The general objective of MODB is to deliver advanced data products to the mediterranean research projects supported by the MAST programme of the European Union. All products are however freely distributed to the whole scientific community. A special effort is being devoted to the preparation and archiving of in situ hydrographic sets which are quality checked, and to the production of gridded climatological fields for temperature and salinity. Software products for data analysis and visualization are also prepared for public release. This project has been developed in collaboration with the MEDATLAS Group.
Among the scientific community involved in the MAST Mediterranean projects, it was felt that research activities could greatly benefit from a comprehensive dataset constructed from all available observations in the Mediterranean Sea. The purpose of this project is to set up the basis for a Mediterranean Oceanic Data Base (MODB) with the associated functions of data gathering, merging, quality-checking, intercalibration, processing and distribution.
Our first mission is to the collect all available in situ observations taken at sea, in order to improve the accessibility, exchange and exploitation of previous or newly acquired data sets. On the climatological time scale, hydrographic measurements are the only parameters offering a fairly good coverage (in space and time) of the entire Mediterranean Sea. Therefore the priority of MODB is to combine the largest possible number of temperature and salinity profiles (multi-bottle, CTD, MBT or XBT type), in order to build the most robust and comprehensive hydrographic data set. The first effort is dedicated to the management of existing (mainly historical) data sets, but continuous upgrading of the data base with additional subsets will be carried out.
The combining of various data sets is not a trivial procedure: one must be very careful to avoid duplication of identical profiles in the combined data set, since it is quite frequent to find the same stationrecorded in various historical data sets. In most cases, the original measurements are identified by their spatial position as well as the date of acquisition; the exact time of measurement is often omitted. Therefore, the header record is usually insufficient to identify unambiguously one given cast, as several measurements might be taken at the same place within a single day. In our experience it can be difficult to identify these occurences, given the rough accuracy of the positioning parameters. In such cases, our merging procedure compares the value of the measured quantities in order to reject the profiles containing the same information.
Any data base unfortunately contains a significant proportion of data corrupted by instrument, operator or communication errors. In order to ensure the quality of the final product, an objective criterion is needed to decide if a measure has to be reconsidered or not. This is the purpose of the quality control procedure. We have divided the checking process into four successive steps: a conformity check to analyse the validity of the profile headers, a physical check to test the physical consistency of each profile separately, a range check to eliminate obviously wrong data, and a statistical check to verify the overall statistical consistency of the data set (by use of the mathematical structure of an analysis method).
Archiving the final product to make it useful to everybody requires the precise definition and documentation of format standards. We have defined an ASCII readable format for the basic representation of hydrographic profile data (MODB/MEDATLAS Profile Data Format). In addition, we have installed several field descriptors to facilite understanding and readability. This leads of course to an increase in the size of the files and the access time. But, on the one hand, good compression algorithms solve the problem of data storage and transmission, and on the other hand, the ASCII bank may be translated to a direct access data bank for use by any kind of model.
From the user's point of view, free access will be given to all quality-controlled data files (within the limits imposed by the data originators, and after extensive validation by the MODB participants). Applicative user interfaces have been developed in order to facilitate the circulation in the data base, and to retrieve the desired information as easily as possible. The physical storage of the data base is performed on a Unix Workstation located on the MODB site (University of Liege). Help-files assist the user with all options and services offered by the data base infrastructure.
Collections of hydrographic data are usually not in a form suitable for straightforward analysis and use. The irregular distribution of hydrographic stations makes the interpretation of the observational information difficult and even hazardous. Conversely, the inspection of synthetic pictures, reconstructed from isolated stations, is very useful for understanding oceanic processes. In addition, the manipulation of data fields on a regular grid is a prerequisite for numerical modelling exercises and for visualisation purposes.
The mapping of data from an irregular network to a regular grid requires the application of adequate interpolation methods. As the observed values are sparse and noisy, the mathematical solution of the projection scheme is generally not unique and the selection of one solution rather than another must be based on an objective criterion. If prior estimates of the statistics are available, the usual method consists in minimizing the expected error variance of the reconstructed field. The separation between the noise and the dominant signal is usually identified after the analysis. A variety of gridding methods have been developed and adapted to the oceanographic problem.
The second task of this project consists in the application of several analysis procedures to the MODB hydrographic data set, in order to reconstruct climatological fields of temperature and salinity in the Mediterranean Sea and the associated seasonal cycle of T/S properties. The resolution of the gridded fields is 1/4 of degree on the horizontal and 19 levels on the vertical.
Our analysis method has been calibrated according to the results of validation and intercomparison exercises. The systematic utilization of the analysis tool produces updates of the climatology as soon as a substantial amount of new data has been accumulated in the MODB. In the terminal phase, the gridded data should be upgraded almost continuously.
The expected benefits from the gridded hydrology will be:
The standard "optimal interpolation" (GANDIN, 1965; BRETHERTON et al., 1976) is probably the most practiced algorithm, but a series of similar schemes have also been derived from the original one (see the list of references below). Normally, only a few parameters have to be adjusted to run these methods, so that it is often easy to accommodate the computer codes into operational "black boxes". This requires a careful validation and intercomparison of the codes, both in simple academic and realistic cases. The analysis scheme chosen for this project is a variational method adapted to the peculiarities of the oceanic problem (BRASSEUR, 1991). The numerical code has been complemented with a user's interface in order to combine data extraction, reduction, gridding and visualization as easily as possible. Quick help and fully documented guides are prepared to assist the user; free access to the analysis package (stored on the MODB workstation) will be made possible through the Internet via ftp anonymous.
With nearly half of the human brain being devoted to visualization, it is worth exploring the data world by means of suitable graphic tools, since this might result in a more efficient use of human capabilities (in particular, faster inspection, analysis, memorisation and comparison), widely extending the possibilities of scientific research. This is particularly true in the case of in situ oceanographic data, which must be represented in a three-dimensional (space) or even four-dimensional (space+time) environment.
The science of data visualization itself constitutes a vast research field which extends beyond the limits of this project. So, our intention here is just to accommodate existing visualization tools, with the aim of getting the best possible perception of the data sets produced at MODB. Two different types of graphic softwares are needed by the scientists. First, relatively simple tools which can be activated on widely available hardware (PC); for instance, the softwares developed for sea-going oceanographers who want to "see" their new data on screen without delay, and to compare them to historical records. Second, more sophisticated systems, based on advanced visualisation techniques, which are designed to explore the three-dimensional structure of the ocean's interior; in particular, the use of animation to explore 3D scenes is of great interest.