E-Locus - Institutional Repository of the University of Crete

Home Search

Results - Details

Search command : Author="Κοσσιώρης" And Author="Γεώργιος"

Current Record: 1 of 2

[Add to Basket]

Identifier

000402393

Title

Data assimilation methods

Alternative Title

Μέθοδοι αφομοίωσης δεδομένων

Author

Σφακιανάκη, Γεωργία

Thesis advisor

Πλεξουσάκης, Μιχαήλ

Reviewer

Μπάγκαβος, Δημήτριος
Κοσσιώρης, Γεώργιος

Abstract

Data Assimilation (DA) methods have been extensively used the past decades in many ﬁelds of science, among which meteorology, hydrology and oceanography, to mention just a few. Generally speaking, data assimilation is a technique to combine past knowledge of the system, in the form of a numerical model, and information about the system’s state, in the form of observations. Two main categories of DA methods can be recognized, variational and sequential. Variational methods are based on optimal control theory and the aim is to minimize a given cost function that measures the model-to-data misﬁt. In sequential methods, on the other hand, observations are assimilated as soon as they become available. In this thesis we study and present two methods from each family. The Three-Dimensional Variational assimilation (3D-Var) and the Four-Dimensional Variational assimilation methods (4D-Var) fall into the ﬁrst category. The most well-known DA method in the sequential family is the Kalman Filtering, and we focus on two of its variants, the Extended and the Ensemble Kalman Filtering. The methods presented in our work are viewed under the prism of numerical weather prediction. In this context, data assimilation is used to produce an analysis of the current state of the atmosphere to be used as initial conditions in a subsequent weather forecast, leading to more accurate predictions. We provide implementations and numerical results for the 3D-Var and Ensemble Kalman Filtering methods applied to the Lorenz-96 model. In the frame of weather forecasting, we also present the Weather Research and Forecasting (WRF) model which is a state-of-the-art atmospheric modeling system designed for numerical weather prediction and is currently being used in operational centers. We focus on the WRF Data Assimilation (WRFDA) module that includes implementations for the 3D-Var and 4DVar methods, as well as a hybrid scheme between the variational and ensemble assimilation methods.

Language

English

Subject

3Dvar

4Dvar

Kalman filter

Optimal interpolation

Issue date

2016-07-22

Collection

School/Department--School of Sciences and Engineering--Department of Mathematics and Applied Mathematics--Post-graduate theses

Type of Work--Post-graduate theses