DINEOF univariate reconstruction of missing satellite data from the North Sea Belcolour-1 database.

Poster (2008)

The Belcolour-1 database holds more than 4 years of uniformly resampled MERIS chlorophyll (CHL), total suspended matter (TSM), MODIS-AQUA CHL, TSM and sea surface temperature (SST) over the North Sea. A ... [more ▼]

The Belcolour-1 database holds more than 4 years of uniformly resampled MERIS chlorophyll (CHL), total suspended matter (TSM), MODIS-AQUA CHL, TSM and sea surface temperature (SST) over the North Sea. A first step of the RECOLOUR* project consists in the univariate reconstruction of missing data with the DINEOF method (Data Interpolating Empirical Orthogonal Functions). In particular, the DINEOF treatment of MERIS CHL and TSM images available for the year 2003 allowed an efficient synthesis of the coherent modes of variability existing at the scale of the whole North Sea. For both parameters, 4 modes were retained by general cross validation as an optimum for the reconstruction of missing data. For CHL, the first spatial mode shows the high influence of coastal nutrients outputs (mainly continental estuaries and diffused coastal sources) and the lower concentration in the well stratified central and northern part of the North Sea compared to the southern bight and the eastern English Channel. The spatial trends described by the first mode are permanent features during the year, although slightly enhanced during the summer and reduced during winter. The second spatial mode correspond to the main algal blooming events (spring and autumn) with increased concentrations in the whole southern bight of the north sea, around the Isle of Wight and in frontal alike structure north-west from Denmark. The third Eofs describes well the dynamics of an early phytoplankton bloom occurring in march along the Norwegian coast, where a strong stratification induced by an output of cold water from Baltic Sea provides good light conditions to phytoplankton. Concerning TSM, the first spatial mode shows the dominant influence of large estuaries and of resuspension from shallow coastal sedimental plains. The patterns suggest a general transport of sediments from south-east England up to the northern Dutch coastal waters, as a clear distinction between the stratified northern part and the well mixed and charged southern and German bights. Although these trends are permanent during the year, the range of the spatial variations are slightly reduced during the summer, following the reduction of resuspension, of total sediment outputed by rivers and of advection along continental coasts. The second mode shows a clear seasonal signal. The winter influence of the second spatial mode can be understood as general sediment enrichment due to higher resuspension, but a clear influence of intense winter terrestrial water outflows leading to lower sediment concentration in the plumes then in the surrounding waters. This is clear for the Elbe river discharge, the whole natural part of the Wadden Sea and the Seine river plume. The Scheelde and Thames rather seems to be just neutralizing the seasonal TSM resuspension signal. The Rhine river discharge seems to make exception as no influence is detected in the second spatial mode. During summer, the contribution of the second EOF is reversed with a general reduction of suspended matter concentration in most part of the area but some local sediment enrichment at specific river discharges. Original MERIS CHL and TSM images were filled and reconstructions were produced at a daily interval based on a linear interpolation of the temporal modes. From this, weekly averages could be calculated at stations such as the turbidity maximum of the Scheelde river plume, showing the onset of the spring bloom co-occurring with a period characterised both by the TSM seasonal reduction and by important TSM temporal variability. [less ▲]

The western mediterranean sea surface temperature dynamics seen through complementary in situ, satellite and modelling approaches over the 1985-1995 period

Conference (2008)

Data Interpolating Empirical Orthogonal Functions (DINEOF), a tool for geophysical data analyses

Conference (2008)

Mathematical and numerical geohydrodynamic models

in Nihoul, Jacques; Chen, Chen-Tung Arthur (Eds.) Encyclopedia of life support systems (EOLSS) (2008)

Three-dimensional analysis of oceanographic data with the software DIVA

in Geophysical Research Abstracts (2008)

In oceanography, the process of gridding data is frequently used for various purposes, e.g. initialization of hydrodynamic models, or graphical representation of sparse data. DIVA (Data-Interpolating ... [more ▼]

In oceanography, the process of gridding data is frequently used for various purposes, e.g. initialization of hydrodynamic models, or graphical representation of sparse data. DIVA (Data-Interpolating Variational Analysis) is designed to perform such gridding tasks. It has the advantage of taking into account the intrinsic nature of oceanographic data, i.e. uncertainty in in situ measurements and anisotropy due to advection and irregular coastlines and topography. Three-dimensional reconstruction of temperature and salinity fields is achieved by stacking horizontal layers where independent analysis with DIVA are performed. Nevertheless, analysis in regions void of data may result in the presence of static instabilities between two or more consecutive layers. The method implemented in DIVA to remove such kinds of instabilities is the object of the present work. It consists of adding pseudo-data from one layer to the upper adjacent layer in order to create stable stratification in the vicinity of instabilities. Two approaches for assigning values to the pseudo data are tested: the first is called the mixing approach and aims at simulating a mixing process between two layers; the second is called the minimal perturbation, as it strives to minimise the perturbations inthe pseudo-data. A realistic application using temperature and salinity profiles in the North Atlantic is carried out and the results are compared with World Ocean Atlas climatologies. [less ▲]

DIVA: a Data Analysis Software with Generalized-Cross Validation and Quality Control

Poster (2007, May 19)

Forecast verification of a 3D model of the Mediterranean Sea. The use of discrete wavelet transforms and EOFs in the skill assessment of spatial forecasts

in Journal of Marine Systems (2007), 65(1-4), 460-483

The quality assessment of a nested model system of the Mediterranean Sea is realised. The model has two zooms in the Provencal Basin and in the Ligurian Sea, realised with a two-way nesting approach. The ... [more ▼]

The quality assessment of a nested model system of the Mediterranean Sea is realised. The model has two zooms in the Provencal Basin and in the Ligurian Sea, realised with a two-way nesting approach. The experiment lasts for nine weeks, and at each week sea surface temperature (SST) and sea level anomaly are assimilated. The quality assessment of the surface temperature is done in a spatio-temporal approach, to take into account the high complexity of the SST distribution. We focus on the multi-scale nature of oceanic processes using two powerful tools for spatio-temporal analysis, wavelets and Empirical Orthogonal Functions (EOFs). We apply two-dimensional wavelets to decompose the high-resolution model and observed SST into different spatial scales. The Ligurian Sea model results are compared to observations at each of those spatial scales, with special attention on how the assimilation affects the model behaviour. We also use EOFs to assess the similarities between the Mediterranean Sea model and the observed SST. The results show that the assimilation mainly affects the model large-scale features, whereas the small scales show little or no improvement and sometimes, even a decrease in their skill. The multiresolution analysis reveals the connection between large- and small-scale errors, and how the choice of the maximum correlation length of the assimilation scheme affects the distribution of the model error among the different spatial scales. (c) 2006 Elsevier B.V. All rights reserved. [less ▲]

Multivariate reconstruction of missing data in sea surface temperature, chlorophyll, and wind satellite fields

in Journal of Geophysical Research. Oceans (2007), 112(C3), 03008

An empirical orthogonal function–based technique called Data Interpolating Empirical Orthogonal Functions (DINEOF) is used in a multivariate approach to reconstruct missing data. Sea surface temperature ... [more ▼]

An empirical orthogonal function–based technique called Data Interpolating Empirical Orthogonal Functions (DINEOF) is used in a multivariate approach to reconstruct missing data. Sea surface temperature (SST), chlorophyll a concentration, and QuikSCAT winds are used to assess the benefit of a multivariate reconstruction. In particular, the combination of SST plus chlorophyll, SST plus lagged SST plus chlorophyll, and SST plus lagged winds have been studied. To assess the quality of the reconstructions, the reconstructed SST and winds have been compared to in situ data. The combination of SST plus chlorophyll, as well as SST plus lagged SST plus chlorophyll, significantly improves the results obtained by the reconstruction of SST alone. All the experiments correctly represent the SST, and an upwelling/downwelling event in the West Florida Shelf reproduced by the reconstructed data is studied. [less ▲]

Filtering inertia-gravity waves from the initial conditions of the linear shallow water equations

in Ocean Modelling (2007), 19(3-4), 204-218

A method for filtering inertia-gravity waves from elevation and depth-averaged velocity is described. This filtering scheme is derived from the linear shallow water equations for constant depth and ... [more ▼]

A method for filtering inertia-gravity waves from elevation and depth-averaged velocity is described. This filtering scheme is derived from the linear shallow water equations for constant depth and constant Coriolis frequency. The filtered solution is obtained by retaining only the eigenvectors corresponding to the geostrophic equilibrium and by discarding explicitly the eigenvectors corresponding to the fast moving inertia-gravity waves. An alternative formulation is derived using a variational approach. Both filtering methods are tested numerically for a periodic domain with constant depth and the variational approach is implemented for a closed domain with large topographic variations. The filtering methods significantly reduce the amplitudes of the inertia-gravity waves while preserving the mean flow. The variational method is compared to the Incremental Analysis Update technique and the benefits of the variational filter are presented. (C) 2007 Elsevier Ltd. All rights reserved. [less ▲]

Study of the combined effects of data assimilation and grid nesting in ocean models – application to the Gulf of Lions

in Ocean Science (2006), 2

Modern operational ocean forecasting systems routinely use data assimilation techniques in order to take observations into account in the hydrodynamic model. Moreover, as end users require higher and ... [more ▼]

Modern operational ocean forecasting systems routinely use data assimilation techniques in order to take observations into account in the hydrodynamic model. Moreover, as end users require higher and higher resolution predictions, especially in coastal zones, it is now common to run nested models, where the coastal model gets its open-sea boundary conditions from a low-resolution global model. This configuration is used in the "Mediterranean Forecasting System: Towards environmental predictions" (MFSTEP) project. A global model covering the whole Mediterranean Sea is run weekly, performing 1 week of hindcast and a 10-day forecast. Regional models, using different codes and covering different areas, then use this forecast to implement boundary conditions. Local models in turn use the regional model forecasts for their own boundary conditions. This nested system has proven to be a viable and efficient system to achieve high-resolution weekly forecasts. However, when observations are available in some coastal zone, it remains unclear whether it is better to assimilate them in the global or local model. We perform twin experiments and assimilate observations in the global or in the local model, or in both of them together. We show that, when interested in the local models forecast and provided the global model fields are approximately correct, the best results are obtained when assimilating observations in the local model. [less ▲]