglio Nazionale delle RIstituto di Fisica Applicata Nello Carrara del Consiglio Nazionale delle R

INTRODUCTIONINTRODUCTION

Th ESA h j t “A li ti f KLIMA Al ith t CO R t i l fThe ESA research project “Application of KLIMA Algorithm to CO2 Retrieval fromIASI/METOP-A Observations and Comparison with TANSO-FTS/GOSAT Products” aimedpto develop a dedicated software, based on the KLIMA inversion algorithm (originallyto develop a dedicated software, based on the KLIMA inversion algorithm (originallyproposed by IFAC CNR for the 6 cycle of ESA Earth Explorer Core Missions) suited forproposed by IFAC-CNR for the 6 cycle of ESA Earth Explorer Core Missions), suited forCO i l d i d i h ESA id b d i l i G idCO2 retrieval and integrated into the ESA grid-based operational environment GridProcessing On-Demand (G-POD) to process Level 1 data acquired by the Infraredg ( ) p q yAtmospheric Sounding Interferometer (IASI) and to perform a comparison with ThermalAtmospheric Sounding Interferometer (IASI) and to perform a comparison with ThermalAnd Near infrared Sensor for carbon Observation Fourier Transform SpectrometerAnd Near-infrared Sensor for carbon Observation Fourier Transform Spectrometer( ) b d f h h b i lli ( ) l(TANSO-FTS), on board of the Greenhouse gases Observing SATellite (GOSAT), Level 2data.data.

I d t bt i bl it t b lk i IASI d t h t i t tIn order to obtain a reasonable capacity to bulk processing IASI data, we choose to integratethe KLIMA code into the G-POD system. For this reason, we investigated an optimizedy , g pversion of the KLIMA algorithm aiming at developing a non-operational retrieval code withversion of the KLIMA algorithm, aiming at developing a non operational retrieval code withadequate features for the integration on the G POD system The optimized version ofadequate features for the integration on the G-POD system. The optimized version ofKLIMA i l d h b l d d i d h G POD i lKLIMA retrieval code has been completed and integrated on the G-POD operationalenvironment and is available for bulk processing of IASI data. Using the KLIMA inversionp g gcode integrated into the ESA G-POD it was possible to perform an extensive comparison ofcode integrated into the ESA G-POD, it was possible to perform an extensive comparison ofa selected set of IASI meas rements collocated ith SWIR TANSO FTS obser ationsa selected set of IASI measurements collocated with SWIR TANSO-FTS observations.

As part of an extension of the collaboration between IFAC-CNR and the GOSAT team inAs part of an extension of the collaboration between IFAC CNR and the GOSAT team inresponse to the second Research Announcement we performed a comparison over aresponse to the second Research Announcement, we performed a comparison over at ll d d d t t f ll t d IASI d TIR TANSO FTS b titemporally reduced dataset of collocated IASI and TIR TANSO-FTS observations overland and ocean. In this work, we present first results and preliminary conclusions from this, p p yactivityactivity.

Fi 1 Gl b l f CO d th f ll Fi 2 T t l f KLIMA Fi 3 A d COFigure 1: Global map of CO2 averaged over the fullid f 2° 2° i l

Figure 2: Total error of KLIMA-IASI CO lti li d b th

Figure 3: Averaged COf th f t tyear on a grid of 2°×2° pixels IASI CO2 multiplied by the square

f h 2 l d f iof the surface temperatu1 Kroot of the χ2 plotted as a function 1 K

of the surface temperature

COMPARISON ACTIVITIESCOMPARISON ACTIVITIES

KLIMA IASI L2 EUMETSAT IASI L2 dKLIMA IASI L2 vs. EUMETSAT IASI L2 productsA perfect coincidence in time and space exists with respect to operational data delivered byp p p p yEUMETSAT The comparison between KLIMA and EUMETSAT datasets is difficultEUMETSAT. The comparison between KLIMA and EUMETSAT datasets is difficultb t i htf d l ti d t d th b f ti t fbecause straightforward correlations do not emerge and the absence of an error estimate forone of the two results prevents a clear conclusions [Cortesi et al., 2013].p [ ]

KLIMA IASI L2 vs SWIR TANSO FTS GOSAT L2 productsKLIMA IASI L2 vs. SWIR TANSO-FTS GOSAT L2 productsTh diff t t t i h b dThree different strategies have been used:• Co-located comparison: comparison of the CO2 total column retrieved from observationsp p 2

of IASI and TANSO-FTS made in contiguous locations in time and space [Cortesi et alof IASI and TANSO FTS made in contiguous locations in time and space [Cortesi et al.,2013];2013];

d i i f h l l d i bl i l• Averaged comparison: comparison of the CO2 total column averaged on a suitable spatialand time interval (Fig. 4);a d t e te va ( g. );

• Seasonal variation comparison: comparison of the seasonal variations of CO from• Seasonal variation comparison: comparison of the seasonal variations of CO2 fromM h 2010 t F b 2011 (Fi 5)March 2010 to February 2011 (Fig. 5).

In the case of the averaged comparison the distribution of the differences between the twog pdatasets shows a negative 7,3 ppm bias of KLIMA-IASI, with a standard deviation of 7 ppmdatasets shows a negative 7,3 ppm bias of KLIMA IASI, with a standard deviation of 7 ppmand an un accounted error of KLIMA IASI of about 6 ppm with respect to the retrieval errorand an un-accounted error of KLIMA-IASI of about 6 ppm with respect to the retrieval error

d h CO h i i bili T ki i h diff A i K land to the CO2 atmospheric variability. Taking into account the different Averaging Kernels ,the negative bias is reduced to 1,17 ppm, as showed in Fig. 4.g , pp , g

KLIMA IASI L2 vs TCCON productsKLIMA IASI L2 vs. TCCON productsAs in the previous case, the comparison with TCCON stations shows a negative bias[Cortesi et al., 2013].[Cortesi et al., 2013].

CONCLUCONCLUThe main outcomes of the project can be summarized in terms of the results we have already cThe main outcomes of the project can be summarized in terms of the results we have already ci f t ti itiin future activities:• Development and implementation of a non-operation inverse model for the retrieval of CO2 pp p p 2 p• Fits of good quality obtained from KLIMA wide-band and multi-target analysis; observed geFits of good quality obtained from KLIMA wide band and multi target analysis; observed ge• On average negative bias of KLIMA retrieved CO relative to other instruments; the average• On average, negative bias of KLIMA retrieved CO2 relative to other instruments; the average

h i i h S A SO S h d d f A AS f• The comparison with SWIR TANSO-FTS showed an un-accounted error of KLIMA-IASI of• The un-accounted continuum retrieval adopted for the comparison with TIR TANSO-Fp p S• The new retrieval settings used for the TIR will be applied in the comparison with SWI• The new retrieval settings used for the TIR will be applied in the comparison with SWI

Th h i h h k D N k S i h f Chib U i i f h ll b i i idThe authors wish to thank Dr. Naoko Saitoh of Chiba University for the collaboration in providU. Cortesi et al. (2013) Finale Report of the Project ‘Sensitivity analysis and application of KL( ) p j y y pp

SI d i i h TANSO FTSSI and comparison with TANSO-FTSSI and comparison with TANSO-FTSpi L M Laurenza and F Barbarai, L.M. Laurenza, and F. BarbaraRicerche Via Madonna del Piano 10 50019 Sesto Fiorentino (Firenze) ItalyRicerche, Via Madonna del Piano 10, 50019 Sesto Fiorentino (Firenze), Italy

ANALISYS OF IASI DATA WITH KLIMA ALGORITHMANALISYS OF IASI DATA WITH KLIMA ALGORITHM

W l d t t l f 240 000 IASI t i G POD ti t i iWe analyzed a total of 240.000 IASI spectra using G-POD computing resources, retrievingthe CO2 both on land and on water and during day and night for a global geographical2 g y g g g g pcoveragecoverage.

Th i i i Fi 1 h h l b l f CO d h f llThe coverage is given in Fig. 1 where the global map of CO2 averaged over the full year ona grid of 2°×2° pixels is shown. In the retrievals, the forward model calculations fit well theg p ,spectra observed by IASI with residual differences comparable with the spectral noise Thespectra observed by IASI with residual differences comparable with the spectral noise. Theresid als are generall m ch smaller than the spectral noise Fe isolated atmosphericresiduals are generally much smaller than the spectral noise. Few isolated atmospheric

2features show average residuals that are larger than the spectral noise. The χ2 values show acorrelation with Earth’s surface temperature and are close to unity at low temperatures,correlation with Earth s surface temperature and are close to unity at low temperatures,while increase monotonically up to 3 at the highest temperatures Values greater than 3while increase monotonically up to 3 at the highest temperatures. Values greater than 3

i ll b t b i tl l t d i d t b bl i t d t doccasionally occur, but being mostly located in desert areas are probably associated to sandstorm scenarios.

The retrieval error of CO varies as a function of the surface temperature with values of 2The retrieval error of CO2 varies as a function of the surface temperature with values of 2hi h d 20 l B i i f IASIppm at high temperature and up to 20 ppm at low temperature. Best precision of IASI

observations is obtained in warm seasons and at low latitudes and the worst precision ispobtained in the cold seasons and at high latitudes A conservative estimate of the totalobtained in the cold seasons and at high latitudes. A conservative estimate of the total

t i l i bt i d b lti l i th t i l ith th t f th 2retrieval error is obtained by multiplying the retrieval error with the square root of the χ2

(see Fig. 2).( g )

Fig 3 shows the average of all retrieved CO values as a function of surface temperatureFig. 3 shows the average of all retrieved CO2 values as a function of surface temperature.O l d ( d i t ) hi h f t t i l t d ith l t th hOver land (red points) a higher surface temperature is correlated with plant growth when aCO2 sink is expected, on the other hand over ocean (blue points) a lower surface2 p , ( p )temperature increases the solubility of CO2 into the sea and produces a sinktemperature increases the solubility of CO2 into the sea and produces a sink.

O f ti Fi 4 S tt di f Fi 5 S l i ti f th CO f M h 2010 tO2 as a functionbi d

Figure 4: Scatter diagrams ofth CO2 KLIMA TANSO

Figure 5: Seasonal variation of the CO2 from March 2010 toF b 2011 i N th H i h Th thure binned over the CO2 KLIMA vs TANSO-

FTS 2 i 2° 2°February 2011 in Northern Hemisphere. The average on theH i h f h CO i d b KLIMA IASI ( d i )FTS v.2 comparison on a 2°x2° Hemispheres of the CO2 retrieved by KLIMA-IASI (red points)

pixel grid related to the whole is compared with the Hemisphere average of SWIR TANSO-year FTS CO2 products; green points refer to L2 TANSO-FTS v.2

original products while blue points refer to v.2 smoothed

KLIMA IASI L2 TIR TANSO FTS GOSAT L2 d tKLIMA IASI L2 vs. TIR TANSO-FTS GOSAT L2 productsp

A different retrieval set-up has been adopted in this comparison:p p p

• Retrieval of CO profile 010/03/08 coincident IASI and TIR TANSO-FTS measure geoloca• Retrieval of CO2 profile 80

010/03/08 coincident IASI and TIR TANSO FTS measure geoloca

IASI

• Use of the same vertical retrieval grid 40 60

eg.]

IASITANSO-FTSg

• Un accounted continuum retrieved simultaneously 0 20 40

e [de

• Un-accounted continuum retrieved simultaneously40

-20 0

itude

-60-40

Lati

-80

-180-150-120 -90 -60 -30 0 30 60 90 120 150 180IASI and TIR TANSO-FTS CO2 Profile Difference

180 150 120 90 60 30 0 30 60 90 120 150 180

Longitude [deg.]100Averaged DifferencesAveraged Differences

Standard Deviation IASI and TIR TANSO-FTS CO2 Total Column

4002

IASI 395

IASITANSO-FTS

1r] 390m]

101

Bar

385[pp

[mB

380

385

O2

re

375

380C

sur

370

375

102res 370

-80 -60 -40 -20 0 20 40 60 8010

Pr

Latitude [deg.]

The un accounted continuum retrievalThe un-accounted continuum retrievalth t d i thremoves the un-accounted error in the

KLIMA results. The new retrieval103

settings will be applied in the-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 settings will be applied in thecomparison with SWIR TANSO FTS

CO2 vmr [ppm]comparison with SWIR TANSO-FTSGOSAT d TCCON L2 d tGOSAT and TCCON L2 products.

USIONSUSIONSconsolidated and of the open and new issues we are investigating at present or plan to addressconsolidated and of the open and new issues we are investigating at present or plan to address

profile and total column from IASI/MetOp-A observationsp peographical and seasonal variability often in good agreement with expectationseographical and seasonal variability often in good agreement with expectationsed comparison with SWIR TANSO FTS shows a bias of 7 3 ppmed comparison with SWIR TANSO-FTS shows a bias of -7,3 ppmf b 6 i h h i l d h CO h i i bilif about 6 ppm with respect to the retrieval error and to the CO2 atmospheric variabilityFTS products removes the un-accounted error in the KLIMA resultsS pIR TANSO-FTS GOSAT and TCCON L2 productsIR TANSO-FTS GOSAT and TCCON L2 products

di h d d f TANSO FTS TIR dding the used dataset of TANSO-FTS TIR data.LIMA algorithm to GOSAT and OCO validation’ (ESA-ESRIN/Contract n. 21612/08/I-OL).g ( )