Defence in the field of computer science, Nigul Olspert, M.Sc.
Nigul Olspert, M.Sc., will defend the dissertation "From periodic to cyclic processes in stellar magnetic activity research: time series analysis methods and their applications" on Friday 16 November at 12 noon at the Aalto University School of Science, lecture hall AS1, Maarintie 8, Espoo.
In this dissertation we introduce and develop time series analysis methods which are dedicated to period and cycle length estimation of magnetically active stars. Knowing both of these quantities is important as it makes observable reality comparable with predictions from the dynamo theory. Magnetic activity is primarily manifested through dark spots on the surface of the star. However, the rotation of the star is not uniform, but differential, which makes the period estimation not a trivial task. Furthermore, over time the number of spots changes, forming a cyclic process. For instance, for the Sun the activity cycle is known to last approximately 11 years, while both the length of each individual cycle as well as the amplitude is varying. Such a behaviour is called quasi-periodic.
In the introduction of the dissertation we give an overview of the relevant questions in the domain of period estimation for unevenly sampled time series and subsequently introduce several practical methods applicable to quasi-periodic as well as nonstationary time series. In the applications we have used datasets of several Sun-like stars, for which we have estimated the mean rotation period, magnetic cycle length, made comparison to earlier studies and discussed the results in the light of dynamo theory. One of the interesting and yet not understood finding from the study is the pattern how the stars group when plotted on a so-called activity diagram.
Besides real world data we have analysed the data from fully 3D global magnetohydrodynamical simulations, which have only quite recently become possible to carry out. The main challenge in the latter analysis is the vast amount of multidimensional data, thus the algorithms used must be well scalable. For the computation we have used parallelisation and the help of supercomputers.
Vastaväittäjä: Professor Ivan Andronov, Odessa National Maritime University, Ukraine
Kustos: Professor Aki Vehtari, Aalto University School of Science, Department of Computer Science