Page 1 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Design Justification v2 overview
Samantha Lavender
Page 2 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Work Packages
15 Jan2006
CDR
Page 3 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Design Justification v2
• 5.2 In situ Characterisation
• 5.3 Coastal Waters
• 5.4 Sensor Cross Characterisation
• 6.3 Merging Algorithm Sensitivity Analysis
Page 4 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
In situ characterisation Samantha Lavender and Yaswant Pradhan
Page 5 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Characterisation review
• In situ data
• MERIS
• MODIS
• SeaWiFS
• Parasol: data not available at present for characterisation
• Overall Conclusions
Page 6 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
In situ data sets
• NOMAD (2002 onwards)• Publicly available SeaBASS (2002 onwards)• NILU database• Boussole buoy
Page 7 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
On-line Database
Page 8 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Spatial Coverage
NOMAD SeaBASS
NILU
Page 9 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
NOMAD In Situ Data Conversion to Fully Normalised Water Leaving
Radiance #412nm
y = 0.6918x + 0.0576
R2 = 0.9383
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5
OBPG nLw
Glo
bC
OL
OU
R n
Lw
#443nmy = 0.7558x + 0.0349
R2 = 0.9544
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5
OBPG nLw
Glo
bC
OL
OU
R n
Lw
#490nmy = 0.8278x + 0.0012
R2 = 0.9671
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5
OBPG nLw
Glo
bC
OL
OU
R n
Lw
#510nmy = 0.8361x - 0.0034
R2 = 0.9859
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5
OBPG nLw
Glo
bC
OL
OU
R n
Lw
#555nmy = 0.8351x - 0.0042
R2 = 0.9922
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5
OBPG nLw
Glo
bC
OL
OU
R n
Lw
#670nmy = 0.8221x - 0.0027
R2 = 0.9976
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5
OBPG nLw
Glo
bC
OL
OU
R n
Lw
OBPG nLw
Glo
bCO
LO
UR
nL
w
412 443 490
510 555 670
Page 10 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Data Processing L2 (M)LAC
L3-DDS Generator
In-situ meta
L3 DDS In-situ dataIn GC-NOMAD template
DDS Match-UpTimediff
<24 hrs
Locationdiff
<=0.02°NO
match-up
NO match-up
Extract 3x3 kernel
L3-DDS Reader GC in-situ Reader
Y N
Import to ExcelStat Template
N
Preparation/GenerationExtraction
Statistics/Result
Tdiff< 24 hrsFLAG !=NoData
Tpix > 5
Match-up Result
Y
Exclude
N
Page 11 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
MERIS MODIS SeaWiFS Total BOUSSOLE 244 210 311 765 NILU 32 41 47 120 NOMAD 151 107 244 502 Total 427 358 602 1387
Data Processing
Number of generated DDS
Page 12 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
MERIS
Page 13 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
MERIS
Page 14 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
MODIS
Page 15 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
MODIS
Page 16 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
SeaWiFS
Page 17 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
SeaWiFS
Page 18 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
SeaWiFS
Page 19 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Normalised Water Leaving Radiance
• Further discussion and analysis is occurring with respect to the derivation of in-situ normalised water leaving radiances as this is a key step in the characterisation process.
• Propose that this work should be ongoing and the characterisations will be updated as additional insitu data becomes available.
• The results presented so far indicate that it is particularly important to seek out datasets with high normalised water leaving radiances.
Page 20 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Chlorophyll & GlobCOLOUR Kd
MERIS MODIS-Aqua (Literature)
SeaWiFS
(Literature)
Chl
Bias and RMS
-0.077
0.479
-0.093
(-0.084)
2.551
(0.644)
-0.112
(0.006)
1.214
(0.657)
Kd (Literature is sensor
algorithms)
Bias and RMS
0.025
0.017
0.065
(-0.018)
0.103
(0.046)
0.051
(0.001)
0.063
(0.520)
Page 21 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
DJF V2.0 Coastal Waters Kai Sørensen and Jo Høkedal
Page 22 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Coastal waters - Guianas CoastMERSEA-IP
• The provinces, Guianas Coastal (GUIA) and Guinea Current Coastal (GUIN) are both coastal stripes influenced by land and river inputs.
• On the African side (GUIN) there is also a strong impact of atmospheric conditions (cloud coverage, biomass burning and desert dust aerosols) on the ocean colour products.
• The two provinces are characterized by the largest differences of the provinces (in this study) between sensor products.
• Between SeaWiFS and MODIS–Aqua the differences (defined as the root mean square relative difference) was as high a 21.3 % and 24.7 % on average for GUIA and GUIN, respectively.
• The differences compared to MERIS are 3-4 % higher.
Page 23 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Coastal water - Baltic SeaMERSEA-IP and FerryBox-EU
• An optically complex water with a high load of CDOM, and summer blooming of Cyanobacteria causing large changes in the IOPs.
• An average difference of MERIS vs SeaWiFS or MODIS-Aqua of around 25%, while between SeaWiFS and MODIS-Aqua of 19.2 %.
• MERIS Algal_1 and Algal_2 show erroneous data in the bloom, but Algal_2 after the 2nd processing gave better agreement.
• Even if the MERIS Neural Network Case 2 products can be trained for this area it will be problematic due to the high IOP variability.
• The validation will also be a challenge during such extreme blooms.
Page 24 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
North Sea – Skagerrak Case1 Chl-a Algorithms, Folkestad, 2005
R2
Sensors compared (small areas of 25 pixels)
All stations
Without #7
MODIS/Aqua vs MERIS 0.60 0.76
SeaWiFS vs MERIS 0.15 0.44
SeaWiFS vs MODIS/Aqua 0.82 0.91
MODIS/Aqua vs MERIS SeaWiFS vs MERIS SeaWiFS vs MODIS/Aqua
Page 25 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
MERIS Skagerrak (2nd processing)Sørensen, 2006.
57.5 58 58.5 59 59.5 600
1
2
3
4
5
6
7
8
9
10200305
KLA
mg/
m- 3
Breddegrad
algal2fluor
MERIS Algal_2 vs Chl-a_HPLC MERIS Algal_2 binned one month vs Chl-a fluorescence from the Ferrybox systems (+/- 1. Stdev.dev.)
Central Skagerrak
Danish Coast
Oslo Fjord
Page 26 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Coast and Open Sea – Spatial variability
400 450 500 550 600 650 700 750 800 850 9000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Wavelength
w
(sr
-1)
2002-07-29 10:07:00, lat=51.3083, lon=2.85
in situ1 of11 of19 of99 of922 of2522 of2537 of4937 of4956 of8156 of81
400 450 500 550 600 650 700 750 800 850 9000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Wavelength
w
(sr
-1)
2002-07-29 10:22:36, lat=63.92, lon=0.7
1 of11 of19 of99 of925 of2525 of2549 of4949 of4981 of8181 of81
Vertical bars: Max-min
Vertical bars: Max-min
Page 27 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Summary• It is clear from the findings by many authors that SeaWiFS and MODIS do not
resolve the true values in Case 2 water and that multivariate complex Case 2 waters need to have complex algorithms e.g. MERIS NN.
• It is presently difficult to give any recommendation on how to solve the issue of combining data from different sensors in coastal water without dealing with all the Case 2 problems.
• The only combining possibilities is then to merge MERIS Case 2 products with Case 1 products, but boundaries will probably be present.
• Alternative are to use Case 1 algorithms into the coast and flag Case2 water. To be discussed.
Page 28 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Sensor Cross Characterisation
Antoine Mangin
Page 29 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Cross characterisation
Cross comparison between MERIS/MODIS/SeaWifs – attempt to detect systematic biases: At global scale and regional scale
Check of the consistency with JRC results
Harmonisation of Kd algorithm
Page 30 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Cross comparison between MERIS/MODIS/SeaWifs – attempt to detect systematic biases: At global scale and
regional scale
comparison
Page 31 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5
SW:MO
SW:ME
MO:ME
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5
SW:MO
SW:ME
MO:ME
March03 06 09 12
03 06 09 12
Slope of the regression
Determination coeff. r2
Med
iter
rane
anSummary for Mediterranean
Page 32 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
0.840.860.880.9
0.920.940.960.98
11.021.04
0 1 2 3 4 5
SW:MO
SW:ME
MO:ME
0.8
0.82
0.84
0.86
0.88
0.9
0.92
0.94
0.96
0 1 2 3 4 5
SW:MO
SW:ME
MO:ME
03 06 09 12
03 06 09 12
Slope of the regression
Determination coeff. r2
Glo
bal
Summary for Global results
Page 33 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
S:A S:M M:A RMS % 11/13/15 14/16/18 14/15/17
Bias 3/5/9 -2/0/4 1/5/9
From JRC’s assessment:
Global
Regional: very fluctuant, seasonal dependency – in agreement with our daily results
Confrontation with other sources
There is a bias between sensors
Page 34 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Either…. …or….
We get a faithful caracterisation of bias wrt season and region and correct for it prior to merging.
We anticipate the impact of using biased data.We apply inputs as is.The impact will be reflected into the error bar estimates wrt to season/region
Not mature enough Recommended
Page 35 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Harmonisation of Kd algorithm
Kd
Page 36 GlobColour CDR Meeting – July 10-11, 2006, ESRIN
Overall Conclusions
• Used some large databases and produced a large number of DDS files (1387), but as is often the case with ocean colour data the number of match-up points is considerably smaller than the number of original insitu points. • The characterisation will undergo additional work within the next couple of months to tie up the loose ends and come to a final set of conclusions.• For now the merging will use the following characterisation results:
• normalised water leaving radiance: GlobCOLOUR• chlorophyll: NASA (will split GlobCOLOUR into low/high groupings)• diffuse attenuation coefficient: GlobCOLOUR
•For Case 2 waters, a decision on the alternatives of using (1) MERIS Case2 products for the coast or (2) using Case1 products only with flagging information must be taken.