Uncertainties in Sentinel-3 Sea and Land Surface Temperature Radiometer Thermal Infrared Calibration

Transcription

1 Uncertainties in Sentinel-3 Sea and Land Surface Temperature Radiometer Thermal Infrared Calibration Dave Smith; Arrow Lee; Mireya Extaluze; Edward Polehampton; Tim Nightingale; Elliot Newman; Dan Peters RAL Space STFC, United Kingdom 2017 RAL Space 1

2 The Remote Sensing Problem A very indirect measurement Noise Responsivity Spectral Response Resolution Coverage Stability Atmosphere (absorption, scattering, emission), surface state, geometry, illumination... Real world e.g. SST, cloud... Instrument Calibration Parameters Uncertainties are introduced at ALL levels and will affect the final physical quantity of interest Validation x v and S v Instrument measurements (y m ) and uncertainty (S y ) Accurate Physics and Environment Retrieval Forward model y(x) Retrieved parameters and uncertainty x and S x A priori information (x a ) And uncertainty (S a ) Knowledge of environment Cost function Understanding of what was missed 2

3 SLSTR instrument Nadir swath >74 (1400km swath) Dual view swath 49 (750 km) Two telescopes 110 mm / 800mm focal length Spectral bands TIR : 3.74µm, 10.85µm, 12µm SWIR : 1.38µm, 1.61µm, 2.25 µm VIS: 555nm, 659nm, 859nm Spatial Resolution 1km at nadir for TIR, 0.5km for VIS/SWIR Radiometric quality NEΔT 30 mk (LWIR) 50mK (MWIR) SNR 20 for VIS - SWIR Radiometric accuracy 0.2K for IR channels 2% for Solar channels relative to Sun On-Board Calibration Blackbody Sources for TIR VISCAL for solar channels 3

4 SLSTR-A Calibration at RAL - Jan-June 2015 Sentinel-3A launch - Feb 2016 First Image - March 2016 In-Orbit Commissioning Review July RAL Space Hurricane Ophelia 15/10/2017 4

5 SLSTR-B Arrived at RAL for calibration Oct 2016 In-Air Tests October Nov 2016 In-Vacuum Tests Nov 2016 Feb 2017 S3B Launch Spring 2018 SLSTR-B = Refurbished Proto-Flight Model (PFMr) Refurb includes: Rebuilt BB1 New flight BB2 Recoated telescope aperture stop to reduce internal strays 2017 RAL Space 5

6 The Goal To ensure the interoperability of satellite datasets it is a requirement for their measurements to be calibrated against standards that are traceable to SI units For temperature this is the International Temperature Scale of 1990 For IR instruments such as SLSTR the traceability is achieved via internal BB sources Instrument Blackbody Source S-PRT Fixed Point Cells 6

7 Calibration Flow Component Detectors Spectral Response Linearity Noise Polarisation Filters Spectral Response Transmission Dichroics Spectral Response Transmission Polarisation Thermometers Temperature vs. Resistance Black-Coating Spectral Emissivity Thermal Impedance Diffuser BRDF Mirrors Reflectance Roughness Subsystem IR-FPA Dynamic Range Linearity Noise Cross-Talk Alignment Polarisation Spectral Response VIS-FPA Dynamic Range Noise Linearity Alignment Polarisation Spectral Response Blackbodies Emissivity Radiance Temperature Gradients Temperature Stability VISCAL Throughput Polarisation Stray-Light Fore-Optics Throughput Alignment Image Quality Cross Talk Polarisation Stray-Light Requirements Instrument SLSTR Radiometric Response Noise Linearity Stray-Light Field-Of-View Polarisation Sensitivity Spectral Response Instrument Model Characterisation Calibration Validation Verification Processing End-to-End Model Retrieval Model Model Product The end-to-end model takes the SST noise and bias requirements and breaks these down into specifications for the individual components Retrieval Coefficients don't get calibrated as such, but nevertheless are maintained using QA procedures The instrument radiometric model takes the asmanufactured instrument and propagates the measured noises and biases to predictions at higher system levels Sea Surface Temperature (ITS-90) Land-Surface Temperature (ITS-90) 7

8 Calibration Model L scene Optics Detector Pre-Amp Offset Adj Gain Integrate V scene ADC DN scene L high L L low DN low DN DN high We obtain calibration coefficients via reference to known calibration sources 8

9 On-Board Calibration systems Thermal InfraRed Blackbodies VIS-SWIR Channels VISCAL Effective e >0.998 Zenith diffuser + T non-uniformity < 0.02 K relay mirrors T Abs. Accuracy 0.07 K Uncertainty <2% T stability < 0.3 mk/s 8 PRT sensors + 32 Thermistors 9

10 SLSTR L1 Processing Processing specification defined by ATBD -> DPM L0 and L1 Product Specifications Each spectral band (5 thermal bands) and detector element (2x2) for each for each earth view (separate for nadir and oblique) has unique set of calibration calibration coefficients = 40 for IR channels alone Contained in Satellite Characterisation and Calibration Database Document (S-CCDB) Configuration controlled by MPC 10

11 SLSTR IR Traceability Tree 2017 RAL Space 11

12 SLSTR Uncertainty Budget 2017 RAL Space 12

13 BB Uncertainties 2017 RAL Space 13

14 SLSTR IR Channel Calibration Budget This is the at-launch flight calibration budget 2017 RAL Space 14

15 Pre-Launch Calibration Objectives Provision of calibration data needed for data processing chain Does the end-to-end flight instrument calibration scheme work? New optical design 2 telescopes not 1, multiple detectors per channel OME thermal design not based on AATSR heritage Does the instrument calibration work over the full field of view and dynamic range? Wider instrument swath compared to AATSR Nonlinearity, Noise performance, Dynamic range Does calibration work in flight representative environment? Nominal BOL EOL (Hot) Orbital temperature variations 2017 RAL Space 15

16 Thermal IR Calibration Facility Earth Shine Plate Alignment Optics Initial Trials with STM completed April 2012 Point source + collimator TV and calibration of S3A instrument March-May 2015 Instrument Electronics Platform Simulator Blackbody Source S3B Calibration Oct 2016 Feb 2017 S3C 2019 ESA requirement to perform calibration tests under flight representative conditions RAL Space Thermal balance Steady State Instrument fully operational S3D

17 TIR calibration- Blackbody Source E S S C o o l a n t P i p e E l l i p t i c a l A p e r t u r e i n E a r t h s h i n e P l a t e E l l i p t i c a l a p e r t u r e i n t a r g e t b a f f l e ( m m m a j o r a x i s, m m m i n o r a x i s ) E S S T a r g e t B a f f l e Standards Precision RIRTs Calibrated to ITS90 < 0.01K M u l t i - L a y e r I n s u l a t i o n B a f f l e R I R T s ( 2 P o s i t i o n s ) Radiometric Accuracy R o t a t i n g F l a n g e C o o l e d S h i e l d G l a s s F i b r e S u p p o r t s ( 3 p o s i t i o n s ) C h a n n e l s f o r R e f r i g e r a n t B a s e p l a t e R I R T s ( 4 p o s i t i o n s ) T a r g e t M o u n t i n g F l a n g e C o o l e d C o p p e r B a f f l e A l u m i n i u m S u p p o r t C y l i n d e r S t a i n l e s s S t e e l S p a c e r s S t r u c t u r e d A l u m i n i u m P l a t e ( C i r c u l a r G r o o v e s, 1 5 h a l f a n g l e ) Emissivity 12µm = µm = µm = < 0.05K Sources previously used for all ATSR instruments ATSR ATSR-2 AATSR S3 SLSTR 17

18 IR Calibration Test Summary Calibration at Nominal BOL conditions Centre of Nadir/Oblique views On-Board BBs at nominal settings (250K, 300K) Test over full dynamic range (5K intervals) Test over full swath (reduced number of scene temperatures) Calibration at Hot EOL conditions Centre of Nadir/Oblique views On-Board BBs at nominal settings (250K, 300K) Test over part dynamic range (10K intervals) Tests with different on-board BB temperatures Test performed at Nominal BOL conditions Currently at low, medium, high power settings +Y and Y BBs will be switched Test over part dynamic range (10K intervals) Orbital simulation tests 2017 RAL Space 18

19 IR Calibration - Counts Vs. Temps 70us integration time shown only Min temperature achieved is 224K Saturation of S7 > 300K (additional step at 305K to confirm) 19

20 TIR Calibration - Measured vs Actual BT Nadir Oblique 20

21 IR Calibration Initial Results Non-Linearity of S8 and S9 consistent with expected behaviour of PC MCT detectors. S3A and S3B show very similar behaviour RAL Space 21

22 Creation of NL Table Measured Counts and BB Radiances normalised to signal corresponding to counts y = L actual DN L(0) L DN ref L(0) x = DN meas DN ref - from SLSTR - from thermometers Polynomial function fitted to data to generate coefficients for NL function NL = n i=2 a i a 1 x i 1 Digital counts are linearized using DN = DN/(1.0 + NL x ) 2017 RAL Space 22

23 Measured - Actual BT SLSTR-B Nadir Oblique 2015 RAL Space 23

24 Measured - Actual BT SLSTR-A Nadir Oblique 2015 RAL Space 24

25 Why the differences? Non-Blackness of optical stops (i.e. ɛ < 0.9) causing non-uniform thermal background Measurements by PTB confirm 2015 investigation Hence modification to stop coatings 2017 RAL Space Temperature gradients in flight BBs Thermal modelling shows asymmetry of baseplate temperatures Analysis of BB radiances in progress 25

26 Black-Body Cross-Over Test Nadir Scanner Post launch we can check BB signals by comparing the signals when the BBs are at the same temperatures. This is achieved by switching the heated BB and allowing their temperatures to cross-over. Test is performed during ground calibration as a baseline 2017 RAL Space 26

27 Comparison DN vs BB Temps 1 st Cross Over Part (RAD06) 1 2 nd Cross Over Part (RAD08) RAL Space 27

28 S3B BB Counts at Cross-Over BB X-Over 1 - Temp = K +YBB -YBB ΔDN ΔT S7 Nadir S8 Nadir S9 Nadir S7 Oblique S8 Oblique S9 Oblique BB X-Over 2 Temp = K +YBB -YBB ΔDN ΔT S7 Nadir S8 Nadir S9 Nadir S7 Oblique S8 Oblique S9 Oblique RAL Space 28

29 SLSTR-A Pre-Launch Part 1 Part RAL Space 29

30 SLSTR-A Post Launch Part 1 Part RAL Space 30

31 S3A BB Counts Comparison at X-Over Post Launch 29 Mar-2016 Pre Launch 29 Mar RAL Space 31

32 Correction for Offset Error (1) From the measured DN we wish to obtain the scene radiance L scene Assuming that the radiometric response of the system is linear with radiance (or adjusted for detector non-linearity), we can derive the gain using two calibration sources of known scene radiance i.e. Blackbodies DN BB = g(l BB ) + DN Offset This gives L scene = XL hbb + (1-X)L cbb + 0 Where X = (DN-DN cbb )/(DN hbb -DN cbb ) Both g and DN offset MUST be constant during the calibration interval. 32

33 Correction for Offset Error (2) What if DN offset is not constant during the scan cycle? Lets consider as a function of pixel position, a scan dependent radiance pertubation of ±ΔL(pos) which in turn gives rise to a pertubation in the background signal ±ΔDN(pos). The calibration model now becomes L scene + ΔL(scene)= X(L hbb + ΔL(hbb)) + (1-X)(L cbb + ΔL(cbb)) where X = ((DN scene +ΔDN(scene))-(DN cbb +ΔDN(cbb)))/ (DN hbb +(ΔDN(hbb) - (DN cbb + ΔDN(cbb))) But we are assuming the ideal calibration model, so the error we observe in the calibration is ΔL error = ΔL(scene) XΔL(hbb) - (1-X) ΔL(cbb) 33

34 Correction for Offset Error (3) Model provides good estimate measured calibration errors. Hence conclusion that this is best explanation for discrepancy. Input parameters derived from instrument temperatures available in HK. These provide an approximation of the stray light source. Model has been coded and tested in Prototype Instrument Processing Facility (IPF-P) Nadir Oblique Early intercomparsions with IASI performed by EUMETSAT suggest that on-orbit stray-light error correction is not necessary RAL Space 34

35 Current State of SLSTR Traceability Pre-Launch reports contains most information Write up as reviewed papers in progress Includes uncertainty estimates of measurements L1 Products Detector noise expressed as NEDT (TIR channels) and NEDL (VIS/SWIR channels) for each scan line Uncertainties in the radiometric calibration are included in the quality annotation datasets as a table of uncertainty vs. temperature type-b (a-priori) estimates based on the pre-launch calibration and calibration model (see later slides) Per pixel estimation of the radiometric uncertainty for either random effects or systematic effects has not been implemented Significant impact on product size and processing time User requirements poorly defined! 35

36 SLSTR Traceability Document This is a key document that needs to be maintained and updated when new information on uncertainty estimates come to light. Much information already exists in different formats which needs collating Content should be published in reviewed journal Identifies key sources of uncertainty Collates known uncertainty estimates e.g. BB temperatures + Emissivity Records the current Traceability Chain I.e. Documents that can link uncertainty estimates to standards e.g. BB emissivity Identify gaps e.g. degradation model for BB electronics 36

37 Conclusions Pre-Launch Calibration allows us to validate the end-to-end instrument flight calibration systems against known reference targets. Not possible after launch Provides a reference dataset against which the processing algorithms can be verified. Papers on calibration results are being prepared L1 products contain basic uncertainty estimates Noise derived from BB sources Estimates of calibration uncertainties from pre-launch characterisation Improvements are foreseen Traceability chain needs to be documented in-order for SLSTR to become a reference sensor RAL Space 37

38 References Calibration Plan David L. Smith, Tim J. Nightingale, Hugh Mortimer, Kevin Middleton, Ruben Edeson, Caroline V. Cox, Chris T. Mutlow, Brian J. Maddison, Peter Coppo Calibration approach and plan for the Sea and Land Surface Temperature Radiometer J. Appl. Remote Sens. 8(1), (Jun 30, 2014). [doi: /1.jrs ] Description of SLSTR design Coppo, B. Ricciarelli, F. Brandani, J. Delderfield, M. Ferlet, C. Mutlow, G. Munro, T. Nightingale, D. Smith, S. Bianchi, P. Nicol, S. Kirschstein, T. Hennig, W. Engel, J. Frerick, J. Nieke, SLSTR: A High Accuracy Dual Scan Temperature Radiometer For Sea And Land Surface Monitoring From Space, Journal of Modern Optics, 57(18), (2010) [doi: / ]. 38