Radiometric and spectral inter-comparison of IASI performed at CNES Presented by Denis Jouglet (CNES) denis.jouglet@cnes.fr Contributions from: - CNES: C. Maraldi, E. Jacquette, T. Phulpin - LMD: N. Scott, L. Crépeau, R. Armante - C-S: C. Quang, X. Lénot - JC Calvel, C. Baqué GSICS Annual Meeting, 16-20 March 2015, New Delhi, India

Introduction Objectives of the inter-comparison For the IASI TEC: External monitoring of the IASI calibration Participation to the IASI-B in-flight commissioning and routine Ensure the continuity of IASI-A, IASI-B and IASI-C For the users (Climatology, Meteorology, GSICS): To ensure the consistency of the IASI calibration with the TIR sensors community Checks the long term data quality 3 independent methodologies / tools Statistics on common geophysical scenes (SNOs and quasi-SNOs) observed by two sensors, to assess the calibration difference only Quasi-SNOs for IASI-A / IASI-B, SNOs for IASI-A(B) / CRIS, IASI-A(B) / AIRS Massive means of IASI-A and IASI-B Comparison of measurements with simulations with NWP + TR (4AOP) for IASI Progresses Methodologies 2 & 3 close to be operational Update with more than 2 years of data now

Introduction IASI-A / IASI-B direct inter-comparison by quasi-SNOs IASI / CRIS, IASI / AIRS direct inter-comparison by SNOs IASI-A / IASI-B inter-comparison via CRIS & AIRS IASI-A / IASI-B inter-comparison by massive means IASI-A / IASI-B inter-comparison via NWP+4AOP Conclusions

Introduction IASI-A / IASI-B direct inter-comparison by quasi-SNOs IASI / CRIS, IASI / AIRS direct inter-comparison by SNOs IASI-A / IASI-B inter-comparison via CRIS & AIRS IASI-A / IASI-B inter-comparison by massive means IASI-A / IASI-B inter-comparison via NWP+4AOP Conclusions

Methodology for direct IASI-A / IASI-B “Similar” scenes: IASI-A and B are on the same orbit with a 180° shift  Numerous common observations (CO) between 2 consecutive tracks, but: never simultaneous: ~50min temporal shift off-nadir: from 0° to 39°, opposite angles Selection on the most relevant scenes Use of geoloc., geom., IIS, AVHRR, ECMWF data Focus on stable and homogeneous scenes = Night, mostly oceans, 0% or 100% clouds Balance “A before B” and “A after B” For each common observation Focus on the central area (same atmospheric thickness) Regional averaging of the soundings (300*300km) ΔT calculated at elementary channel level Mean and stdev computed over the dataset IASI-A Track 2 IASI-B Track 1 IASI-A Track 1

Direct IASI-A / IASI-B inter-comparison: results Biases and standard deviation over the selected dataset (homogeneous and stable scenes, night, as many “A before B” as “A after B”) 2013 With 1 year of data: Biases < ~0.1K  Very good cross calibration Compliant with the radiometric absolute specification of 0.5K Highest bias in B1  Shape still under investigation 2014

Temporal evolution The inter-comparison IASI-B / IASI-A is very stable with time B1 B3 B2

Trend with other parameters ● Is the effect in B1 due to non-linearity? Plot of NEDT vs BT (spectrally integrated in B1, B2 and B3) + Sliding means Small increase in B1, amplitude -0.1K (cold scenes) to 0 (warm scenes) Slight non-linearity residuals between IASI-A and B B1 B2 B3

Direct IASI-A / IASI-B spectral inter-comparison Specification for each IASI: Δν/ν < 2 ppm Methodology: Definition of 30 spectral windows For each window, cross-correlation of the IASI-A and -B spectra for different spectral shifts The maximum of correlation gives the actual spectral shift Based on the same dataset as the radiometric calibration Results:  Performances largely compliant with the requirement  Very stable with time (after cal/val)

Introduction IASI-A / IASI-B direct inter-comparison by quasi-SNOs IASI / CRIS, IASI / AIRS direct inter-comparison by SNOs IASI-A / IASI-B inter-comparison via CRIS & AIRS IASI-A / IASI-B inter-comparison by massive means IASI-A / IASI-B inter-comparison via NWP+4AOP Conclusions

Reminder of AIRS, IASI, CRIS characteristics Typical IASI, AIRS and CRIS spectra

Methodology for IASI / AIRS, IASI / CRIS Similar scenes: SNOs (Simultaneous Nadir Overpasses) Tolerance in simultaneity : 20 min ~30 scenes every 3 days for IASI / AIRS (12000 in 5 years) Always at high latitudes Spatial match:  Regional averaging of the soundings pixels over a 300km*300km area around the orbit crossing point Spectral match: Construction of 33 broad pseudo-bands Each PB = intelligent averaging of ~100 elementary channels to get the similarity of the PB spectral functions The AIRS missing channels and varying spectral resolution are considered when calculating the IASI coefficients NB: the convolution of IASI by the CRIS or AIRS SRFs has been performed but is still under exploitation For each pseudo-band,

IASI-A / CRIS & IASI-B / CRIS inter-comparison 2014 Biases and standard deviations (no filtering) Biases < ~0.2K  Very well cross calibrated Same shape, highest bias in B1, stronger for IASI-B. Spectral slope? Similar datasets of IASI and CRIS:

IASI-A / CRIS & IASI-B / CRIS inter-comparison 2013 Biases and standard deviations (no filtering) Similar to 2014

IASI-A / AIRS & IASI-B / AIRS inter-comparison 2014 Biases and standard deviations (no filtering) Biases < ~0.2K  Very well cross calibrated Same shape, highest bias in B1, stronger for IASI-B. Atmospheric shape? Similar datasets between IASI and AIRS:

IASI-A / AIRS & IASI-B / AIRS inter-comparison 2013 Biases and standard deviations (no filtering) Similar to 2014

Temporal Evolution of IASI/AIRS and IASI/CRIS B1 B2 B3 IASI-A / CRIS IASI-B / CRIS IASI-A / AIRS IASI-B / AIRS  All couples are very stable with time (a bit lower in B1)

Non-linearity effects of IASI/AIRS and IASI/CRIS ? Trend of B1 NeDT wrt B1 BT  No obvious trend?

Introduction IASI-A / IASI-B direct inter-comparison by quasi-SNOs IASI / CRIS, IASI / AIRS direct inter-comparison by SNOs IASI-A / IASI-B inter-comparison via CRIS & AIRS IASI-A / IASI-B inter-comparison by massive means IASI-A / IASI-B inter-comparison via NWP+4AOP Conclusions

Indirect IASI-A / IASI-B through AIRS and CRIS Combination of IASI / AIRS and IASI / CRIS for IASI-B / IASI-A IASI-B / CRIS – IASI-A / CRIS 2014 (1 year) IASI-B / AIRS – IASI-A / AIRS 2014 (1 year) IASI-B / CRIS – IASI-A / CRIS 2013 (1 year) IASI-B / AIRS – IASI-A / AIRS 2013 (1 year)  Very similar from year to year

Indirect IASI-A / IASI-B through AIRS and CRIS Combination of IASI / AIRS and IASI / CRIS for IASI-B / IASI-A IASI-B / CRIS – IASI-A / CRIS 2014 (1 year) IASI-B / AIRS – IASI-A / AIRS 2014 (1 year) All biases agree: ~0.1K  Confirms the very good cross calibration Always an effect on B1, IASI origin? Small differences in B1: dataset selection, e.g. colder? Direct IASI-B / IASI-A in PB

Introduction IASI-A / IASI-B direct inter-comparison by quasi-SNOs IASI / CRIS, IASI / AIRS direct inter-comparison by SNOs IASI-A / IASI-B inter-comparison via CRIS & AIRS IASI-A / IASI-B inter-comparison by massive means IASI-A / IASI-B inter-comparison via NWP+4AOP Conclusions

Difference of massive means IASI-B - IASI-A (from T. Phulpin) Independent method: averaging of all L1C radiance spectra for IASI-B and IASI-A Hyp: in 1 year, both sounders globally sees the same scenes Data from Feb 2014 to Jan 2015 ~65% cloudy; surf. temp. distrib. (1σ): 250K-290K Difference IASI-B – IASI-A (NEDT) from these mean radiances: Same amplitude (bias ~0.1K in B1) and same shape as IASI-B / IASI-A from SIC Influence of the dataset: here also the input dataset may be a bit different between A and B (Ext. Cal., different AVHRR cloud flags, etc.) Massive means on all scenes

Difference of massive means IASI-B - IASI-A (from T. Phulpin) Independent method: averaging of all L1C radiance spectra for IASI-B and IASI-A Hyp: in 1 year, both sounders globally sees the same scenes Data from Feb 2014 to Jan 2015 ~65% cloudy; surf. temp. distrib. (1σ): 250K-290K Difference IASI-B – IASI-A (NEDT) from these mean radiances: The AVHRR cloud mask differs from MetOp-A to MetOp-B, causing larges differences in massive means (calibration of channel 5?) Massive means on all scenes Massive means on clear scenes Massive means on cloudy scenes

Introduction IASI-A / IASI-B direct inter-comparison by quasi-SNOs IASI / CRIS, IASI / AIRS direct inter-comparison by SNOs IASI-A / IASI-B inter-comparison via CRIS & AIRS IASI-A / IASI-B inter-comparison by massive means IASI-A / IASI-B inter-comparison via NWP+4AOP Conclusions

LMD Protocol (*) for the Quality assessment of satellite L1 and L2 data: Stand Alone Approach (*) Since mid 1990s and the NOAA/NASA TOVS Pathfinder Programme Scene classification: Airmass? Clear? Cloud? Aerosols? Obs’d Satellite Data (BTs) ECMWF Analysis + Simulated Satellite Data (BTs) Statistics Residuals « simulated_observed » BTs Land/Sea/Day/Night/Secant, Airmass…

Work dedicated to IASI The tool IASI stand-alone is being operationalized to be operated by CNES 1 year has been processed (Dec 2013 – Nov 2014) Space (2.5km) and time coloocated with ECMWF analysis (0h, 6h, 12h, 18h; 137 vertica levels) but future with reanalysis or ARSA Entire globe -180°:+180° / -90°:90° Here only see at night, but future land/sea and day/night Only clear scenes (clouds determined by the LMD Cloud/Aerosol Detection Alg. Based on MetOp IASI/AMSU/MHS/HIRS data Done for IASI-A and IASI-B  Double differences

Calc – Obs (IASI-A, IASI-B) and double differences The inter-comparison IASI-B – IASI-A by double differences is almost< 0.1K Exceptions where the TR is not well modeled No larger bias in B1, but the scene distribution is different from quasi-SNO method

Residuals for B1 wrt Brightness Temperature For Channel 894.25 cm-1 Left : IASI-A, Right : IASI-B In cyan : mean +-stdv of residuals within 10 BTs classes [210:310:10] (in K) The trend wrt brightness temperature seen by quasi-SNO method is not clear here

CONCLUSIONS ● Progresses: 3 operational tools, independent methodologies Trends confirmed with >2 years of data mission Major result: very accurate cross-calibration! IASI-B very close to IASI-A (bias < ~0.1K)  continuity of the IASI mission IASI / AIRS / CRIS: Bias between 0K and 0.2K, < radiometric absolute specification of 0.5K All are very stable with time 3 methodologies agree: very confident Larger bias in IASI B1, stronger in IASI-B: non-linearity in IASI? The observed bias is still high for climatic time series On-going work: Go further in the interpretation of the shape of the bias curves: Impact of the linearity tables and maybe update in IASI board processing Get the bias curves for each IASI pixel Fine tuning of the selection rules to reduce the IASI-B/IASI-A quasi-SNOs dataset IASI complete uncertainty budget (GRWG_14.13)

Thank you for your attention

Backup slides

Methodology for direct IASI-A / IASI-B Similar scenes: IASI-A and B are on the same orbit with a 180° shift  Numerous common observations between 2 consecutive tracks, but: never simultaneous: ~50min temporal shift off-nadir: from 0° to 39°, opposite angles The numerous common observations makes a pre-filtering compulsory  On the most relevant scenes IASI-A Track 2 IASI-B Track 1 IASI-A Track 1 Spatial match: Regional averaging of the soundings (area 300km * 300km or less) Spectral match: ΔT calculated at elementary channel level  Mean and stdev computed over the dataset

IASI-A / IASI-B : selection of scenes Selection aims at reducing the effects of : 50 min time delay Need for stable scenes Need for homogeneous atmospheres (same atm. even if moving) Off-nadir geometry = The lines of sight A & B are always different Focus on the central area  same atmospheric thickness Need for homogeneous atmospheres (typical size ~16km) Selection by a relevance index Assessment of several parameters, each parameter is given a mark, then a global mark is computed with weights For homogeneous scenes: Low inter-pixel and intra-pixel variance of the IIS imager for A & B Clouds & snow: none or full in A & B For stable scenes: Focus on oceans at night Low differences in IIS imager A & B temperatures Clouds & snow: same amount between A & B Low variations in ECMWF profiles (“Geophysical NeDT”) “Geophysical” daily NeDT 0.74K 19.92K

IASI-B BB and CS1 measurements The radiometric calibration uses two reference points (internal Cold Space and warm black body) CS is colder for IASI longest wavelengths  The non-linearity is likely to be higher in IASI longest wavelengths

IASI / AIRS spectral matching Work with IASI L1C, AIRS L1B Method: 33 broad pseudo-bands (PBs) 1 PB = summation of ~100s of elementary channels (most widths between 23 and 63 cm-1) Reduces noise and spectral resolution differences AIRS spurious channels: taken into account through a weighted summation of the IASI channels (weighs are computed to make the resulting PB response functions similar in IASI and AIRS) Comparison of ΔT = TIASI - TAIRS in each PB Other methods under progress similar channels (statistical similar behavior) convolved channels Instrumental functions of one PB for AIRS (including spurious channels), for IASI without weighting in the channels summation and for IASI with weighting

Necessity of tuning the dataset selection Nominal case (night, oceans, as many “A before B” as “A after B”): Only “A before B”: Stringent quality index:  Standard deviation mainly due to geophysics Only diurnal data:  Major impact of the input dataset, all parameters must be balanced

Features of the selected dataset Goal for these parameters: diversity Latitudes Longitudes IASI BT @ 870cm-1 Sat Zen Angle Goal for these parameters: minimization Intra-pixel IIS variance Inter-pixel IIS variance Difference of mean IIS Difference of cloud fraction Cloud fraction Continental fraction