The sensors studied

1
Merging of ocean color datafrom multiple
satellite missionsC. Pottier1,2, V. Garçon1, A.
Turiel3, G. Larnicol2, J. Sudre1, P.Y. Le Traon4
, P. Schaeffer2 1 LEGOS 2 CLS 3 ICM
4 IFREMER
The sensors studied
SeaWiFS MODIS-Aqua
Launch date August 1st, 1997 May 4, 2002
Equator crossing 12h 13h30
Spectral coverage (nm) 402 885 (8 bands) 405 14385 (36 bands)
Data available since 04/09/1997 (Level 3) (OC4v4 R51) 29/11/2002 (Level 3) (OC3M R11)
Original spatial resolution 1100 m 1000 m
Level-3 data put on a regular grid 1/12x1/12 1/24x1/24
Globcolour 1st User Consultation Meeting,
Villefranche sur Mer, December 4-6, 2006
2
Methods and zone of interest

• Error weighted-averaging
• Objective analysis
• Global application for 2003
• Wavelets (ongoing work)
• gt Gulf Stream area

Biogeochemical provinces
Longhurst, 1998
3
SeaWiFS 2410 matchups

• YEAR 2003
• AMT 12 and AMT 13 (Atlantic Meridional Transect,
  UK)
• SeaBASS (SeaWiFS Bio-optical And Storage System,
  USA)
• National Oceanographic Data Center (USA)
• International Council for the Exploitation of the
  Sea
• National Institute of Oceanography (India)
• OISO
• MINERVE 0203-R2, 0204-R4, 0304-R0, and 0304-R1
• ARGAU3
• DIAPAZON 7, 8, and 9
• DYFAMED
• Other cruises (Yves Dandonneau, pers. com.)

MODIS/Aqua 2789 matchups
4
SeaWiFS error measurement
MODIS/Aquas error measurement
of chlorophyll
of chlorophyll
5
SeaWiFS bias
MODIS/Aquas bias
of chlorophyll
of chlorophyll
6
Error-weighted averaging Objective
analysis in Pottier et al., IEEE TGRS, 2006
7
Error-weighted averagingPottier et al., IEEE
TGRS, 2006Gregg et al., IOCCG Report, 2006

• Input parameters for each sensor
• Map of chlorophyll for the ith day
• Map of spatial errors (according to matchups)
• Computation made on logtransformed values
• Output products
• Map of combined chlorophyll for the ith day
• Map of associated errors

8
Combined data
SeaWiFS
MODIS/Aqua
Chlorophyll-a concentration (mg/m3)
Error
Spatial repartition of the data
of chlorophyll
03/26/2003
9
Objective analysisPottier et al., IEEE TGRS,
2006Gregg et al., IOCCG Report, 2006

• Aim to determine the value of a field ? at a
  point in space and time, given various
  measurements of the field unevenly spread over
  time and space ?obsi (i1n)
• Best least squares linear estimator ?est(x)
  (Bretherton et al., 1976)
• A priori knowledge of
• The covariance function (i.e. variance
  correlation function)
• The measurement noise and the bias (obtained from
  the matchups)

10
SeaWiFS
Combined data
MODIS/Aqua
Chlorophyll-a concentration (mg/m3)
Error
Spatial repartition of the data
of chlorophyll
03/26/2003
11
Wavelets
Use of multi-resolution algorithm to compute the
wavelet coefficients
 h  convolution with a low-pass filter  g 
convolution with a high-pass filter  ?2  keep
1 row (or col.) on 2
12
Wavelets Pottier, Turiel and Garçon1st step
estimation of the missing data
Vertical details
13
Wavelets Pottier, Turiel and Garçon2nd step
combination
14
MODIS/Aqua
SeaWiFS
Original image (1/12)
Original image (1/24)
03/26/2003
Reconstructed image
Reconstructed image
Weighted averaging (1/12)
Objective analysis (1/12)
Wavelets (1/24)
15
Ongoing work

• Weighted averaging and objective analysis
• Fully operational for near real time at the
  global scale
• Statistics computation needed for all time series
• Addition of other sensors (MERIS)
• Wavelets very promising ! But still a lot of
  work to do !

16
Oceanographic applicationusing combined SeaWiFS
and MODIS/Aqua ocean color data

• Southern ocean (35S 50S)
• What are the dominant modes of variability in sea
  level anomalies and surface chlorophyll
  concentrations in this circumpolar belt ?

(Pottier et al., 2004 Pottier et al., 2006)
17
2002-2006 period 3/0 mode of variability
Band 35S-45S
Weekly sea level anomalies
Weekly combined chlorophyll SeaWiFS MODIS/Aqua
Monthly SeaWiFS chlorophyll
Band 45S-50S
18
(No Transcript)
19
Multi-resolution wavelets algorithm
20
In conclusion

• Within MERSEA,

21
Error