Variational assimilation of tomography data in ocean models : twin experiments
Abstract
Acoustic tomography is a promising tool for monitoring the ocean interior. It provides integral measurements of temperature and velocity with high time resolution. The analysis of tomographic data using inversion show the ability to estimate the temporal evolution of heat content along the observed section. Tomography has been also used in global assimilation of different data sets to give a three dimensional estimate of the circulation. We propose to assess the impact of variational assimilation of those integral data in a realistic model. We used the adjoint technique to assimilate simulated observations, first it was done in quasi-geostrophic model, later in a primitive equation model. The goal of those experiments is to know what can be controled in the model by the tomographic data and at which resolution. The results clearly show the ability of tomographic constraint to recover temporal evolution of heat content in each layer and improve some component of the circulation described by the model.
1 Introduction
The range of scales resolved by the data is determined by the size of the cells defined by the observational sections. Only in the case of a mesoscale array, an estimation of the three dimensional circulation can be obtained by inversion of the data. When going toward larger scales, with coarser arrays, it is no longer possible to give a complete description of the circulation without using other sources of information. An appealing approach is to combine the observations and the knowledge contained in the equations of the numerical model which gives a continous representation of the ocean circulation. This can be done an assimilation method. As the final goal is to study the circulation consistent with all over the period where observations are available, we chose the variational assimilation approach. It gives at each instant an estimation of the state of ocean coherent with all past and future observations of the assimilation period. Here the question is what type of information is given by the tomographic constraint in term of model variables and how it improves the model solution. We focus on the case of a basin scale observational array, as the one which was deployed in the western Mediterranean sea for Thetis 2 experiment.
2 Ocean acoustic tomography
Acoustlc tomography is a technique to explore the ocean interior. The variations of sound speed in sea water are mainly due to the variations of temperature. For each acoustic ray from the source to the receiver, a travel time τ0 is calculated with a mean sound speed profile C0(z). The time difference between τ0 and the measured travel time is due to the variations of sea water properties and the
