Titulo Estágio
Machine learning for forecasting geomagnetic activity at COI
Local do Estágio
Observatório Geofísico e Astronómico da Universiade de Coimbra
Enquadramento
Geomagnetic activity mainly results from the interaction between plasma arriving from the Sun and our planet's magnetosphere and ionosphere. Major geomagnetic disturbances (also called storms) driven by the solar wind and coronal mass ejections can last up to several days. These storms pose a significant threat to power distribution and global navigation systems and can cause temporary disruptions. The nine hours Hydro-Québec blackout on March 1989 is one of the most famous examples. Recently, a dedicated field of study emerged called Space Weather (SW), in an attempt to understand, predict and mitigate the impact of these events (e.g. project MAG-GIC: www.uc.pt/en/org/maggic).
The physical forward modelling of geomagnetic storms is formidable, with the need to take into account the plasma magnetohydrodynamics of the solar-interplanetary-magnetosphere system. For this reason, as a first approach to geomagnetic storm modelling, empirical models that relate geomagnetic storm series to solar wind parameters have been used. This is the case of Tsyganenko & Sitnov, 2005 model (e.g. Castillo et al., 2017). Another possibility is the use of machine learning techniques such as Artificial Neural Networks (e.g. the seminal article Gleisner & Lundsted, 2001) or other state of the art methodologies (e.g. the review article Camporeale, 2019).
It is believed that the basic mechanism of formation of storms impacting mid-latitude observatories is the strengthening of Earth's ring current in response to changing solar wind conditions. The solar wind parameters most important for strengthening the ring current are the southward component of the inter-planetary magnetic field (IMF), velocity, and plasma density. At the OMNI database at GSFC/SPDF OMNIWeb Interface (https://omniweb.gsfc.nasa.gov), there is an abundance of high-quality data for these parameters obtained from high-altitude satellites.
Simultaneously, at the University of Coimbra there is a magnetic observatory (COI) that is part of the Geophysical and Astronomical Observatory of the University of Coimbra (OGAUC), providing for continuous local measurements of the geomagnetic field. This is one of the oldest geomagnetic observatories in the world still in operation, with a dataset going back to 1866, and the only one in the Portuguese mainland. At the time geomagnetic storms occur, they are registered by OGAUC magnetometers.
The development of computational tools for the forecast of SW related phenomena is the goal of a collaboration between researchers from CITUEC and Centre for Mathematics of the University of Coimbra.
References:
Castillo, Y. Pais, M.A.; Fernandes, J., et al., 2017, Geomagnetic activity at Northern Hemisphere's mid-latitude ground stations: How much can be explained using TS05 model, Journal of Atmospheric and Solar-Terrestrial Physics, Vol.165-166, pp. 38-53
Gleisner, H., Lundstedt, H., 2001; A neural network-based local model for prediction of geomagnetic disturbances Journal of Geophysical Research Atmospheres, 106(A5):8425-8434
Tsyganenko, N.A., Sitnov, M.I.: 2005, Modeling the dynamics of the inner magnetosphere during strong geomagnetic storms. J. Geophys. Res. 110, A03208.
Camporeale, E.: 2019, The challenge of machine learning in Space Weather: Nowcasting and forecasting. Space Weather, 17, 1166–1207.
Objetivo
The student will learn the basics of geomagnetism and of the geomagnetic activity's driving mechanisms. At the end of the project the student will have created computational tools for the nowcast/forecast of local geomagnetic activity at Coimbra, including geomagnetic series during storms and local geomagnetic activity indices. These will be compared with COI station observations.
Plano de Trabalhos - Semestre 1
In the beginning of the project the student will carry out a bibliographic research in order to become acquainted with the problem in hand. In this manner, the student will learn the basics on solar activity, solar wind emission and propagation and of the Sun-Earth interaction. Additionally, the student will become familiar with OMNI and OGAUC databases and their interest for SW studies, performing an exploratory analysis of the data of interest for geomagnetic studies (e.g. checking for possible data correlations that will be useful for the second half of the project).
At the end of the semester the student will perform a short presentation to the members of the CITEUC and CMUC Space Weather collaboration.
Plano de Trabalhos - Semestre 2
During the second semester, the student will employ recent developments on deep learning techniques for the forecasting of geomagnetic indices. In particular, the research work will focus on the development of convolutional neural networks that take as input data from aforementioned databases to predict the geomagnetic activity.
The last two months of the semester will be dedicated to the writing of the dissertation. Culminating in a short presentation of the work performed and of the results obtained during this project to the members of the CITEUC and CMUC collaboration.
Condições
The work will be carried out at Centro de Investigação da Terra e do Espaço da Universidade de Coimbra, hosted at the Geophysical and Astronomical Observatory of the University of Coimbra. CITEUC has been the host institution of several MSc and PhD projects, including in the Space Weather field. At CITEUC the student will have at his disposal all the tools (hardware and software) required.
This project counts with CITEUC's expertise in Geomagnetism and Space Weather and CMUC's expertise in Computational Mathematics.
Observações
This project will be co-supervised by Prof. Sílvia Barbeiro from FCTUC's Mathematics Department and CMUC (Centre for Mathematics of the University of Coimbra).
Orientador
Fernando Jorge Gutiérrez Pinheiro
fjgpinheiro.astro@gmail.com 📩