ANTARES 4 - SeaWifs data
[Version française]

The present pages provide access to the SeaWifs data specially acquired and processed by NASA in support of cruise ANTARES 4. From mid-December 1998 to the end of the cruise (until about 20st February 1999), the individual scenes recorded over the area (35-65°S, 50-80°E) were received by Jacqueline Etcheto and Viviane Leboucher (LODyC, Paris), who retransmitted them onboard Marion Dufresne to Jacques Le Fèvre (UMR 6539 "Bioflux", Brest) who performed subsequent processing. Receipt onboard generally occurred within 3-4 days of data acquisition by the satellite. The immediate role of these quasi-real time data was to help, together with some other high-resolution or quasi-synoptic data (XBT transects, Tow-Yo), in conducting the cruise, and especially in choosing the position of process-study stations. Their exploitation a posteriori is a source of information on the time variations of phytoplankton biomass distribution and on frontal dynamics (evolution of meanders and other secondary structures associated with ageostrophic circulation), thus providing a means of putting the field data from the cruise into their regional perspective.

Processing applied to the data

The individual scenes provided by NASA are geometrically corrected images whose colour range approximately corresponds to the scale used by NASA to globally express chlorophyll concentrations from SeaWifs data. The original files were all named with a suffix indicative of the JPEG format, but most of them actually were in the RAS format (Sun Raster: uncompressed bitmap graphics, 256-colour indexed palette, RGB encoding). These individual scenes are both too restricted in their spatial coverage to be directly usable for the purpose stated above and beset with a high noise level. The noise sources are twofold: a spatial noise due to the fragmentation of cloud-free areas, and a signal noise showing up in the form of strange pixel colours unrelated to the scale of pigment concentration.

The fragmentary character of individual scenes is usually circumvented by constructing composites. The latter integrate (generally by computing a mean value) the data for a given pixel (or group of pixels) over a certain duration. This technique results in some smoothing of the spatial structures, all the more so if the integration time is long. It also requires access to the numerical data from the sensor, which were not available onboard. Mosaics were thus constructed instead, which are spatial combinations of several images covering different areas, with a single signal value (colour) from a single image being retained for a given pixel. This technique preserves the full resolution of spatial structures, as they may appear on individual scenes, but the result may give the illusion of being more synoptic than it actually is, if the component images were collected over too long a duration. The choice made was to construct a series of successive mosaics, updated daily by retaining the most recent data for any given pixel: the areas still empty at day D were filled up with the relevant significant pixels from day D+1, and the pixels significant at day D were replaced with the relevant significant pixels from day D+1 if any. The series of succesive daily mosaics thus allows for a detailed follow-up of the evolution in time of the structures under study. In addition to this main time series, a few mosaics were also constructed to reflect the situation over a short duration (of the order of a few days) whenever a relevant set of high quality images were available. During the cruise a HTML site was set up on a Unix server onboard Marion Dufresne, to provide self-service access to individual images and mosaics. The site was updated as frequently as possible, as soon as new data were received and processed.

The mosaics were in practice constructed by using image-editing software. The basic technique roughly consists in transparently pasting the most recent images over the immediately preceding mosaic. "Transparent" pasting essentially means that the empty areas in the new image should leave as they are the significant pixels in the old one, if any are found in the same place. As stated above, however, individual images have a high background noise and include a large number of pixels whose colour is unrelated to any chlorophyll content, but where the nominal signal is non-null. Such pixels may replace more significant ones, sometimes over relatively large areas, thus jeopardizing the quality of the mosaics obtained. A filtering technique was accordingly applied prior to image pasting, in an attempt to remove the background noise. Filtering was effected by manipulating the 256-colour palette of the RAS files, after the few files received in JPEG format were converted to RAS. Noise reduction is followed by an adjustment "to the nearest colour" of the RAS palette to the palette from the global NASA colour scale for chlorophyll concentration in SeaWifs images. Mosaics are saved in the GIF format, whose characteristics are similar to the RAS format (bitmap graphics, 256-colour indexed palette, RGB encoding), but where a compression algorithm allows for a reduction in file size without altering the data (contrary, in this respect, to the JPEG format). Three filters of increasing efficiency were successively devised. The protocol used during most of the cruise did not include a colour adjustment to the NASA scale for chlorophyll concentration and the final residual noise was quite noticeable. The data were reprocessed, and the onboard HTML site updated, between the call at Crozet on 18th February 1999 and the arrival at La Réunion on 23rd February 1999, with an improved filter and an adjustment to the NASA scale performed in fine on the mosaics.

A complete data reprocessing was finally effected prior to the build up of the present site. The filter is more efficient and the adjustment to the NASA scale is performed before pasting the individual images. The smallest pixel groups in the latter are removed, in order to also decrease the spatial noise. The mosaics were reconstructed by incorporating 17 images (mainly from December 1998) that had failed to be retransmitted to the ship. A new scale of crorrespondence between colours and chlorophyll concentrations has been devised. This scale is more specific of the ANTARES 4 data range than the NASA global scale and better displays variations in phytoplankton biomass as well as some hydrodynamic structures. The mosaics reconstructed after the cruise are available here in both their NASA scale version and their ANTARES 4 scale version. In both cases, however, it is advisable not to be too confident in the absolute values of pigment concentration. The series of daily mosaics has also been used for constructing an animated presentation of the time evolution of the sytem under study, which can be displayed by WWW browsers. The starting point of the animation, which is displayed as a loop, is the situation on 31st December 1998. Subsequent mosaics are used in sequential order as the component frames in a movie.

Contents of the present pages

These pages were modelled after the site set up onboard Marion Dufresne. The images are accessed from an index page by clicking on thumbnails. This loads a presentation page for the image, where the time references of the latter are displayed, together with the scale of chlorophyll concentration and a few links (to image enlargement, return to the index, if applicable to the version with the alternative chlorophyll scale). Three index pages, bearing themselves a return link to a general index, are available:

Reprocessed data: This page provides access to the mosaics reconstructed a posteriori. The latter are presented by default in their ANTARES 4 scale version but can also be displayed in their NASA scale version. The page also provides access to time series made up of mosaic sectors from selected dates spread over the cruise duration and to animations covering the first two months of 1999. The perimeter of the area covered by the Tow-Yo operation is superimposed on the short-time mosaics (cf. supra) and on the component images of the time series.

Source images: As stated above, the individual images are presented raw, without any other processing than the conversion of the files to the GIF format with the purpose of decreasing their size.

Marion Dufresne site: Mirror of the site set up onboard, in its final condition, i.e. including the very last images received in the early morning of 23rd February, while Marion Dufresne was already docked in La Réunion, which were processed immediately. A single index page gives access to mosaics, time series (NASA scale only) and individual images. This page is partly redundant, since the preceding one provides more complete access to the individual images and since cruise-time mosaics are in effect replaced by those, also more complete in several respects, recontructed a posteriori. It is retained both to keep trace of the protocol set up onboard and on methodological grounds. Comparing cruise-time mosaics and time series with their counterparts reconstructed a posteriori (for date correspondence, see the reference tables mentioned below) shows, in particular, the effect of the gaps in the series of images received onboard and the increasing efficiency of the successive attempts at noise removal.

In addition to these three index pages giving direct access to images, reference tables provide the list of available images, their date and time of acquisition, the list of mosaics with their component images.

The image files (individual scenes and mosaics) are all stored in the same directory, from which they can be imported directly (click on the link to display the directory contents). This requires knowledge of the exact name of the relevant file (cf. infra) and finding it among more than 100 other ones. It is easier to import the file when connected to the presentation page of the image, by clicking on the latter with the mouse's right button.

File names are compatible with the MS-DOS norm (8 characters or less, followed with a period and a 3-character suffix indicative of the file format) and are therefore acceptable by all operating systems on all platforms supporting WWW browsers. The files for individual images are named after their date of acquisition. The 4 digits of the year are followed with the 3 digits of the calendar day (often called "Julian day", which is improper since reference is actually made to the Gregorian calendar, not the Julian calendar) and, if applicable, with a letter indicative of the order of acquisition of the image whenever several of them are available for a given day. Thus, 1998356.gif is the file name for the single image recorded on 22nd December 1998 and 1999045c.gif the file name for the third image recorded on 14th February 1999. A similar convention is used for the mosaics constructed a posteriori. Thus, m1999006.gif is the file name for the mosaic obtained by cumulating all available images up to, and including, 6th January 1999. The leading m is actually used for mosaics whose colours are coded according to the ANTARES 4 scale (cf. supra). A leading n is used instead in file names for mosaics in NASA scale version (i.e. n1999006.gif in this example). The mosaics constructed onboard were named with sequential numbers beginning with mosaic00.gif (cumulation of available images up to 6th January 1999, considered the starting point) and incremented up to mosaic33.gif (cumulation of all images up to the last ones, recorded on 21st February 1999). A new mosaic was constructed for each day where new images were avalaible, but gaps reaching up to 6 days sometimes occurred in the acquisition or retransmission (cf. supra) of data. These gaps are not reflected in the file names: mosaic19.gif of 7th February 1999 thus directly follows mosaic18.gif of 1st February. For more detail, e.g. file names for short-duration mosaics, please consult the reference tables.


A technical note may be consulted concerning the pages on this site.

Back