Space weather has become an international issue due to the catastrophic impact it can have on modern societies. Solar flares are one of the major solar activities that drive space weather and yet their occurrence is not fully understood. Research is required to yield a better understanding of flare occurrence and enable the development of an accurate flare prediction system, which can warn industries most at risk to take preventative measures to mitigate or avoid the effects of space weather. This thesis introduces novel technologies developed by combining advances in statistical physics, image processing, machine learning, and feature selection algorithms, with advances in solar physics in order to extract valuable knowledge from historical solar data, related to active regions and flares. The aim of this thesis is to achieve the followings: i) The design of a new measurement, inspired by the physical Ising model, to estimate the magnetic complexity in active regions using solar images and an investigation of this measurement in relation to flare occurrence. The proposed name of the measurement is the Ising Magnetic Complexity (IMC). ii) Determination of the flare prediction capability of active region properties generated by the new active region detection system SMART (Solar Monitor Active Region Tracking) to enable the design of a new flare prediction system. iii) Determination of the active region properties that are most related to flare occurrence in order to enhance understanding of the underlying physics behind flare occurrence. The achieved results can be summarised as follows: i) The new active region measurement (IMC) appears to be related to flare occurrence and it has a potential use in predicting flare occurrence and location. ii) Combining machine learning with SMART's active region properties has the potential to provide more accurate flare predictions than the current flare prediction systems i.e. ASAP (Automated Solar Activity Prediction). iii) Reduced set of 6 active region properties seems to be the most significant properties related to flare occurrence and they can achieve similar degree of flare prediction accuracy as the full 21 SMART active region properties. The developed technologies and the findings achieved in this thesis will work as a corner stone to enhance the accuracy of flare prediction; develop efficient flare prediction systems; and enhance our understanding of flare occurrence. The algorithms, implementation, results, and future work are explained in this thesis.

Coronal Mass Ejections (CMEs) and solar flares are energetic events taking place at the Sun that can affect the space weather or the near-Earth environment by the release of vast quantities of electromagnetic radiation and charged particles. Solar active regions are the areas where most flares and CMEs originate. Studying the associations among sunspot groups, flares, filaments, and CMEs is helpful in understanding the possible cause and effect relationships between these events and features. Forecasting space weather in a timely manner is important for protecting technological systems and human life on earth and in space. The research presented in this thesis introduces novel, fully computerised, machine learning-based decision rules and models that can be used within a system design for automated space weather forecasting. The system design in this work consists of three stages: (1) designing computer tools to find the associations among sunspot groups, flares, filaments, and CMEs (2) applying machine learning algorithms to the associations' datasets and (3) studying the evolution patterns of sunspot groups using time-series methods. Machine learning algorithms are used to provide computerised learning rules and models that enable the system to provide automated prediction of CMEs, flares, and evolution patterns of sunspot groups. These numerical rules are extracted from the characteristics, associations, and time-series analysis of the available historical solar data. The training of machine learning algorithms is based on data sets created by investigating the associations among sunspots, filaments, flares, and CMEs. Evolution patterns of sunspot areas and McIntosh classifications are analysed using a statistical machine learning method, namely the Hidden Markov Model (HMM).

Made available in DSpace on 2016-03-15T19:35:55Z (GMT). No. of bitstreams: 1 Odilon Moura Guimaraes Junior.pdf: 3129591 bytes, checksum: ed51f5550fa4fa1c61e9e8289b9b9671 (MD5) Previous issue date: 2015-12-10 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / This work aims the study of the stochastic characteristics of solar flares with data in the sub millimeter wavelength range, in order to identify the flare evolving mechanisms. We shall apply the wavelet transformation, as a mathematical tool. The wavelet coefficients are squared and normalized by the squared wavelet coefficient mean to get specific indices known as Local Intermittence Measure (LIM). We generate 2D graphics with time and timescale axes to identify episodes that can represent intermittent behavior over the flare development, trying to find participative structures that could be associated to energy release models: either from small structures increasing in number and size resulting in a large explosive region, the so called Avalanche models; or on the other way, when a single energy release fragments into smaller and smaller regions, recognized as a Cascade . In both cases we try to verify whether the flare evolution follows a deterministic behavior rather than simply random episodes. / Este trabalho tem o propósito de estudar características aparentemente estocásticas das explosões solares, a partir de dados observacionais na faixa dos comprimentos de ondas submilimétricos procurando identificar a forma como elas evoluem. Aplicaremos para esse estudo a transformação wavelet como ferramenta matemática. Os coeficientes wavelet são elevados ao quadrado e normalizados para obter índices específicos denominados de Medida de Intermitência Local (LIM = Local Intermittence Measure). Estes geram gráficos em um plano com eixos temporal e escalar, onde podemos identificar episódios durante uma explosão que definitivamente representem comportamentos intermitentes permitindo visualizar ao longo do desenvolvimento da explosão os mecanismos de evolução das explosões solares: seja a partir de pequenas estruturas que aumentando em número e dimensões terminam formando uma grande região explosiva, seguindo um modelo do tipo Avalanche ; ou no caminho inverso, onde uma única explosão segue se fragmentando em regiões cada vez menores seguindo a tendência de um modelo do tipo Cascata . Em ambos os casos procuraremos notar se a evolução das explosões segue comportamentos determinísticos ao invés de simplesmente aleatórios.

Made available in DSpace on 2016-03-15T19:38:06Z (GMT). No. of bitstreams: 1 Tatiana Ferreira Raffaelli.pdf: 1372071 bytes, checksum: 274f2a97f290810c43d6e7c6e0730d1a (MD5) Previous issue date: 2007-09-18 / The existence of a highly reliable prediction system to detect the occurrence of large solar flares (class X) is still an unsolved problem. Despite many studies performed so far, no such a system has been found yet. In this work, we have developed a method using Bayesian Network - an Artificial Intelligence technique for the detection of giant solar flares. The Bayesian Networks software learned the relation among the variables that describe the sunspots within an active region and built a network with the relationships among them based on conditional probabilities. The studies were divided into two stages one to detect whether the sunspot would produce a big flare or not and another phase where some networks were built to discover the day the flare would occur. The first phase results were very satisfactory reaching a reliability of 77%. The second phase was more complex and the results were about 77% (with day constraints) and 54% (a wider range of days). / A existência de um sistema de previsão, de alta confiabilidade, para a detecção de ocorrência de grandes explosões solares (classe X) ainda é um problema sem solução. Existem diversos estudos nesta área, porém ainda não foi encontrado nenhum sistema eficiente. Para este trabalho foi desenvolvido um método utilizando-se redes Bayesianas, técnica de Inteligência Artificial, para a previsão das grandes flares (explosões) solares. O software de redes Bayesianas aprendeu a relação entre as variáveis que descrevem as regiões ativas e constroem uma rede com os relacionamentos entre elas baseados em probabilidades condicionais. Os estudos foram divididos em duas etapas, uma rede para detectar se a mancha solar irá produzir uma grande explosão ou não, e uma outra etapa em que foram construídas redes para prever o dia em que a explosão irá ocorrer. Os resultados obtidos na primeira etapa foram bem satisfatórios, atingindo 84% de confiabilidade. Já a segunda etapa do trabalho mostrou-se mais complexa e os resultados obtidos foram de 77% (com restrições de dias) e 54% (sem restrições de dia).

Les éruptions solaires sont des tempêtes de rayonnement électromagnétique, de particules relativistes et parfois de masse coronale provoquées par la libération d’énergie magnétique provenant de la couronne solaire. Si ces tempêtes atteignent l'environnement terrestre, elles peuvent poser un danger à la santé des astronautes en hautes orbites et causer des perturbations importantes sur les systèmes GPS. Dans certains cas, elles peuvent même induire des dommages aux infrastructures technologiques, dont les réseaux électriques. La prédiction des éruptions solaires est donc considérée comme un des plus importants défis de la météorologie spatiale. Par contre, à ce jour, aucune méthode présentée dans la littérature n’est capable de produire des prédictions fiables, ce qui met en évidence la nature complexe du déclenchement des éruptions solaires. Nous présentons donc dans ce mémoire une méthode alternative aux techniques statistiques habituelles, basée sur l'assimilation de données couplée avec des modèles rapides en automate cellulaire appelés modèles d'avalanche. Les modèles d'avalanche sont une simplification drastique de la physique du déclenchement des éruptions solaires. Malgré leur simplicité, ils reproduisent assez bien les statistiques à long terme de la taille des éruptions. Nous présentons dans ce mémoire des analyses empiriques de la capacité prédictive de quatre modèles: le modèle de Lu et Hamilton (LH) (Lu & Hamilton, 1991, ApJ, 412, 841-852), deux modèles à forçage déterministes (D) (Strugarek & Charbonneau, 2014, SoPh, 289(8), 2993-3015) et finalement deux modèles maximisant l'énergie libérée, appelées modèles DMC, qui sont fortement inspirés du modèle présenté par Farhang et al. (2018, ApJ, 859(1), 41). Les modèles DMC ont été développés dans le cadre de cette maîtrise et donc un chapitre de ce mémoire est dédié à leur présentation et aux analyses plus détaillées de leurs caractéristiques. Nous montrons que pour les modèles D ainsi que les modèles DMC, une mémoire existe entre les évènements simulés de grandes tailles, malgré la forte stochasticité de chacun de ces modèles. Nous présentons de plus dans ce mémoire un nouveau protocole de prédiction des éruptions solaires, utilisant l'assimilation de données couplée avec les modèles d'avalanches. Notre protocole se base sur une méthode de recuit simulé pour ajuster la condition initiale du modèle jusqu'à ce qu'elle reproduise de façon satisfaisante une série d'évènements observés. Une fois cette condition initiale optimisée produite, la simulation qui en résulte représente notre prédiction. Nous montrons dans ce mémoire le succès de notre protocole à bien assimiler une centaine d'observations synthétiques (produit par les modèles d'avalanche eux-mêmes). / Solar flares are sudden releases of electromagnetic radiation, relativistic particles and occasionally coronal mass, caused by the release of magnetic energy from the solar corona. They pose a danger to astronauts in high orbits and directly impact the Earth, including significant disturbances on GPS systems, and can even cause damage to technological infrastructures, including electrical networks. Predicting solar flares is therefore considered to be one of the most critical challenges in space weather. However, no method presented in the literature can produce reliable predictions, highlighting the complex nature of the triggering of solar flares. We, therefore, present in this thesis an alternative method to the usual statistical forecasting techniques. Our method is based on data assimilation coupled with computationally inexpensive cellular automaton models called avalanche models. Avalanche models are a drastic simplification of the physics underlying the triggering of solar flares. Despite their simplicity, they reproduce reasonably well the long-term statistics of solar flares sizes. In this thesis, we present empirical analyses of the predictive capabilities of four models: the Lu and Hamilton (LH) model (Lu & Hamilton, 1991, ApJ, 412, 841-852), two deterministic-driven (D) models (Strugarek & Charbonneau, 2014, SoPh, 289(8), 2993-3015) and finally two models using the principle of minimum energy during magnetic reconnection, called DMC models, which are strongly inspired by the models presented by Farhang et al. (2018, ApJ, 859(1), 41). The DMC models were developed during this project; therefore, a chapter of this thesis is dedicated to their presentation and more detailed analyses of their characteristics. We show that for D and DMC models, a memory exists between large simulated events, despite the high stochasticity present within each of these models. We finally present in this thesis a new protocol for predicting solar flares, using data assimilation coupled with avalanche models. Our protocol is based on a simulated annealing method to adjust the initial condition of the model until it satisfactorily reproduces a series of observed events. Once this optimal initial condition is found, the resulting simulation produces our prediction. In this thesis, we show our algorithm's success in assimilating hundreds of synthetic observations (produced by the avalanche models themselves).

Coronal Mass Ejections (CMEs) and solar flares are energetic events taking place at the Sun that can affect the space weather or the near-Earth environment by the release of vast quantities of electromagnetic radiation and charged particles. Solar active regions are the areas where most flares and CMEs originate. Studying the associations among sunspot groups, flares, filaments, and CMEs is helpful in understanding the possible cause and effect relationships between these events and features. Forecasting space weather in a timely manner is important for protecting technological systems and human life on earth and in space. The research presented in this thesis introduces novel, fully computerised, machine learning-based decision rules and models that can be used within a system design for automated space weather forecasting. The system design in this work consists of three stages: (1) designing computer tools to find the associations among sunspot groups, flares, filaments, and CMEs (2) applying machine learning algorithms to the associations¿ datasets and (3) studying the evolution patterns of sunspot groups using time-series methods. Machine learning algorithms are used to provide computerised learning rules and models that enable the system to provide automated prediction of CMEs, flares, and evolution patterns of sunspot groups. These numerical rules are extracted from the characteristics, associations, and time-series analysis of the available historical solar data. The training of machine learning algorithms is based on data sets created by investigating the associations among sunspots, filaments, flares, and CMEs. Evolution patterns of sunspot areas and McIntosh classifications are analysed using a statistical machine learning method, namely the Hidden Markov Model (HMM).

No / A crucial challenge to successful flare prediction is forecasting periods that transition between "flare-quiet" and "flare-active." Building on earlier studies in this series in which we describe the methodology, details, and results of flare forecasting comparison efforts, we focus here on patterns of forecast outcomes (success and failure) over multiday periods. A novel analysis is developed to evaluate forecasting success in the context of catching the first event of flare-active periods and, conversely, correctly predicting declining flare activity. We demonstrate these evaluation methods graphically and quantitatively as they provide both quick comparative evaluations and options for detailed analysis. For the testing interval 2016-2017, we determine the relative frequency distribution of two-day dichotomous forecast outcomes for three different event histories (i.e., event/event, no-event/event, and event/no-event) and use it to highlight performance differences between forecasting methods. A trend is identified across all forecasting methods that a high/low forecast probability on day 1 remains high/low on day 2, even though flaring activity is transitioning. For M-class and larger flares, we find that explicitly including persistence or prior flare history in computing forecasts helps to improve overall forecast performance. It is also found that using magnetic/modern data leads to improvement in catching the first-event/first-no-event transitions. Finally, 15% of major (i.e., M-class or above) flare days over the testing interval were effectively missed due to a lack of observations from instruments away from the Earth-Sun line.

Thèse numérisée par la Division de la gestion de documents et des archives de l'Université de Montréal.

Thèse numérisée par la Division de la gestion de documents et des archives de l'Université de Montréal

Made available in DSpace on 2016-03-15T19:38:53Z (GMT). No. of bitstreams: 1 ALBERTO FARES AKEL JUNIOR.pdf: 5112998 bytes, checksum: f18fc33d2f9508c3ec265c0efa016b43 (MD5) Previous issue date: 2015-08-21 / Fundação de Amparo a Pesquisa do Estado de São Paulo / We study the behavior of the Earth-ionosphere waveguide through the modeling of the propagation of very low frequency radio waves (VLF). We use the computational model LWPC (Long Wave Propagation Capability) to estimate changes in amplitude and phase of the VLF signals detected by the SAVNET network (South America VLF NETwork), and thus try to understand the behavior of the lower ionosphere under different ionization conditions. The research was divided into two parts. The first part investigates the behavior of the VLF signals in quiescent regimes of ionization. Amplitude and phase simulations for the were carried out, modifying adapting polynomials for the β and h parameters (or Wait s parameters) as a function of the zenithal angle. The second part of this research, uses these polynomials in the study of the lower ionosphere under transient ionization regimes in two distinct conditions: first during of solar flares and second during solar eclipse. For the simulations under solar flare conditions, we calculate the changes in β and ℎ′ parameters during the 25/03/2008 solar explosion. With these values, we calculate the electronic density profile through an exponential model and we find that the electronic density at 75 km is ∼ 104 cm−3, that is twenty times higher than during quiescent conditions. To evaluate our parameter estimates, we calculate the variation of the Wait s parameters for the case of twelve solar events of different classes. We note that the variations Δℎ′ found in this work are larger than that in Muraoka, Murata e Sato (1977) because they consider the variations in the conductivity gradient. For the solar eclipse simulations on 11/07/2011, we investigate its effect on the VLF phase. For this, we use the obscuration coefficient to estimate the guide height variation along the whole path during the eclipse. The simulations reproduce the phase behavior during the eclipse. However, a delay of about twenty four minutes was observed between the simulated and observed measurements. The observed delay is a direct consequence of own estimates of the perturbed ionospheric height and it causal relation with the obscuration during the eclipse. lower ionosphere, VLF, modeling, ionospheric disturbances, solar flares, solar eclipse. / Neste trabalho realizamos o estudo do comportamento do guia de ondas terra-ionosfera através da modelagem da propagação ondas de rádio de frequência muito baixa (VLF). Para isto, utilizamos o modelo computacional LWPC (Long Wave Propagation Capability) para estimar as variações de amplitude e fase de sinais de VLF detectados nos trajetos da rede SAVNET (South America VLF NETwork) e assim compreender o comportamento da baixa ionosfera em diferentes regimes de ionização. A pesquisa foi dividida em duas partes. A primeira parte, investigou o comportamento do sinal VLF em regimes quiescente de ionização, assim realizou-se simulações de amplitude e fase adaptando polinômios que definem os parâmetros β e ℎ′ (ou parâmetros de Wait) em função do ângulo zenital solar. Na segunda parte desta pesquisa, aplicou-se os polinômios no estudo da baixa ionosfera sob regimes transientes de ionização em duas condições distintas. A primeira para o caso de explosões solares e a segunda um para eclipse solar. Nas simulações relativas a explosões solares, calculamos as variações dos parâmetros β e ℎ′ durante o evento do dia 25/03/2008. Com esses valores, calculamos o perfil de densidade eletrônica, através de um modelo exponencial e observamos que a densidade eletrônica em 75 km é ∼ 104 cm−3, ou seja, vinte vezes maiores que antes da explosão. Para avaliar nossas estimativas, calculamos a variação dos parâmetros de Wait para doze eventos de diferentes classes. Observamos que as variações Δℎ′ neste trabalho são sempre maiores do que as descritas em Muraoka, Murata e Sato (1977), devido elas considerarem as variações no gradiente de condutividade. Nas simulações relativa ao eclipse solar do dia 11/07/2011, investigamos seu efeito na fase observada. Para esse estudo, utilizou-se o coeficiente de obscurecimento para realizar as simulações, desta forma foi possível estimar a variação da altura do guia ao longo de todo o trajeto durante o eclipse. As simulações reproduziram o comportamento da fase durante o eclipse. Entretanto, foi observado um atraso entre as medidas calculadas e observadas de aproximadamente ∼ vinte e quatro minutos. O atraso observado é diretamente decorrente da estimativa da altura de referência da ionosfera pertubada e de sua relação causal com o obscurecimento durante o eclipse.