S3-A OLCI Cyclic Performance Report...S01 sequence on 21/10/2016 10:03 to 10:05 (absolute orbit...

PREPARATION AND OPERATIONS OF THE MISSION PERFORMANCE

CENTRE (MPC) FOR THE COPERNICUS SENTINEL-3 MISSION

S3-A OLCI Cyclic Performance Report

Cycle No. 010

Start date: 14/10/2016

End date: 10/11/2016

Ref.: S3MPC.ACR.PR.01-010

Issue: 1.0

Date: 22/11/2016

Contract: 4000111836/14/I-LG

Customer: ESA Document Ref.: S3MPC.ACR.PR.01-010

Contract No.: 4000111836/14/I-LG Date: 22/11/2016

Issue: 1.0

Project: PREPARATION AND OPERATIONS OF THE MISSION PERFORMANCE CENTRE (MPC)

FOR THE COPERNICUS SENTINEL-3 MISSION

Title: S3-A OLCI Cyclic Performance Report

Author(s): OLCI ESLs

Approved by: L. Bourg, OLCI ESL

Coordinator

Authorized by Frédéric Rouffi, OPT Technical

Performance Manager

Distribution: ESA, EUMETSAT, S3MPC consortium

Accepted by ESA S. Dransfeld, MPC Deputy TO

for OPT

P. Féménias, MPC TO

Filename S3MPC.ACR.PR.01-010 - i1r0 - OLCI Cyclic Report 010.docx

Disclaimer

The work performed in the frame of this contract is carried out with funding by the European Union. The views expressed herein can in no way be taken to reflect the official opinion of either the European Union or the

European Space Agency.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: iii

Changes Log

Version Date Changes

1.0 22/11/2016 First Version

List of Changes

Version Section Answers to RID Changes

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: iv

Table of content

1 INSTRUMENT MONITORING ......................................................................................................................... 1

1.1 CCD TEMPERATURES ......................................................................................................................................... 1

1.2 RADIOMETRIC CALIBRATION ............................................................................................................................... 2

1.2.1 Dark Offsets [OLCI-L1B-CV-230] .............................................................................................................. 2

1.2.2 Instrument response and degradation modelling [OLCI-L1B-CV-250]..................................................... 5

1.2.3 Ageing of nominal diffuser [OLCI-L1B-CV-240] ...................................................................................... 10

1.2.4 Updating of calibration ADF [OLCI-L1B-CV-260] ................................................................................... 10

1.3 SPECTRAL CALIBRATION [OLCI-L1B-CV-400] ..................................................................................................... 10

1.4 SIGNAL TO NOISE ASSESSMENT [OLCI-L1B-CV-620] ........................................................................................... 11

1.4.1 SNR from Radiometric calibration data................................................................................................. 11

1.4.2 SNR from EO data. ................................................................................................................................. 12

1.5 GEOMETRIC CALIBRATION/VALIDATION .............................................................................................................. 12

2 OLCI LEVEL 1 PRODUCT VALIDATION ...........................................................................................................14

3 LEVEL 2 LAND PRODUCTS VALIDATION ........................................................................................................26

4 LEVEL 2 WATER PRODUCTS VALIDATION .....................................................................................................32

5 LEVEL 2 SYN PRODUCTS VALIDATION ...........................................................................................................38

6 EVENTS ........................................................................................................................................................39

7 APPENDIX A .................................................................................................................................................40

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: v

List of Figures

Figure 1: long term monitoring of CCD temperatures using minimum value (top), time averaged values

(middle), and maximum value (bottom) provided in the annotations of the Radiometric Calibration Level

1 products, for the Shutter frames, all radiometric calibrations so far. --------------------------------------------- 1

Figure 2: Same as Figure 1 for diffuser frames. --------------------------------------------------------------------------- 2

Figure 3: Dark Offset for band Oa1 (top) and Oa21 (bottom), all radiometric calibrations so far except

the first one (orbit 183) for which the instrument was not thermally stable yet. -------------------------------- 3

Figure 4: map of periodic noise for the 5 cameras, for band Oa21. X-axis is detector number (East part,

from 540 to 740, where the periodic noise occurs), Y-axis is the orbit number. The counts have been

corrected from the west detectors mean value (not affected by periodic noise). Periodic noise amplitude

is high in camera 2, 3 and 4. It is lower in camera 4 and small in camera 1. We see that the drift of the

periodic noise tends to stabilize during the last radiometric calibrations. ----------------------------------------- 4

Figure 5: Dark Current for band Oa1 (top) and Oa21 (bottom), all radiometric calibrations so far except

the first one (orbit 183) for which the instrument was not thermally stable yet. -------------------------------- 5

Figure 6: Gain Coefficients for band Oa1 (top) and Oa21 (bottom), all diffuser 1 radiometric calibrations

so far except the first one (orbit 183) for which the instrument was not thermally stable yet. -------------- 6

Figure 7: time evolution of the camera-averaged gain coefficients for bands Oa1, Oa7, Oa14 and Oa21

(from left to right and top to bottom). -------------------------------------------------------------------------------------- 6

Figure 8: camera averaged gain relative evolution with respect to most recent calibration, as a function

of elapsed time since first calibration acquired after the fix of the Start Trackers issue; one curve for

each band (see colour code on plots), one plot for each module. --------------------------------------------------- 7

Figure 9: Across-track profiles of Gains relative evolution with respect to most recent calibration for time

distance about 2, 6, 13, 19, 27 and 34 weeks (from left to right then top to bottom) -------------------------- 8

Figure 10: same as Figure 8 after normalization by band Oa18. ------------------------------------------------------ 9

Figure 11: Signal to Noise ratio as a function of the spectral band for the 5 cameras. These results have

been computed from radiometric calibration data. All calibrations except first one (orbit 183) are

presents with the colours corresponding to the orbit number (see legend). The SNR is very stable with

time: the curves for all orbits are almost superimposed. --------------------------------------------------------------11

Figure 12: long-term stability of the SNR estimates from Calibration data, example of channel Oa1. -----12

Figure 13: Geometric Validation in-house tool functional breakdown. --------------------------------------------13

Figure 14: location and dates for the five Geometric Validation scenes. ------------------------------------------13

Figure 15: summary results for each product, in meters. -------------------------------------------------------------13

Figure 16: Location of the Cal/Val-test sites used to perform this activity ----------------------------------------17

Figure 17: RGB Quick-Looks from S3A/OLCI over Pacific-NE-Optimum site. Red box indicates the ROI. --17

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: vi

Figure 18: Rayleigh calibration coefficients from S3A/OLCI averaged over two acquisitions over Pacific-

NW-Optimum and SIO_Optimum calval-sites as a function of wavelength. Error bars indicate the

associated uncertainty of the calibration coefficients ------------------------------------------------------------------19

Figure 19: Interband calibration coefficients from S3A/OLCI from one acquisitions over SIO_Optimum

calval-site as a function of wavelength. Error bars indicate the associated uncertainty of the calibration

coefficients ------------------------------------------------------------------------------------------------------------------------20

Figure 20: Quick-looks of ERR-L1B S3A/OLCI products generated by DIMITRI from top-left to bottom-

right over Algeria3, Algeria5, Libya1, Libya4, Mauritania1 and Mauritania2. The red squares indicate the

position of the ROIs. ------------------------------------------------------------------------------------------------------------21

Figure 21: The estimated gain coefficients (observed/simulated signal) from S3A/OLCI over the 6 PICS

sites identified by CEOS (black diamond) as a function of wavelength. Dashed-green, orange and red

lines indicate the 2%, 5% and 10% respectively. Dashed-black lines indicate the significant gaseous

absorption bands which are mis-represented in the BRDF model of PICS method in DIMITRI. Error bars

indicate uncertainty associated to the estimated gain coefficients. ------------------------------------------------22

Figure 22: Time series of the relative difference defined as observed reflectance - simulated one over

the later in per-cent from (black) MSI_A and (blue) OLCI_A for bands 665 nm over (top) Algeria-3,

(middle) Algeria-5 and (bottom) Libya-1 CalVal-sites. Error bars indicate the method uncertainty. -------23

Figure 23: The estimated gain coefficients (observed/simulated signal) from (black) S2A/MSI and (blue)

S3A/OLCI over (top) Algeria-3 and (bottom) Libya-1 PICS as a function of wavelength. Dashed-green,

orange and red lines indicate the 2%, 5% and 10% respectively. Dashed-black and blue lines indicate the

significant gaseous absorption bands which are mis-represented in the BRDF model of PICS method in

DIMITRI. Error bars indicate uncertainty associated to the estimated gain coefficients. ----------------------24

Figure 24: Scatter-plot between match-up OLCI/FAPAR daily values and FAPAR derived from MODIS at

250m. -------------------------------------------------------------------------------------------------------------------------------29

Figure 25: FAPAR time series from OLCI (blue) and MODIS (green) over the US-Ne1 CV site. ---------------29

Figure 26 FAPAR time series from OLCI (blue) and MODIS (green) over the IT-Cat CV site -------------------30

Figure 27: FAPAR time series from OLCI (blue) and MODIS (green) over the SP-Val CV site. Inner panels

time series are MERIS FAPAR for Summer and Fall 2001 (top panel) and 2009 (bottom panel). ------------30

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 1

1 Instrument monitoring

1.1 CCD temperatures

The monitoring of the CCD temperatures is based on MPMF data extractions not yet operational. In the

meantime, we monitor the CCD temperatures on the long-term using Radiometric Calibration

Annotations (see Figure 1). Variations are very small (0.08 C peak-to-peak) and no trend can be

identified. Data from current cycle (2 rightmost data points) do not show any specificity.

Figure 1: long term monitoring of CCD temperatures using minimum value (top), time averaged values (middle),

and maximum value (bottom) provided in the annotations of the Radiometric Calibration Level 1 products, for

the Shutter frames, all radiometric calibrations so far.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 2

Figure 2: Same as Figure 1 for diffuser frames.

1.2 Radiometric Calibration

Two OLCI Radiometric Calibration Sequences have been acquired during Cycle 010:

S01 sequence on 21/10/2016 10:03 to 10:05 (absolute orbit 3531)


This section presents the overall monitoring of the parameters derived from radiometric calibration data

and highlights, if present, specificity of current cycle data.

1.2.1 Dark Offsets [OLCI-L1B-CV-230]

Dark offsets.

Dark offsets are continuously affected by the global offset induced by the Periodic Noise on the OCL

convergence. Current Cycle calibrations (orbits 3531 & 3736) are affected the same way as others. The

amplitude of the shift varies with band and camera from virtually nothing (e.g. camera 2, band 0a1) to

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 3

up to 5 counts (Oa21, camera 3). The Periodic Noise itself comes on top of the global shift with its

known signature: high frequency oscillations with a rapid damp. This effect remains more or less stable

with time in terms of amplitude, frequency and decay length, but its phase varies with time, introducing

the global offset mentioned above.

There is no significant evolution of this parameter during the current cycle.

Figure 3: Dark Offset for band Oa1 (top) and Oa21 (bottom), all radiometric calibrations so far except the first

one (orbit 183) for which the instrument was not thermally stable yet.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 4

Figure 4: map of periodic noise for the 5 cameras, for band Oa21. X-axis is detector number (East part, from 540

to 740, where the periodic noise occurs), Y-axis is the orbit number. The counts have been corrected from the

west detectors mean value (not affected by periodic noise). Periodic noise amplitude is high in camera 2, 3 and 4.

It is lower in camera 4 and small in camera 1. We see that the drift of the periodic noise tends to stabilize during

the last radiometric calibrations.

Dark Currents.

Dark Currents are not affected by the global offset of the Dark Offsets, thanks to the clamping to the

average blind pixels value. However, the oscillations of Periodic Noise remain visible. There is no

significant evolution of this parameter during the current cycle.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 5

Figure 5: Dark Current for band Oa1 (top) and Oa21 (bottom), all radiometric calibrations so far except the first

one (orbit 183) for which the instrument was not thermally stable yet.

1.2.2 Instrument response and degradation modelling [OLCI-L1B-CV-250]

Figure 6 below shows the gain coefficients of every pixel for two OLCI channels, Oa1 (400 nm) and Oa21

(1020 nm), highlighting the significant evolution of the instrument response since early mission.

Figure 7 on the other hand displays the time evolution of the cross-track averaged gain, for each

module, as a function of time. It shows that if a significant evolution occurred during the early mission,

the trends tend to stabilize. In particular the last two calibrations provide very stable results. This is

further illustrated on Figure 8 and Figure 9. The latter shows that radiometric gains are becoming very

stable: within 0.2% since 4 weeks without spectral signature, and within 0.5% since 9 weeks, still

without spectral signature; beyond that a significant evolution can be seen globally as a “white”

curvature of the AC profile, as well as spectrally with channel Oa1 raising out of the general trend.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 6

Figure 6: Gain Coefficients for band Oa1 (top) and Oa21 (bottom), all diffuser 1 radiometric calibrations so far

except the first one (orbit 183) for which the instrument was not thermally stable yet.

Figure 7: time evolution of the camera-averaged gain coefficients for bands Oa1, Oa7, Oa14 and Oa21 (from left

to right and top to bottom).

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 7

Figure 8: camera averaged gain relative evolution with respect to most recent calibration, as a function of

elapsed time since first calibration acquired after the fix of the Start Trackers issue; one curve for each band (see

colour code on plots), one plot for each module.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 8

Figure 9: Across-track profiles of Gains relative evolution with respect to most recent calibration for time

distance about 2, 6, 13, 19, 27 and 34 weeks (from left to right then top to bottom)

In order to get rid of the white variability (not spectrally dependant) caused by the BRDF model, all

bands are normalized by band Oa18. Oa18 was chosen because NIR degrades slowest and because Oa20

and Oa21 are subject to Periodic Noise, e- leaks, etc … Results are presented Figure 10.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 9

Figure 10: same as Figure 8 after normalization by band Oa18.

In Figure 10, we see that the ugly oscillations of Figure 8 have disappeared. However it is still surprising

that some bands show an increase of sensitivity with time, while a decrease is expected since we are

monitoring a ‘degradation’ of the instrument. Using the diffuser 2 results, we can say that this sensitivity

increase cannot be explained by the ageing of diffuser 1. Moreover, we have checked that the spectral

assignment drift cannot explain either this increase of sensitivity.

Thus there is still something that remains unexplained concerning the evolution of the sensitivity of the

instrument. Investigations are on-going.

The time elapsed until the beginning of the mission is still too small to be able derive a degradation

model.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 10

1.2.3 Ageing of nominal diffuser [OLCI-L1B-CV-240]

There has been no new calibration sequence S05 (reference diffuser) acquired during cycle 010.

1.2.4 Updating of calibration ADF [OLCI-L1B-CV-260]

There has been no OL_1_CAL_AX update during cycle 010.

1.3 Spectral Calibration [OLCI-L1B-CV-400]

There has been no Spectral Calibration acquisition during cycle 010.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 11

1.4 Signal to Noise assessment [OLCI-L1B-CV-620]

1.4.1 SNR from Radiometric calibration data.

Figure 11: Signal to Noise ratio as a function of the spectral band for the 5 cameras. These results have been

computed from radiometric calibration data. All calibrations except first one (orbit 183) are presents with the

colours corresponding to the orbit number (see legend). The SNR is very stable with time: the curves for all orbits

are almost superimposed.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 12

Figure 12: long-term stability of the SNR estimates from Calibration data, example of channel Oa1.

1.4.2 SNR from EO data.

There has been no update on SNR assessment from EO data during the cycle. Last figures (cycle 9) are

considered valid.

1.5 Geometric Calibration/Validation

In-house tools for Geometric Validation have been set-up and used during Cycle 10 in order to

compensate for the unavailability of the GeoCal tool at MPC-CC. These tools provide estimates of the

co-registration between Sentinel-2/MSI L1c and Sentinel-3/OLCI L1b products. The methodology is

based on automatic detection of GCPs by blob detection using SURF (Speeded Up Robust Features, blob

detection based on the determinant of the Hessian instead of a Laplacian), and inter-correlation of sub-

images to determine the spatial translation. In this first run, 3 bands have been used, as sketched on

Figure 13.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 13

Figure 13: Geometric Validation in-house tool functional breakdown.

Five products have been selected, at various locations on Earth, as shown on Figure 14.

Figure 14: location and dates for the five Geometric Validation scenes.

The results show a large dispersion of the retrieved shifts, about 150m, with an average of 130m, after

exclusion of the SW Australia scene due to very low correlation. Note that the average translation for all

scenes (as determined from average x and y displacements, not as the average distance) is about 40m,

but its quality cannot be verified on such a limited data set.

Figure 15: summary results for each product, in meters.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 14

2 OLCI Level 1 Product validation

[OLCI-L1B-CV-300] – Radiometric validation

In the frame of the ESA SEOM project called S2 Radiometric Validation, ACRI-ST is developing a

radiometric inter-band validation method based on the observation of Deep Convective Clouds. The

critical path of the method is now defined and ACRI wishes the S3 MPC project to benefit from this

development adapting the method to OLCI. The main adaptations to be done (LUTs computation, etc.)

have been identified and the activities will start in the next days. The DCC data selection is already done

by the S3ETRAC service.

This method is complementary to the other methods already implemented in the frame of the S3 MPC

(Rayleigh, Sunglint, DIMITRI methods, etc.).

[OLCI-L1B-CV-310] – Radiometric validation

S3ETRAC OLCI processing

The S3ETRAC service used during the commissioning phase with OLCI and SLSTR products has been

adapted and deployed in the S3 MPC environment. Several adaptations and improvements have been

performed: (1) the processing has been updated to take into account the difference of Manifest file

format from IPFP and IPF/PUG; (2) update of the front door of the L1 input products; (3) setup of the ftp

to deliver the S3ETRAC products; (4) database setup and management; (5) service management: sending

of an automatic daily email to the S3ETRAC manager summarising the S3 products received, the errors

of processing (if any) and the S3 products with no site identified; (6) optimisation of the runs

management on the cluster. Moreover, the configuration of the processing has been updated to take

into account more sites than for the commissioning phase (now, 49). Some minor bugs have also been

fixed.

The S3ETRAC service is open since 12th October. It extracts and filters OLCI L1 RR and SLSTR L1 RBT data,

filters and computes associated statistics over 49 sites corresponding to different surface types (desert,

snow, ocean maximizing Rayleigh signal, ocean maximizing sunglint scattering and deep convective

clouds). The S3ETRAC products available at ftp://ftp.acri-cwa.fr are used for the assessment and

monitoring of the L1 radiometry by the L1 Rad Val ESLs.

763 S3ETRAC products have been generated (31/10/2016) including

509 S3ETRAC/OLCI products (354 for desert, 44 for snow, 79 for Rayleigh, 13 for sunglint and 19

for DCC)

254 S3ETRAC/SLSTR products (133 for desert, 2 for snow, 115 for Rayleigh and 4 for Rayleigh)

The development of automatic scripts calculating monthly statistics on the number of products

processed and on the records extracted per site and per surface type is in progress. Examples of first

statistics graphs are presented below: (top, left) number of OLCI L1 products received and processed per

month, (top, right) number of OLCI Rayleigh records extracted per month, (bottom, left) number of

products intersecting the SADE-Libya4 desert site and processed in yellow, number of products giving

ftp://ftp.acri-cwa.fr/

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 15

one record in blue, number of products not giving record in green and number of products with an error

of processing.

All these statistics shall allow to quickly identify processing anomalies, issue on site coverage, etc. It is

planned to generate all these statistics on a monthly basis and to display them on a specific page of the

S3ETRAC website.

A S3ETRAC service website (http://s3etrac.acri.fr) has also been open with only for the moment

informative webpages.

The home page with general information on the S3ETRAC service

The service-version page with information on the S3ETRAC processing and ADF versions and

content, as well as sites description

The processing-info page with processing information such as list of products reprocessed,

anomalies, etc.

Radiometric validation with DIMITRI (ARGANS)

This activity is dedicated to the assessment of the absolute and relative radiometry measurements of

OLCI on board of Sentinel-3A over the wavelength spectrum (Oa1-Oa21). Due to the lack of in-situ

measurements to perform this radiometric vicarious validation, we have restricted our

verification/validation (for now) only over the VIS spectral range (Oa01 to Oa10) using Rayleigh and sun-

glint methodologies supported by DIMITRI.

http://s3etrac.acri.fr/

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 16

Regarding the assessment of the relative radiometry measurements of OLCI, we use the desert-

calibration method (so-called PICS) for Pseudo-Invariant Calibration Sites, also supported by DIMITRI

toolbox.

1- Absolute Radiometry Vicarious Validation:

We have acquired about 445 acquisitions from Baseline N001 over March until July 2016 over the six

ocean-CalVal-sites (see Table 1 and Figure 16). After the ingestion of the whole dataset successfully into

DIMITRI and the automatic cloud screening, quick-looks were generated (e.g. Figure 17). The Rayleigh

method was performed over all the sites and all the time series. We found only 13 acquisitions which

are compliant with Rayleigh method criteria using (Option:Pixel-per-Pixel).

Then we have acquired 102 acquisitions between September and November 2016 from baseline N002.

We found only 2 acquisitions which are compliant with Rayleigh method criteria using (Option: ROI-

average).

Table 1: Ocean test sites coordinates

SITE_ID TYPE NORTH_LAT SOUTH_LAT EAST_LON WEST_LON

Atlantic North-

West (ANW)

Ocean 27.04 17.05 -57.25 -60.09

Atlantic North-

West Optimum

(ANWO)

Ocean 27.04 22.5 -66.5 -67.5

Atlantic South-

West Optimum

(ANWO)

Ocean -13.5 -14.5 -23.5 -24.5

Pacific North-East

(PNE)

Ocean 23.49 15.01 -162.69 -176.36

Pacific North-East

Optimum (PNEO)

Ocean 18.5 17.5 -151.5 -152.5

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 17

Figure 16: Location of the Cal/Val-test sites used to perform this activity

Figure 17: RGB Quick-Looks from S3A/OLCI over Pacific-NE-Optimum site. Red box indicates the ROI.

In order to perform Rayleigh method, the following conditions shall be satisfied:

To avoid any signal contamination by clouds, haze or cloud shadows, clear sky conditions (0%

cloud coverage at AOI level) are mandatory for proper computation of the vicarious coefficients.

A low wind speed is required for ensuring no presence of whitecaps; typically it is limited to

5 m/s.

A small content of aerosol must be ensured to avoid any error propagation in the atmospheric

path radiance. DIMITRI methodology considers a very stringent threshold on Rayleigh-Corrected

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 18

Normalized Radiance at 865 nm of 0.002, which ensure the former condition and also avoid

using further data screening for sun glint.

DIMITRI allows the user to choose/modify these parameters. Rayleigh calibration coefficients and their

associated uncertainties over the targeted ROI in DIMITRI are summarized in Table 2 and presented on

Figure 18.

Table 2: Rayleigh calibration coefficients and their associated standard deviation from two acquisitions over the

NW_Pacific_Optimum and SIO_Optimum calval-sites from the baseline N002.

S3A-

OLCI

Wave

length

(nm)

Rayleigh Cal-

Coefficient

Standard

deviation

Oa01 400 1.058 0.058

Oa02 412 1.057 0.060

Oa03 443 1.040 0.041

Oa04 490 1.039 0.042

Oa05 510 1.022 0.042

Oa06 560 1.003 0.040

Oa07 620 0.999 0.040

Oa08 665 1.001 0.031

Oa09 674 1.006 0.037

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 19

Figure 18: Rayleigh calibration coefficients from S3A/OLCI averaged over two acquisitions over Pacific-NW-

Optimum and SIO_Optimum calval-sites as a function of wavelength. Error bars indicate the associated

uncertainty of the calibration coefficients

These results are obtained from the baseline N-002 and from only two acquisitions over two different

sites. However, the results are not suficiantly mature to draw any conclusions. We believe that more

acquisition over the six sites do improve the results and their uncertainties. The more the number of

processed products the more statistically significant are the results.

2- Interband Vicarious Calibration over Sung lint

The sun glint method is performed over the six ocean CalVal sites (Table 1), 174 acquisitions from

baseline N001 and 102 acquisitions from baseline N002 have been processed. Only one acquistion over

SIO_Optimum is compliant with the method criteria. The rations interbands and their associated

uncertainties over the targeted ROI in DIMITRI are presented on Figure 19.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 20

Figure 19: Interband calibration coefficients from S3A/OLCI from one acquisitions over SIO_Optimum calval-site

as a function of wavelength. Error bars indicate the associated uncertainty of the calibration coefficients

3- Relative Radiometry Vicarious Validation

We have acquired about 400 acquisitions from Baseline N001 from March until July 2016 and 68

acquisitions from baseline N002 from August until November 2016 over the six desert-CalVal-sites

(Figure 20 and Table 3). All the dataset is ingested successfully into DIMITRI and automatically cloud

screened, quick-looks were generated (e.g. Figure 20). The results below are obtained from 26

acquisitions from Baseline N002 which are compliant with PICS method conditions.

Table 3: PICS test-sites coordinates

Latitude (°) Longitude (°)

Name min max min max

Algeria 3 29.82 30.82 7.16 8.16

Algeria 5 30.52 31.52 1.73 2.73

Libya 1 23.92 24.92 12.85 13.85

Libya 4 28.05 29.05 22.89 23.89

Mauritania 1 -9.8 -8.8 18.8 19.9

Mauritania 2 -9.28 -8.28 20.35 21.35

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 21

Figure 20: Quick-looks of ERR-L1B S3A/OLCI products generated by DIMITRI from top-left to bottom-right over

Algeria3, Algeria5, Libya1, Libya4, Mauritania1 and Mauritania2. The red squares indicate the position of the

ROIs.

The PICS method supported by DIMITRI is applied successfully. We observe a clear consistency

overall the 6 PICS. The estimated gain coefficients overall the PICS -defined as observed reflectance to

the simulated one- show values within <3% and uncertainty about ±2.5% for almost all the spectral

range of OLCI from Oa01-Oa19 (Figure 21). The spectral bands with significant absorption of water

vapour and O2 (Oa11, Oa13) show outlier gain and higher uncertainty, while bands Oa01, Oa05, Oa10,

Oa12 and Oa14 display gain within 3%-5%. As the PICS-method in DIMITRI is a statistical-based method,

we think that such results over short time lapse are satisfactory for now. However, the results from both

methods (Rayleigh and PICS) are consistent and show good agreement over the VIS-spectral range. Surly

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 22

these results need to be consolidated with more products over longer time series to allow us to draw

reliable conclusions.

Figure 21: The estimated gain coefficients (observed/simulated signal) from S3A/OLCI over the 6 PICS sites

identified by CEOS (black diamond) as a function of wavelength. Dashed-green, orange and red lines indicate the

2%, 5% and 10% respectively. Dashed-black lines indicate the significant gaseous absorption bands which are

mis-represented in the BRDF model of PICS method in DIMITRI. Error bars indicate uncertainty associated to the

estimated gain coefficients.

4- Cross-mission Intercomparison

ARGANS proposed to use “Sensor-to-Sensor inter-calibration” method over DomeC CalVal-site. The

method has been tested on S2A/MSI. It works correctly. We could not perform this method on OLCI

during this period due to the lack in acquisitions over DomeC site. Meanwhile we performed an

Intercomparison with MSI_A using PICS method over Algeria-3, Algeria-5 and Libya-1 PICS. The results

are displayed on Figure 22 below. Again the results show that the observed reflectance by OLCI_A is

higher than that one observed by MSI_A.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 23

Figure 22: Time series of the relative difference defined as observed reflectance - simulated one over the later in

per-cent from (black) MSI_A and (blue) OLCI_A for bands 665 nm over (top) Algeria-3, (middle) Algeria-5 and

(bottom) Libya-1 CalVal-sites. Error bars indicate the method uncertainty.

However, as it could be seen on the time-series (Figure 22) that S3A_OLCI shows higher fluctuation than

S2A_MSI and higher bias of about 3-4%. The temporal average of the relative gain coefficients of PICS

method are summarized below on Figure 23.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 24

Figure 23: The estimated gain coefficients (observed/simulated signal) from (black) S2A/MSI and (blue)

S3A/OLCI over (top) Algeria-3 and (bottom) Libya-1 PICS as a function of wavelength. Dashed-green, orange and

red lines indicate the 2%, 5% and 10% respectively. Dashed-black and blue lines indicate the significant gaseous

absorption bands which are mis-represented in the BRDF model of PICS method in DIMITRI. Error bars indicate

uncertainty associated to the estimated gain coefficients.

5- Synthesis of DIMITRI results

After the processing of the dataset over PICS, we suspected somehow that the radiometry calibration of

OLCI is slightly high. Then the absolute calibration coefficient over Rayleigh scattering confirmed such

thoughts. It strongly suggests that the calibration of OLCI products requires further refinement.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 25

Table 4: Synthesis of the estimated absolute, interbands and relative vicarious calibration coefficients

Rayleigh Glint PICS

S3A/OLCI channel

Wavelength (nm)

Vic. Cal. Coeff.

Uncertainty Interband.

Coeff.

Uncertainty Vic. Cal.

Coeff. Uncertainty

Oa01 400 1.058 0.030 0.978 0.01 1.04377 0.040

Oa02 412 1.057 0.030 0.983 0.01 1.0234 0.020

Oa03 443 1.040 0.029 0.987 0.01 1.02496 0.020

Oa04 490 1.039 0.029 0.982 0.01 1.03085 0.020

Oa05 510 1.022 0.029 0.991 0.01 1.02832 0.020

Oa06 560 1.003 0.028 1.006 0.01 1.02283 0.020

Oa07 620 0.999 0.028 1.01 0.01 1.02753 0.020

Oa08 665 1.001 0.028 1 0.01 1.02322 0.020

Oa09 674 1.006 0.028 0.996 0.01 1.02498 0.020

Oa10 681 NA NA 1.003 0.01 1.03399 0.020

Oa11 708 NA NA 1.003 0.01 1.06999 0.045

Oa12 753 NA NA 1.182 0.012 1.03536 0.020

Oa13 761 NA NA 1.014 0.01 0.834571 0.122

Oa14 764 NA NA 1.005 0.01 0.913419 0.041

Oa15 768 NA NA 1.009 0.01 1.03598 0.020

Oa16 778 NA NA 0.978 0.01 1.02434 0.020

Oa17 865 NA NA 0.983 0.01 1.02178 0.020

Oa18 885 NA NA 0.987 0.01 1.02213 0.020

Oa19 900 NA NA NA NA 1.00803 0.163

Oa20 940 NA NA NA NA NA NA

Oa21 1020 NA NA NA NA NA NA

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 26

3 Level 2 Land products validation

[OLCI-L2LRF-CV-300]

Task Description This task encompasses activities associated with overall initial

assessment of performance over land and in particular TCI and

troubleshooting activities.

Summary of activities Three activities were performed within this task: (i) overall data

format, (ii) geometric accuracy and (iii) radiometric quality.

Percentage of completion (phase E1) 20%

Description of activities done and results

i) Overall format

Unlike MERIS L2 data, individual Rayleigh corrected spectral bands are not incorporated in the OLCI L2

products. These would be useful to allow their relative contribution to be understood, and to enable

non-canopy factors such as atmospheric aerosol content and soil background to be investigated.

ii) Geometric accuracy

Substantial issues with the geometry of the L2 products were identified, with pixel displacement of up to

5 km in some cases. These issues were found to be restricted to NT products, so further analysis has

been conducted on NR products.

iii) Radiometric quality

No radiometric discrepancies have been observed through visual examination. OTCI data for 15th June

were compared with historical MERIS MTCI for same day of acquisition. The distribution and shape of

the histograms were near to identical, demonstrating the consistency of the product.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 27

To provide an initial indication of data quality, we investigated the time-series of the OTCI between July

and October for the UK-NFo site. The OTCI demonstrates a slight reduction towards the latter part of

the time-series, consistent with the onset of senescence. Throughout this period an increased number

of products were available, enabling day to day variability in OTCI values to be assessed. The limited

variability in OTCI values over this period indicates that the product is stable.

Time-series of OTCI values over the UK-NFo site (3 x 3 macro-pixel)

To provide another indication of algorithm performance, we have investigated the agreement between

the OGVI and OTCI, as both provide a measure of vegetative vigour. Moderate to strong relationships

have been observed over the core validation sites, indicating that the algorithms themselves are free

from any major issues.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 28

Relationship between OGVI and OTCI over the UK-NFo site (3 x 3 macro-pixel).

Validation of the OLCI FAPAR (OGVI)

It must be recalled here that the OLCI FAPAR is based on generic retrieval that can been optimized to

any optical sensors, as it has been already done for SeaWiFS, MERIS, and MODIS. The resulting products

are thus directly comparable at the same place and time.

Level 2 OLCI FAPAR at 300 m is compared against its JRC counterpart derived from MODIS Collection 6

data at 250 m with the following steps:

1. The data are remapped over same common area at 250 m (or 300 m) using SNAP for OLCI

FAPAR and MRT for MODIS.

2. (Short-term) seasonality trend is then checked against previous years using MERIS-

derivedFAPAR

Benchmark of the two sensors FAPAR products is made using 3x3 averaged values over only when they

are all valid to avoid remaining cloud/cloud shadow effects and minimize geo-location/re-projection

problems.

These comparisons have been done so far above the IT-Cat, IT-Tra, SP-Val, US-Ne1, DE-Geb and UK-NFo

Cal/Val sites. Scatter plots of MODIS vs. OLCI FAPAR results for all sites are presented on Figure 24. They

show a relative good agreement with a correlation agreement of 0.9571, a RMSE equal to 0.07198, and

a bias less than 0.05 which is the range of theoretical uncertainties of the algorithm.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 29

Figure 24: Scatter-plot between match-up OLCI/FAPAR daily values and FAPAR derived from MODIS at 250m.

Further examples of inter-mission comparisons are provided below.

The temporal evolution over US-Ne1 (maize field) of the OLCI FAPAR after the month of August shows

the expected senescence, i.e. decrease (Figure 25).

Figure 25: FAPAR time series from OLCI (blue) and MODIS (green) over the US-Ne1 CV site.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 30

The temporal evolution over IT-Cat (Orange trees) of the OLCI FAPAR after the month of august shows

the expected increase (Figure 26).

Figure 26 FAPAR time series from OLCI (blue) and MODIS (green) over the IT-Cat CV site

The temporal evolution over SP-Val of the OLCI FAPAR after the month of august shows stable values

(Figure 27). Values are significantly lower than those from MERIS for year 2001 but similar to year 2009.

Figure 27: FAPAR time series from OLCI (blue) and MODIS (green) over the SP-Val CV site. Inner panels time

series are MERIS FAPAR for Summer and Fall 2001 (top panel) and 2009 (bottom panel).

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 31

[OLCI-L2LRF-CV-410 & OLCI-L2LRF-CV-420] – Cloud Masking & Surface Classification for Land Products

A first analysis on Sentinel-3 Level 2 water and land product flagging quality was conducted for the

S3MPC OPT Commissioning Phase Report (S3MPC.CLS.REP.012 - i1r0 - STM Commissioning Phase

Report; 29.06.2016). A second, dedicated analysis was then conducted afterwards and summarised in a

technical note (Sentinel-3 Flag Quality Analysis – Technical Note; 19.10.2016). The first and second

analysis of Sentinel-3 OLCI level 2 FR/RR land and water products cloud and snow flags have shown that

the flag quality is reasonable, but still needs improvements.

The main issues include false detection of clouds as snow/ice, omission of snow/ice over land or false

detection as cloud, omission of clouds in sun glint areas, omission of some cloud border pixels, false

detection of desert and beaches as clouds, false commission of bright cloud tops as snow/ice as well as

omission of some small clouds and haze by the cloud flag. Some issues are specific for the product (land

or water), others are more related to the resolution or location, while most of the issues concern all

products and resolutions.

The following list shows all product and resolution specific issues.

Land FR products:

The coastline at boarder of ocean and desert and most beaches are falsely detected as clouds.

Bright sand areas are sometimes detected as cloud.

Top of bright clouds is often detected as snow/ice.

Some issues are limited to higher latitudes.

Classification as cloud or snow/ice seems to be less reliable and can’t be retraced. Overall the

classification of either cloud or snow/ice seems arbitrary.

Dark, wet sea ice is mainly not detected as snow/ice.

Land RR products:

The coastline at boarder of ocean and desert and most beaches are falsely detected as clouds.

Sometimes small clouds and haze are omitted.

Bright sand areas are sometimes detected as cloud.

Top of bright clouds is often detected as snow/ice.

Some issues are limited to higher latitudes.

Classification as cloud or snow/ice seems to be less reliable and can’t be retraced. Overall the

classification of either cloud or snow/ice seems arbitrary.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 32

4 Level 2 Water products validation

[OLCI-L2WLR-CV-300, OLCI-L2WLR-CV-310, OLCI-L2WLR-CV-32, OLCI-L2WLR-CV-330, OLCI-L2WLR-CV-

340, OLCI-L2WLR-CV-350, OLCI-L2WLR-CV-360 and OLCI-L2WLR-CV-370] – Level 2 Water-leaving

Reflectance product validation

Activities done

In situ data collection, matchup extraction and statistics computation is now operational for AERONET-OC data.

First BOUSSOLE (Mediterranean Sea) and IML (Gulf of Saint Lawrence, Canada) data have been provided by LOV and ARCTUS. Matchups extraction have been performed for both.

Matchup extraction and statistics computation is now operational and refreshed every Sunday for all available datasets.

Side-by-side in situ and OLCI trends are now generated at level-2 for available stations or moorings (AERONET-OC, BOUSSOLE, IML ; total of 17 stations)

Up to 200 useful matchups after only 5 months of OLCI Level-2 production

OLCI performances against in situ data are refreshed once a week to produce the plot and statistics

presented below.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 33

lambda N RPD |RPD| slope intercept r2

400 21 26% 32% 0.7037 0.0109 0.1005

412 188 140% 151% 0.7562 0.0060 0.4104

443 188 73% 81% 0.8815 0.0046 0.5895

490 156 45% 48% 0.9963 0.0039 0.6495

510 69 35% 38% 0.8331 0.0035 0.8501

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 34

560 182 20% 25% 1.0896 0.0012 0.7371

620 21 44% 46% 0.5018 0.0010 0.2147

665 137 19% 45% 0.7603 0.0008 0.5132

681 28 46% 54% 1.0062 0.0003 0.6614

709 20 112% 121% -0.2540 0.0005 0.0105

These first validation results show that OLCI reflectance products are not within the requirements but

the other all picture is nonetheless very promising. Statistics in terms of slope and r2 are good for this

stage of the mission and it turns out that blue and green bands mainly show a positive biais with a small

dispersion. This indicates that providing the Level1 products is stabilised, the implementation of a

vicarious adjustment would significantly improve the results.

As expected, we can observe an increasing dispersion with the shorter wavelength. This kind of issues

will not be sorted by vicarious adjustment. L2 processing steps like BPAC should be assessed in priority

to investigate this phenomenon. A closer look in to level2 flags will also be necessary to ensure that the

dispersion is not caused by erroneous flagging.

It is important to note that the number of usable matchup is significant after only a few months of OLCI

operation with up to 188 mathcups at band 412and 443.

The four graphs presented below show the in situ and satellite time series for AERONET-OC Galata

station. As it could be observed on time series, the two time series are in good agreement. OLCI L2 data

are nonetheless mission at the beginning of the mission for a complete picture.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 35

Below is another set of examples for AERONET-OC AAOT, Gloria, Helsinki Lighthouse and Gustav dalen

Tower. It is very promising to observe that despite the complexity of the sampled water bodies (Adriatic

Sea, Black Sea and Baltic Sea), both in situ and satellite Time Series are in good agreement. It is recalled

here

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 36

Activities planned during the next period

OLCI-L2WLR-CV-200 and OLCI-L2-CV-210

Initiate the implementation of NIR band vicarious adjustment routines and test them providing

we have sufficient L2 RR marine data over South Pacific and Indian Ocean regions

Difficulties encountered

Difficulty to retrieve OLCI Level-2 products for CalVal activities

[OLCI-L2WLR-CV-510 & OLCI-L2WLR-CV-520] – Cloud Masking & Surface Classification for Marine

Products

Summary of the Sentinel-3 flag quality analysis

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 37

A first analysis on Sentinel-3 Level 2 water and land product flagging quality was conducted for the

S3MPC OPT Commissioning Phase Report (S3MPC.CLS.REP.012 - i1r0 - STM Commissioning Phase

Report; 29.06.2016). A second, dedicated analysis was then conducted afterwards and summarised in a

technical note (Sentinel-3 Flag Quality Analysis – Technical Note; 19.10.2016). The first and second

analysis of Sentinel-3 OLCI level 2 FR/RR land and water products cloud and snow flags have shown that

the flag quality is reasonable, but still needs improvements.

The main issues include false detection of clouds as snow/ice, omission of snow/ice over land or false

detection as cloud, omission of clouds in sun glint areas, omission of some cloud border pixels, false

detection of desert and beaches as clouds, false commission of bright cloud tops as snow/ice as well as

omission of some small clouds and haze by the cloud flag. Some issues are specific for the product (land

or water), others are more related to the resolution or location, while most of the issues concern all

products and resolutions.

The following list shows all product and resolution specific issues.

FR Water products:

Some clouds are arbitrarily detected as snow/ice.

Clouds are sometimes not detected in sun glint areas

Snow/ice over land is often undetected or detected as cloud

Sometimes cloud border pixel are omitted

False detection of snow/ice which could be a false commission of cirrus clouds as snow/ice

Water RR products:

Clouds are sometimes not detected in sun glint areas

Sun glint is flagged over land

Reflectance values are over corrected at undetected clouds

Snow/ice over land is often undetected or detected as cloud

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 38

5 Level 2 SYN products validation

[SYN-L2-CV-100]

Description of activities done and results

This activity has been started recently with the verification of the GPP-like tool.

Two main objectives are gathered in this requirement. The first one deals with the consistency between

SYN L1c and L1B parameters (same parameters included and same order of magnitude). Note that this

verification is only possible using the SYN breakpoints as the L1B parameters are no longer included in

the SYN L1 product.

During the S3 commissioning, we checked the well-behaviour of the GPP-like tool. From the SYN L1

breakpoints, this tool re-creates the SYN L1c product in the same format that the one specified in the O-

GPP project (including L1B parameters on the acquisition grid). This verification process includes the

self-consistency between OLCI/SLSTR L1B and SYN L1c parameters.

The second objective of this requirement is to ensure that the de-registration, i.e. the retrieval of the

OLCI and SLSTR original projection, is performed correctly. This activity has not been yet performed.

Note that these activities can only be performed using GPP-like tool. As consequences, a generic

procedure is currently set up with S3 MPC operators. Each week, 4 or 5 land granules will be

reprocessed with SYN L1c by enabling the production of SYN L1 breakpoints. These breakpoints will be

then taken as inputs by the GPP-like tool and archived to be compared statistically to the original SLSTR

and OLCI L1b products. In addition, each month, these parameters will be fully compared and the de-

registration checked.

Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 39

6 Events

Two OLCI Radiometric Calibration Sequences have been acquired during Cycle 010:



Sentinel-3 MPC


Cycle No. 010


Issue: 1.0

Date: 22/11/2016

Page: 40

7 Appendix A

Other reports related to the Optical mission are:

S3-A SLSTR Cyclic Performance Report, Cycle No. 010 (ref. S3MPC.RAL.PR.02-010)

All Cyclic Performance Reports are available on MPC pages in Sentinel Online website, at:

https://sentinel.esa.int

End of document

https://sentinel.esa.int/

S3-A OLCI Cyclic Performance Report...S01 sequence on 21/10/2016 10:03 to 10:05 (absolute orbit...

Documents

Transcript of S3-A OLCI Cyclic Performance Report...S01 sequence on 21/10/2016 10:03 to 10:05 (absolute orbit...