Tufan Colak

Machine-learning algorithms are applied to explore the relation between significant flares and their associated CMEs. The NGDC flares catalogue and the SOHO/LASCO CME catalogue are processed to associate X and M-class flares with CMEs based on timing information. Automated systems are created to process and associate years of flare and CME data, which are later arranged in numerical-training vectors and fed to machine-learning algorithms to extract the embedded knowledge and provide learning rules that can be used for the automated prediction of CMEs. Properties representing the intensity, flare duration, and duration of decline and duration of growth are extracted from all the associated (A) and not-associated (NA) flares and converted to a numerical format that is suitable for machine-learning use. The machine-learning algorithms Cascade Correlation Neural Networks (CCNN) and Support Vector Machines (SVM) are used and compared in our work. The machine-learning systems predict, from the input of a flare’s properties, if the flare is likely to initiate a CME. Intensive experiments using Jack-knife techniques are carried out and the relationships between flare properties and CMEs are investigated using the results. The predictive performance of SVM and CCNN is analysed and recommendations for enhancing the performance are provided.

This paper describes the application of a new method to calculate energies of solar active regions and sunspots in satellites images. Digital images from the Solar & Heliospheric Observatory (SOHO) satellite are used I n this study and energies are calculated using the Ising model, which has been modified for our application. The new algorithm has been integrated with our

In this work we study the association between eruptive filaments/prominences and coronal mass ejections (CMEs) using machine learning-based algorithms that analyse the solar data available between January 1996 and December 2001. The support vector machine (SVM) learning algorithm is used for the purpose of knowledge extraction from the association results. The aim is to identify patterns of associations that can be represented using SVM learning rules for the subsequent use in near real-time and reliable CME prediction systems. Timing and location data in the US National Geophysical Data Center (NGDC) filament catalogue and the Solar and Heliospheric Observatory/Large Angle and Spectrometric Coronagraph (SOHO/LASCO) CME catalogue are processed to associate filaments with CMEs. In the previous studies, which classified CMEs into gradual and impulsive CMEs, the associations were refined based on the CME speed and acceleration. Then the associated pairs were refined manually to increase the accuracy of the training dataset. In the current study, a data-mining system is created to process and associate filament and CME data, which are arranged in numerical training vectors. Then the data are fed to SVMs to extract the embedded knowledge and provide the learning rules that can have the potential, in the future, to provide automated predictions of CMEs. The features representing the event time (average of the start and end times), duration, type, and extent of the filaments are extracted from all the associated and not-associated filaments and converted to a numerical format that is suitable for SVM use. Several validation and verification methods are used on the extracted dataset to determine if CMEs can be predicted solely and efficiently based on the associated filaments. More than 14 000 experiments are carried out to optimise the SVM and determine the input features that provide the best performance.

Machine-learning algorithms are applied to explore the relation between significant flares and their associated CMEs. The NGDC flares catalogue and the SOHO/LASCO CME catalogue are processed to associate X and M-class flares with CMEs based on timing information. Automated systems are created to process and associate years of flare and CME data, which are later arranged in numerical-training vectors and fed to machine-learning algorithms to extract the embedded knowledge and provide learning rules that can be used for the automated prediction of CMEs. Properties representing the intensity, flare duration, and duration of decline and duration of growth are extracted from all the associated (A) and not-associated (NA) flares and converted to a numerical format that is suitable for machine-learning use. The machine-learning algorithms Cascade Correlation Neural Networks (CCNN) and Support Vector Machines (SVM) are used and compared in our work. The machine-learning systems predict, from the input of a flare’s properties, if the flare is likely to initiate a CME. Intensive experiments using Jack-knife techniques are carried out and the relationships between flare properties and CMEs are investigated using the results. The predictive performance of SVM and CCNN is analysed and recommendations for enhancing the performance are provided.

In this paper, we introduce two novel models for processing real-life satellite images to quantify and then visualise their magnetic structures in 3D. We believe this multidisciplinary work is a real convergence between image processing, 3D visualisation and solar physics. The first model aims to calculate the value of the magnetic complexity in active regions and the solar disk. A series of experiments are carried out using this model and a relationship has been indentified between the calculated magnetic complexity values and solar flare events. The second model aims to visualise the calculated magnetic complexities in 3D colour maps in order to identify the locations of eruptive regions on the Sun. Both models demonstrate promising results and they can be potentially used in the fields of solar imaging, space weather and solar flare prediction and forecasting.

In this paper, a machine-learning-based system that could provide automated short-term solar flare prediction is presented. This system accepts two sets of inputs: McIntosh classification of sunspot groups and solar cycle data. In order to establish a correlation between solar flares and sunspot groups, the system explores the publicly available solar catalogues from the National Geophysical Data Center to associate sunspots with their corresponding flares based on their timing and NOAA numbers. The McIntosh classification for every relevant sunspot is extracted and converted to a numerical format that is suitable for machine learning algorithms. Using this system we aim to predict whether a certain sunspot class at a certain time is likely to produce a significant flare within six hours time and if so whether this flare is going to be an X or M flare. Machine learning algorithms such as Cascade-Correlation Neural Networks (CCNNs), Support Vector Machines (SVMs) and Radial Basis Function Networks (RBFN) are optimised and then compared to determine the learning algorithm that would provide the best prediction performance. It is concluded that SVMs provide the best performance for predicting whether a McIntosh classified sunspot group is going to flare or not but CCNNs are more capable of predicting the class of the flare to erupt. A hybrid system that combines a SVM and a CCNN is suggested for future use.

In this paper, Associations algorithms and Support Vector Machines (SVM) are applied to analyse years of solar catalogues data and to study the associations between eruptive filaments/prominences and Coronal Mass Ejections (CMEs). The aim is to identify patterns of associations that can be represented using SVM learning rules to enable real-time and reliable CME predictions. The NGDC filaments catalogue and the SOHO/LASCO CMEs catalogue are processed to associate filaments with CMEs based on timing and location information. Automated systems are created to process and associate years of filaments and CME data, which are later arranged in numerical training vectors and fed to machine learning algorithms to extract the embedded knowledge and provide learning rules that can be used for the automated prediction of CMEs. Features representing the filament time, duration, type and extent are extracted from all the associated (A) and not-associated (NA) filaments and converted to a numerical format that is suitable for machine learning use. The machine learning system predicts if the filament is likely to initiate a CME. Intensive experiments are carried out to optimise the SVM. The prediction performance of SVM is analysed and recommendations for enhancing the performance are provided.

The Halloween storm, which occurred late October early November 2003, caused serious problems including damaging 28 satellites, knocking two out of commission, diverting airplane routes and causing power failures. In this paper, we tested our fully Automated Solar Activity Prediction (ASAP) tool with the solar data corresponding to this period. The prediction capability of the tool is evaluated using various performance measures. With this study we are aiming to answer if solar flares during Halloween storm could have been predicted using ASAP and if ASAP can be used for the prediction of such extreme events in the future.

Since the Solar Dynamics Observatory (SDO) began recording ~ 1 TB of data per day, there has been an increased need to automatically extract features and events for further analysis. Here we compare the overall detection performance, correlations between extracted properties, and usability for feature tracking of four solar feature-detection algorithms: the Solar Monitor Active Region Tracker (SMART) detects active regions in line-of-sight magnetograms; the Automated Solar Activity Prediction code (ASAP) detects sunspots and pores in white-light continuum images; the Sunspot Tracking And Recognition Algorithm (STARA) detects sunspots in white-light continuum images; the Spatial Possibilistic Clustering Algorithm (SPoCA) automatically segments solar EUV images into active regions (AR), coronal holes (CH) and quiet Sun (QS). One month of data from the SOHO/MDI and SOHO/EIT instruments during 12 May - 23 June 2003 is analysed. The overall detection performance of each algorithm is benchmarked against National Oceanic and Atmospheric Administration (NOAA) and Solar Influences Data Analysis Centre (SIDC) catalogues using various feature properties such as total sunspot area, which shows good agreement, and the number of features detected, which shows poor agreement. Principal Component Analysis indicates a clear distinction between photospheric properties, which are highly correlated to the first component and account for 52.86% of variability in the data set, and coronal properties, which are moderately correlated to both the first and second principal components. Finally, case studies of NOAA 10377 and 10365 are conducted to determine algorithm stability for tracking the evolution of individual features. We find that magnetic flux and total sunspot area are the best indicators of active-region emergence.

... noise for large numbers of electron emissions [4]. This noise comes from the quantum nature oflight ... types of noise and for different applications (ie solar imaging, remote sensing andmammography, etc ... 4] MG Lfdahl, MJ v. Noort, and C. Denker, "Solar image restoration," F. Kneer ...

In this paper, a new method is applied to calculate the magnetic energies of active regions and sunspots in satellites images. Digital images from the Solar & Heliospheric Observatory (SOHO) satellite are used in this study. The magnetic energies are calculated using the Ising model, which has been modified for our application. The new algorithm is integrated with our existing

This paper presents a hybrid system for automatic detection and McIntosh-based classification of sunspot groups on SOHO/MDI white-light images using active-region data extracted from SOHO/MDI magnetogram images. After sunspots are detected from MDI white-light images they are grouped/clustered using MDI magnetogram images. By integrating image-processing and neural network techniques, detected sunspot regions are classified automatically according to the McIntosh classification system. Our results show that the automated grouping and classification of sunspots is possible with a high success rate when compared to the existing manually created catalogues. In addition, our system can detect and classify sunspot groups in their early stages, which are usually missed by human observers.

The importance of real-time processing of solar data especially for space weather applica-tions is increasing continuously, especially with the launch of SDO which will provide sev-eral times more data compared to previous solar satellites. In this paper, we will show the initial results of applying our Automated Solar Activity Prediction (ASAP) system for the short-term prediction of significant solar flares to SDO data. This automated system is cur-rently working in real-time mode with SOHO/MDI images and its results are available online (http://spaceweather.inf.brad.ac.uk/) whenever a new solar image available. This system inte-grates image processing and machine learning to deliver these predictions. A machine learning-based system is designed to analyse years of sunspots and flares data to extract knowledge and to create associations that can be represented using computer-based learning rules. An imaging-based real time system that provides automated detection, grouping and then clas-sification of recent sunspots based on the McIntosh classification and integrated within this system. The results of current feature detections and flare predictions of ASAP using SOHO data will be compared to those results of ASAP using SDO data and will also be presented in this paper.

Space weather forecasting is a very challenging task and investigating the associations between properties (i.e., shape, scale, location) of the related solar features, appearing in solar images, are usually complicated because of the variation in their physical and visual properties. Establishing the correlations among the occurrences of solar activities and solar features is a long-standing problem in solar imaging. This work is an attempt to shed more light on the driving forces behind the initiations of Coronal Mass Ejections (CMEs). This is still a big mystery in this field and in this work we have analysed years of data relating to one particular feature, filaments, to determine if an association between filaments and the eruptions of CMEs can be drawn. The resulting association set has been fed to a powerful machine learning algorithm to determine if CMEs can be predicted solely based on filaments. Our learning algorithm, AdaBoost, is used because of robust and accurate performance. Three of the most common versions of the Adaboost algorithm are used in this work, which are the Gentle AdaBoost, the Real AdaBoost and the Modest AdaBoost.

In this paper, Hidden Markov Models (HMMs) are used to study the evolution of sunspots and to develop a model that can be used to predict the McIntosh class and the sunspot area for the sunspot under investigation for the next 24 hours. The testing results show accuracy in the prediction of next-day area and McIntosh classification reaching up to 71% and 60% respectively, when studied on the period from 18/08/1996 till 31/03/2006.

ABSTRACT Solar imaging is currently an active area of research. In this study a 3D modeling technique for magnetic field lines on Sun is provided. Magnetic field foot prints are detected from SOHO (Solar and Heliospheric Observatory) / MDI (Michelson Doppler Imager) Magnetogram images. Using negative and positive footprints possible dipole pairs are found according to their proximity. Then using this data, 3D models are built on bipolar coordinates and placed on detected pairs after transformations.