Internal waves (IWs) owe their existence to the stratification in the medium. These waves affect acoustic transmission greatly. Impact of these waves on acoustic transmission in deep water is fairly well understood due to better performance of well-celebrated Garrett-Munk (GM) model. However, in shallow waters, predicting these waves is not as easy, because of interactions with the bottom and surface. Hence two experiments, one during October 2002 and the other during October 2004 were conducted to characterize IWs in the shallow waters of northeastern Arabian Sea. The first experiment was carried out during October 2002 south of Gulf of Kutch (GOK) and the second experiment during October 2004 both south and north of GOK. During these experiments CTD moorings were deployed and temperature and salinity (TS) data were collected at 5 seconds interval. CTD Yo-Yo collected vertical profiles of TS at a sampling interval of 2.5 minutes for 3.5 hours during October 2002 and 1 hour during October 2004 experiment. In addition, during the first experiment, currents were measured using a vessel mounted Acoustic Doppler Current Profiler (ADCP), and in both experiments CTD TS profiles were taken from the ship. This data set has been used for characterizing internal waves in the northeastern Arabian Sea. Experiment conducted during October 2002, south of GOK has revealed large tidal ranges. The barotropic tidal range at the experimental site was 1.5m. Current observations made using the vessel mounted ADCP, along the shore and across the shore, showed signs of first mode (baroclinic) oscillations; currents in the top and bottom layers were in opposing directions. They were found to be southwesterly in the top layer and northeasterly in the bottom layer. Time - depth sections of TS profiles from CTD yo-yo data, revealed the presence of high frequency internal waves and solitons overriding on low frequency trend. Moored CTD time series of temperature records showed the presence of internal solitons, which caused a vertical displacement of about 8m in the isotherms, which is equivalent to 3OC change in temperature, in less than 10 minutes. Passage of internal solitons induced vertical mixing causing the mixed layer to deepen by about 10m and current speed increased by about 0.1 m/s. Internal solitons were traveling towards northwest and current vectors suggest that they were generated when the internal tide is reflected from the bottom. Vertical displacement spectra agreed well with GM spectra when solitons were not present. However, when the solitons were present the displacement spectra had higher energy levels compared to the GM spectra. Another experiment was done in October 2004, mainly aimed at characterizing internal solitons and to verify the consistency of the results obtained during October 2002 experiment. This experiment also showed that IWs of both high and low frequency along with internal solitons were present at the experimental site. It was found that internal solitons were more energetic during spring tide than the neap. The observed amplitudes of these solitons were around 12m and were not rank ordered suggesting that the experimental site is close to the generation point. It is believed that, generally, solitons get phase locked to the barotropic tide’s trough and travel. Such phase locking was not observed at the experimental site. They were observed riding on both troughs and crests of barotropic tide. One of the aims of this thesis is to develop a simulation model based on Garrett-Munk steady state internal wave spectrum. Hence, an internal wave model IWAVE was developed to simulate the sound speed structure due to internal waves. Sound speed structure is simulated instead of TS structure, because of their direct utility in sonar range prediction models. Since the GM model is a deep-water and mid-latitude model, it was calibrated to suite shallow-water tropical environment by incorporating the site and region specific parameters. EOFs and Dynamical modes estimated using TS profiles were used to identify the site-specific parameters of the GM model. Values for characteristic mode number and spectral slope used in the GM model are 3 and 2 respectively. However, it was found that they are different in the northeastern Arabian Sea. At this site, the characteristic mode number was found to be 1 and the spectral slope was found to be 3. The modified model was validated against the measured sound speed profiles. In the first case, the first sound speed profile (TS) of the CTD yo-yo data (20 October 2002) was used for predicting the remaining profiles and compared them with observations. This was done to verify the model’s ability to predict high frequency case (TS profiles are measured at every 2.5min.). In the second case, during October 2004, TS profiles collected at every one-hour for 24 hours were used. This gives an idea of the model’s performance for the low frequency case. The variances of the measured and simulated sound speed profiles matched well in both cases with the modified GM model.

Les modèles numériques d'océans régionaux tridimensionnels sont basés sur la résolution des équations primitives et utilisent pour la plupart des méthodes de résolution eulérienne de type différences finies sur des grilles décalées. Ces modèles doivent représenter fidèlement les transports et transferts d'énergie. L'amélioration de ces modèles numériques exige donc (i) l'identification des processus prépondérants, notamment en terme de dissipation, dans ces transferts et (ii) la construction de méthodes numériques respectant un certain nombre d'équilibres. La première partie du travail se concentre sur la propagation des ondes externes et internes de gravité. Nous nous intéresserons en premier lieu à la stabilité de la séparation en mode rapide (barotrope) et lents (baroclines) et montrons qu'elle peut être ameliorée en levant certaines hypothèses traditionnellement effectuées. Dans un second temps, nous étudions l'impact de la discrétisation (ordre des schémas, grilles décalées ou non) sur la propagation des ondes internes de gravité provenant du couplage vitesse pression. Une décomposition en modes verticaux nous permet également de proposer un schéma espace temps très efficace. La seconde partie étudie en détail les schémas d'advection de quantité de mouvement et de traceurs, tout particulièrement dans l'objectif d'une réduction de la diffusion diapycnale (diffusion dans les directions orthogonales aux couches de densité constante). Ce travail nous amène tout d'abord à porter notre attention sur les schémas d'advection verticaux souvent négligés au regard de la dimension horizontale. Les bonnes propriétés d'un schéma compact (et de ses variantes espace temps et monotones) sont mises en avant. Enfin nous analysons le comportement multidimensionnel de ces schémas d'advection. / Three-dimensional regional ocean numerical models are based on solving the primitive equations and mostly use Eulerian finite differences methods of resolution on staggered grids. These models must accurately represent transports and energy transfers. Improving these numerical models therefore requires (i) the identification of predominant process, particularly in terms of dissipation in these transfers and (ii) the construction of numerical methods respecting a number of balances. The first part of the work focuses on the propagation of external and internal gravity waves. We focus primarily on the stability of the separation in fast mode (barotropic) and slow (baroclinic) and show that it can be improved by removing certain assumptions traditionally made. In a second step, we study the impact of the discretization (order of schemes, staggered grids or not) on the propagation of internal gravity waves coming from the coupling velocity pressure. A decomposition into vertical modes also allows us to offer a highly effective space-time scheme. The second part examines in detail the numerical advection schemes of momentum and tracers, especially with the aim of reducing the diapycnal diffusion (diffusion in the orthogonal direction of constant density layers). This work leads us first to focus our attention on the vertical advection schemes often overlooked in front of the horizontal dimension. The good properties of a compact schema (and its space-time and monotonous variants ) are highlighted. Finally we analyze the multidimensional behavior of these advection schemes.