Sitemap Contact Kalender Nieuw Home

Projectendatabank FEDRA





Reconstruction of colour scenes (RECOLOUR)

Onderzoeksproject SR/00/111 (Onderzoeksactie SR)

Personen :

  • Dr.  BECKERS Jean-Marie - Université de Liège (ULG)
    Coördinator van het project
    Betoelaagde Belgische partner
    Duur: 1/12/2006-30/11/2008
  • Dr.  RUDDICK Kevin - Koninklijk Belgisch Instituut voor Natuurwetenschappen (KBIN)
    Betoelaagde Belgische partner
    Duur: 1/12/2006-30/11/2008

Beschrijving :

Context and objectives

The project RECOLOUR is a spin-off project from BELCOLOUR. The synoptic imagery of ocean from satellite provides unique spatial coverage to help answer environmental management and research questions. For numerous applications it is more useful to have a complete time series of data at all points rather than data which is gappy in space and time because of clouds, masking of low confidence data and absence of acquisitions. The objectives of RECOLOUR are to reanalyze BELCOLOUR and Pathfinder scenes. Specific objectives are to exploit Empirical Orthogonal Functions (EOFs) to provide an optimal methodology for fillling missing data in space and time and generate complete series of gridded products of SST, Chl a, TSM and associated error fields. RECOLOUR will also synthesise the variability of the studied systems (southern North Sea, Mediterranean Sea around Corsica) by identifying dominant modes of variability. RECOLOUR will provide a methodology to detect anomalous pixels resulting from imperfect cloudmasking. Multivariate approaches will extend the EOF to include informations from wind-stress and tidal amplitudes. RECOLOUR will verify that a multivariate approach provides better estimates than univariate approaches.


Standard Optimal Interpolation (O.I) calculates the analysed field x from the available data d, the observational error-covariance matrix R , the background covariance matrix B and the covariance between the data points and analysis points C according to x = C (B + R)-1 d. When a series of clouded images is to be filled in, the repeated observation on a single grid can be exploited to improve the specification of the covariance functions. This is done with the Data INterpolating Empirical Orthogonal Functions method (DINEOF), where the time series of images provided a mean to calculate principal components of incomplete data as eigenvectors of a covariance matrix, and simultaneously filling in the missing data. The basic idea is that if we knew an EOF expansion of a data set, the truncated expansion profides a filter (or analysed) field everywhere, including at the clouded points. The error and analysed fields can also be used for a posteriori verification of data quality. If the difference of the original data vs analysed data does not fit the error statistics, it means the original data is suspicious and one should look at the data.

Results expected

D1. Southern North Sea satellite data : gridded datasets with documentation on data format and processing
D2. Updated database BELCOLOUR1 produced after the first RECOLOUR processing. New data from BELCOLOUR2
D3. Southern North Sea hydrodynamical model : gridded data from the COHERENS hydrodynamical model
D4. Mediterranean satellite data : the Pathfinder gridded dataset for the Mediterranean
D5. DINEOF reconstructions: EOF patterns and time-evolution
D6. time interpolation at moments not covered by an image
D7. multitemporal averages will be calculated, avoiding bias due to inhomogeneous data coverage
D8. error-covariance matrices of each analysis. Those matrices allow to provide error maps for each image.
D9: mask of suspicious pixels provided, based on the statistics of difference between analysed and original pixels
D10. SeaWifs TSM data will be used in combination with COHERENS model outputs to provide filled images
D11. A tool estimating surface TSM by least square approach on the EOF amplitudes from wind-stress and tide
D12. RECOLOUR products on the BELCOLOUR WWW pages, Mediterranean products on GHER WWW pages

Documentatie :

Over deze website


© 2018 POD Wetenschapsbeleid