Astroinformatics

Magnetism is ubiquituous in the universe - planets, stars, galaxies and even the intergalactic medium are all magnetized. The modern society is strongly affected by the magnetic activity manifestations of our star, the Sun. The space missions, air traffic, power grids and telecommunication networks are vulnerable to disturbances caused by bad space weather, i.e. increased solar magnetic activity. The magnetized objects share a few common features: at least they are rotating and the matter in which the magnetic fields are generated (interior of the Earth, solar convection zone, interstellar matter) is turbulent. Understanding the generation of magnetic fields, therefore, is intrinsically linked to studying turbulent fluids under extreme conditions. Such studies are out of the scope of analytic methods and laboratory experiments, the only viable tool being numerical modelling. Even that approach is extremely challenging: to be able to understand, explain, and predict the behavior of the global-scale magnetic field of the Sun, one needs to build a numerical model of the fine details of the motions within the star. This can be achieved only through by developing the fastest and the most accurate possible algorithms that are suitable for massively parallel computation, and continuously improving them to take advantage of the new emerging ICT technologies (such as cloud and GPU computing).

As a result of the modelling, huge datasets with three-dimensional spatial information of the basic physical quantities (such as velocity, magnetic field, density, temperature) are obtained. As the magnetohydrodynamic equations are integrated over time, a statistically steady state is usually obtained, but the solutions most often show systematic behavior (large-scale circulation, magnetic cycles) making it necessary to study also the time dimension of the data. In this way, even the moderated-sized simulations create big data when all the information is build into a spatiotemporal stack. With increasing amount of data, the development of efficient and computationally economical data analysis methods is an urgent requirement in our research group.

The lines of study described above constitute a major part of a discipline called computational astrophysics. In the year 2012, computational astrophysics user group, of which roughly half of the projects are lead by the CMDAA group researchers, consumed roughly 250 000 core hours per user of the national computing centre CSC resources. This is the top number amongst all the disciplines, and reflects the large volume of computational resources required for the execution of the models.

Systematic observations of the Sun have been carried out for over four hundred years, during the last century with increasing amount of space-borne machinery. Complemented with the data obtained for other solar-type stars coming from many observational infrastructures scattered on the best observing sites over the globe and some satellites, the amount of data is accumulating with increasing speed also in the observational frontier. Investigation of these objects is important, as this sheds some light into the history of the Sun, as most of the targets observable to Earthly astronomers are rapid rotators, resembling the Sun at an earlier age. The most common method to search for systematic cyclic behavior (as observed on the Sun), indicative of a magnetic activity cycle, is to use time-series analysis methods on stellar photometric data. For astronomical observations in the optical range of wavelengths made from the ground, the observations are inevitably interrupted during daytime and due to bad weather. In astronomical community, therefore, a special class of time series analysis methods capable of effective functioning even in the presence of long time intervals of missing data (called as gaps) has been developed and are extensively applied in the stellar community.

One of the major goals of our research group is to achieve a better understanding of solar activity. A complementary path to the intensive numerical modelling is provided by the time series analysis of solar activity tracers utilizing the vast amount of spatiotemporal data that is continuously collected by satellite and ground-based observations. Although our modelling efforts indicate that predictability of the future solar magnetic activity, in the generation process of which the stochastic turbulent effects play a key role, is low, it is very important to study and try to determine the key factors leading to long-term increased/decreased magnetic activity level of the Sun. Such epochs are known as grand minima (such as the Maunder minimum) and maxima (such as the present Grand Modern Maximum that seems to have come to its end).

The Sun is the only stellar object the surface of which we can directly observe - the other stars, even the closest and biggest ones, are seen as point sources even with the most powerful telescopes. As mentioned above, one can always study the lightcurve variations caused for instance by large starspots from photometric datasets, but then one recovers quite unreliable information on the actual latitudinal location of the spot. Is it possible at all to "see" the surfaces of other stars?

Almost magically, spectroscopy and spectropolarimetry offer an indirect way of mapping the surfaces of stars - the starspots make a finger print (in practise a bump) in the absorption lines and moreover the structures being magnetized causes a Zeeman effect in the polarized spectra. Due to these effects, systematic consequtive observations of the stars can be transformed, with the aid of sophisticated inversion methods, into surface temperature and magnetic field maps of the objects. Almost 20-year long time-series of spectroscopic and ten-year long time-series of spectropolarimetric observations of rapidly rotating late-type stars has been collected with the SOFIN-spectrograph at the Nordic Optical telescope, La Palma, Spain. The NOT data set is complemented with the observing programme carried out at the ESO La Silla 3.6m telescope with HARPSPol spectropolarimeter. Rapidly spinning young Suns have revealed to exhibit increased magnetic activity level, large high-latitude spots with highly non-axisymmetric configuration, and time evolution characterized by an azimuthally directed dynamo wave and sudden jumps of the activity level from one active longitude to the other. This is to be contrasted with the present-day Sun with small spots occurring at lower latitudes migrating in the latitudinal direction.

The group members participate in the development of the Pencil code, a high-order finite-difference solver for compressible hydrodynamic flows with magnetic fields and particles. It is highly modular and can easily be adapted to different types of problems. The code runs efficiently under MPI on massively parallel shared- or distributed-memory computers. The code is freely available under the Gnu General Public License, and new users/developers are welcomed to join the effort.

Pencil code

ISDA, Irregularly Spaced Data Analysis, is a software package designed to perform various tasks on irregularly spaced time series. ISDA can help you to analyse data sets from such diverse fields of application as astronomy, geophysics, biophysics, business and so forth. Astronomers, amateurs as well as professionals, form the largest group of current/prospective users. ISDA helps by automating routine tasks and by providing new and effective methods of analysis. Some algorithms that are used in the system were developed in Tartu University and in Tartu Observatory (Estonia) by Dr. Jaan Pelt and his group, close collaborators of the CMDAA group. The CMDAA group uses the ISDA package as a standard tool for the analysis of solar and stellar datasets, ranging from sunspot numbers and locations to stellar photometric time series with gaps due to observational constraints.

ISDA

Astaroth is code developed within the group to test the speed up the execution of finite-difference type solvers using Graphical Processing Units (GPUs).