Hunting elusive excess variance in big LOFAR data
| PhD ceremony: | Ms H.Y. (Hyoyin) Gan |
| When: | October 06, 2022 |
| Start: | 09:00 |
| Supervisor: | prof. dr. L.V.E. (Léon) Koopmans |
| Co-supervisor: | M.H.F. (Michael (he/him)) Wilkinson, Dr |
| Where: | Academy building RUG |
| Faculty: | Science and Engineering |
The Epoch of Reionisation (EoR) is a watershed period of the universe when the predominantly neutral intergalactic medium was ionised and the first luminous sources formed. LOFAR (Low Frequency Array) is a radio interferometer which can detect the 21-cm signal from the EoR. The detection is challenging due to the strong astrophysical foregrounds, radio frequency interference, ionospheric effects and instrumental effects. Even after calibration, the remaining residuals are above the estimated thermal noise, known as the "excess variance".
My thesis is dedicated to studying complex correlations between excess variance and its sources. In Chapter 2, I found that the excess variance has a Local Sidereal Time dependence related to distant and bright sources in the sky such as Cassiopeia A and Cygnus A. In Chapter 3, I compared the performance of a new direction-dependent calibration method, DDECAL, to our current method, SAGECAL on an unexplored field around our target field, the North Celestial Pole. Similar imprints from Cassiopeia A and Cygnus A are shown in this analysis as well. To further identify the contribution of bright sources in sky images more efficiently, I introduce a new data analysis tool, Self-Organising Attribute Maps. This method explores clusters in vector attributes of a component tree, the max-tree, with an unsupervised machine learning technique, self-organising maps (SOMs). The applications on medical and LOFAR sky images show that this method is promising for exploring morphological features in images without manually thresholding vector attributes.