The CarbonSat candidate mission - Radiometric and Spectral Performances over Spatially Heterogeneous Scenes

Transcription

1 The CarbonSat candidate mission - Radiometric and Spectral Performances over Spatially Heterogeneous Scenes J. Caron, B. Sierk, J.-L. Bézy, A. Loescher, Y. Meijer ESA-Estec (Netherlands) Earth Observation Directorate, Instrument Pre-Development Division CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 1

2 Overview of CarbonSat mission Status of CarbonSat candidate mission for ESA Earth Explorer 8, in competition with FLEX (Fluorescence Explorer) feasibility studies with 2 consortia (Airbus DS GmbH / OHB) Mission goals monitor atmospheric CO 2 and CH 4 high precision (noise) = 1-3 ppm for CO 2, 6-12 ppb for CH 4 very high accuracy (syst. errors) = 0.5 ppm for CO 2, 5 ppb for CH 4 high spatial resolution = 2km x 3km, large swath = km Technical implementation Push-broom instrument (common telescope, 3 spectrometers) measuring Earth high resolution : NIR OMI NO 2 over Europe SWIR-1 SWIR-2 Image by Pepijn Veefkind, KNMI CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 2

3 Outline 1. Error mechanisms 2. Derivation of requirements for Instrument Spectral Response Function (ISRF) distortions 3. Mitigation for along-track (ALT) ISRF distortion 4. Summary OMI NO 2 over Europe Image by Pepijn Veefkind, KNMI CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 3

4 Outline 1. Error mechanisms 2. Derivation of requirements for Instrument Spectral Response Function (ISRF) distortions 3. Mitigation for along-track (ALT) ISRF distortion 4. Summary OMI NO 2 over Europe Image by Pepijn Veefkind, KNMI CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 4

5 Heterogeneous Scene (HS) analysis Fact: strong measurement errors have been reported over real, non-uniform scenes in along-track (ALT) with OMI (App.Opt., pp , 2006) in along + across-track (ALT+ACT) with MAMAP (Atmos. Meas. Tech., 4, , 2011) Heterogeneous scene analysis is now a standard component of phase 0/phase A studies for EO missions at ESA Sentinel 4 Software correction for ISRF distortion due to ALT non-uniformities (1 error mechanism) Sentinel 5, CarbonSat (on-going) Many more error mechanisms analysed S5 is the first ESA EO mission with a systematic mitigation of scene heterogeneity effects CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 5

6 Error mechanisms over HS They can be organised in 3 families Family # Error name Description ISRF distortion Lack of correction before binning Misregistration errors 1 Non-uniform slit illumination (ALT) 2 Tilted slit image (ACT) The slit illumination is not uniform in ALT. The slit image is tilted wrt detector lines (spatial direction). 3 PRNU + ACT binning Pixel response non-uniformity is corrected only after ACT binning. 4 NL + ACT binning Non-linearity is corrected only after ACT binning. 5 NL + temporal (ALT) co-addition Non-linearity is corrected only after temporal (ALT) co-addition. 6 Mis-registration Barycenter of instrument footprint varies spectrally. 7 Variable footprint Shape of instrument footprint varies spectrally. 8 Polarisation scrambler 9 Detector memory effect (ALT) Scrambler generates a polarisationdependent pointing. Measurement N carries a fraction of N-1 measurement. Subject of this presentation Lack of correction before binning sub-pixels are binned on-board sub-pixels see different levels of signal gain or non-linearity correction done after binning Mis-registration errors Different part of a measured spectrum correspond to different areas on ground. CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 6

7 Error #1: non-uniform slit illumination (ALT) two situations with a spectrometer flying over a uniform/non-uniform scene same spectrometer measures different lineshapes depending on scene Case 1 = uniform Case 2 = non-uniform Slit illumination grating camera collimator slit telescope ISRF Flight direction Flight direction L>0 L>0 L=0 CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 7

8 Errors #2: Tilted slit image (ACT) Mechanism the slit image is locally tilted w.r.t. detector line (spatial direction) over uniform scenes, the ISRF is widened over non-uniform scenes, the ISRF is narrowed and shifted requirements on spectrometer smile / detector positioning 1 pixel CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 8

9 Outline 1. Error mechanisms 2. Derivation of requirements for Instrument Spectral Response Function (ISRF) distortions 3. Mitigation for along-track (ALT) ISRF distortion 4. Summary OMI NO 2 over Europe Family # Error name ISRF distortion Lack of correction before binning 1 Non-uniform slit illumination (ALT) 2 Tilted slit image (ACT) 3 PRNU + ACT binning 4 NL + ACT binning 5 NL + temporal (ALT) co-addition Image by Pepijn Veefkind, KNMI CarbonSat Copernicus SENTINEL-5 J.-L. Bézy ICSO SPIE 2014, Remote Tenerife, Sensing, 6-10 October Amsterdam, September 2014 Slide 9

10 Radiometric vs spectral approach ISRF distortion errors can be evaluated spectrally or radiometrically Radiometric approach (NIR) Spectral approach (NIR) undistorted k distorted ISRF = ISRF undistorted + k.isrf distorted Radiometric approach accurate = mapping radiometric errors at level-2 with a gain vector [talk CarbonSat B.Sierk] BUT unpractical = the gain vector depends on the observed location and instrument footprint! Requirement on ISRF distortion? convenient = elegant parameterization of distorted ISRF BUT approximate = available requirements on ISRF stability/knowledge assume perturbations with slow spectral variations (e.g. thermo-mechanical effects, ). For fast spectral variations, the definition of ISRF raises new difficulties CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 10

11 New terminology: ISRF vs ISMF Mono- stimulus (laser at fixed ) Widening + sampling by pixels The ISRF (Instrument Spectral Response Function) is the instrument response to a monochromatic stimulus at a fixed. The ISMF (M=Measured) describes the spectral origin of the photons measured in one spectral channel. Spectral origin of the measured photons for one spectral channel CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 11

12 ISRF/ISMF over heterogeneous scenes 2D spectral response function showing measured vs incident ISRF/ISMF are vertical/horizontal slices Slow spectral variations ISRF and ISMF are mirror images CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 12

13 ISRF/ISMF over heterogeneous scenes 2D spectral response function showing measured vs incident ISRF/ISMF are vertical/horizontal slices Slow spectral variations ISRF and ISMF are mirror images Heterogeneous scenes (effect magnified >100x) ISRF undergoes a spectrally fast variable shift ISMF is distorted (distorted) (just shifted) CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 13

14 ISRF/ISMF over heterogeneous scenes 2D spectral response function showing measured vs incident ISRF/ISMF are vertical/horizontal slices Slow spectral variations ISRF and ISMF are mirror images Heterogeneous scenes (effect magnified >100x) ISRF undergoes a spectrally fast variable shift ISMF is distorted (distorted) (just shifted) Difficulty = traditional requirements make no distinction between ISRF and ISMF CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 14

15 Spectral approach selected for ISRF errors CarbonSat ISRF knowledge error < than 1% (onground)/2% (inflight) of peak Assumption: this requirement can also be used over heterogeneous scenes In ACT, a requirement on slit image slope is derived based on a 0.1% allocation band Slit image slope NIR 10 % SWIR % SWIR-2 5 % Slit image local slope in % of spectral pixel per spatial sample Demanding value in NIR, SWIR-2 Relaxed in SWIR-1 thanks to weak absorption In ALT, no design requirement possible. Impact is part of ISRF knowledge error. How to mitigate it? CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 15

16 Outline 1. Error mechanisms 2. Derivation of requirements for Instrument Spectral Response Function (ISRF) distortions 3. Mitigation for along-track (ALT) ISRF distortion 4. Summary OMI NO 2 over Europe Family # Error name ISRF distortion Lack of correction before binning 1 Non-uniform slit illumination (ALT) 2 Tilted slit image (ACT) 3 PRNU + ACT binning 4 NL + ACT binning 5 NL + temporal (ALT) co-addition Image by Pepijn Veefkind, KNMI CarbonSat Copernicus SENTINEL-5 J.-L. Bézy ICSO SPIE 2014, Remote Tenerife, Sensing, 6-10 October Amsterdam, September 2014 Slide 16

17 Slit homogenizer Software correction is not possible in SWIR bands: due to strong detector read-out noise and impact on SNR, no temporal oversampling Slit homogenizer considered in S5 / CarbonSat Slit homogenizer (abbrev. SH) is a thick slit made with two parallel highly reflective mirrors. It is based on astigmatic imaging - in ALT telescope and collimator focal planes match the input/output SH planes - in ACT telescope and collimator focal planes coincide (normal imaging) Telescope Homogenizer Collimator Telescope Homogenizer Collimator Along track (homogenizing) Across track CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 17

18 Slit homogenizer modelling Interference pattern at output Monochromatic source point at SH input Due to multiple reflections, an interference pattern is created at SH output. Transfer function Intensity at output (e.g. with sampling N=1000 points) is calculated for each position at input (N=1000 points). The resulting N*N matrix is called a transfer function. It fully describes SH wavelength. Telescope pupil Mirror sources Parameters = 1.0 m / F# = 6.0 SH width = 100 m SH depth = 2.4 mm n=2 n=1 n=0 reflections input output CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 18

19 SH performance How to calculate SH performance? ISRF distortion due to a contrast jump from L dark to L bright at SSD center is estimated with L bright L dark ISRF, 0,, slit Lbright 0 Ldark 0 ISRFlinear 0 ISRFuniform 0 smear L L max ISRF, Slit/smear ratio bright 0 dark 0 uniform 0 Spectral contrast of the nonuniform scene SH intrinsic distortion using ISRFs for a uniform (ISRF uniform ) or linear (ISRF linear ) illumination profile varying from 0 to 1 at SH input SH intrinsic distortion Green = ISRF linear Blue = ISRF uniform Traditional slit intrinsic distortion ~ 65.7% Slit homogenizer intrinsic distortion ~ 3.78% Possible SH implementation Reduction factor ~ 17 in NIR CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 19

20 Outline 1. Error mechanisms 2. Derivation of requirements for Instrument Spectral Response Function (ISRF) distortions 3. Mitigation for along-track (ALT) ISRF distortion 4. Summary OMI NO 2 over Europe Image by Pepijn Veefkind, KNMI CarbonSat Copernicus SENTINEL-5 J.-L. Bézy ICSO SPIE 2014, Remote Tenerife, Sensing, 6-10 October Amsterdam, September 2014 Slide 20

21 Summary Many error mechanisms over non-uniform scenes, to be considered for EO spectrometers design Deriving requirements for ISRF distortion errors is challenging Scene variability makes the link to level-2 (using gain vectors) unpractical Spectral requirements can be used. Unfortunately this is approximate since ISRF relies on assumption of slow spectral variability On-going work Mitigation of ALT (along-track) ISRF distortion needs dedicated solutions Software correction (Sentinel-4) with temporally over-sampled data available on ground Hardware correction with slit homogenizer is considered for Sentinel-5 and CarbonSat CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 21

22 Thanks for your attention! Any questions? CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 22

23 BACK-UP CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 23

24 Scene parametrization The simplest parametrization is a superposition of 2 radiance spectra, corresponding to the same geophysical scenario but with different albedos (L dark and L bright ), with a spatially varying weight w(x,y) x y w x y w x y L,, 1, L dark, L For simple analyses: contrast scenes are used When realistic cases are needed: MODIS or AVIRIS data are converted into spatially variable weights giving equivalent reflectance values bright CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 24

25 Overview of Sentinel-4/5 missions Elements of the Copernicus Programme Sentinel-4: GEO based on-board MTG-S Sentinel-5: LEO based on-board MetOp-SG Scientific goal Sentinel-4 Hourly revisit time over Europe Mainly air quality Diurnal cycle of tropospheric composition Sentinel-5 Daily revisit time global coverage Climate, air quality, ozone & UV Tropospheric & stratospheric composition Technical implementation Push-broom imager spectrometer measuring Earth reflectance FOV 108 deg Swath ~ 2650 km CarbonSat ICSO 2014, Tenerife, 6-10 October 2014 Slide 25