Oceanography is a branch of Physics that studies the evolution of the oceans
of the Earth. Since the Earth environment is a complex and extended system, to accurately estimate its status and its evolution, many observational systems that measure the physical parameters on a global scale are needed. In this context, the European Union started an advanced program that aims to monitor and forecast the environmental status of the Earth: the Copernicus program (http://www.copernicus.eu). The work of this thesis falls under this framework, in particular the Data Assimilation technique for the marine biogeochemistry forecasts.
Data Assimilation (DA) is a technique widely used in computational sciences to incorporate observational data into a prediction model. A large use of DA algorithms is done, for example, in weather forecasting simulations and physical oceanography. Recently, DA has been applied also in biogeochemical ocean modeling, with additional issues than in other geophysical disciplines due to the complexity of biogeochemical models and the poor availability of large data sets.
The code examined in this thesis implements a three-dimensional variational DA scheme (3DVar) for the assimilation of satellite observations in a biogeochemical model (further details will be found in Teruzzi et al. (2014)).
The model is a medium complexity biogeochemical model (OGSTM-BFM) which describes the plankton bacteria and nutrient dynamics in the Mediterranean Sea (further details will be found in Lazzari et al. (2010), Cossarini, Lazzari & Solidoro (2015)). The code is called 3DVarBio and is used to constrain the three-dimensional fields of the four phytoplankton functional types in the OGSTM-BFM. The main purpose of this thesis is to parallelize and optimize the DA scheme within the 3DVarBio code.
Thesis not available.