Session 13 - System Science: Application to space weather analysis, modelling and forecasting

Richard Boynton (Univ of Sheffield); Homayon Aryan (Goddard Space Flight Center); George Balasis (IAASARS, NOA); Enrico Camporeale (CWI)
Friday 01/12, 9:45 - 13:00
Delvaux


KEYWORDS - System science, machine learning, data assimilation, information theory, signal processing

The construction of accurate dynamical models is fundamental to forecasting the many aspects of space weather. The process of deducing models for forecasting traditionally involved breaking the system into component parts and applying the laws of physics to each part to build up a description of that system. However, for complex space weather systems, we do not have enough knowledge about some of the processes involved to build an accurate model solely from first principles. Alternatively, complex systems science based methods have been developed to deduce dynamical models from input-output data. The techniques developed from system science, such as system identification, machine learning, data assimilation, information theory, signal processing, among others, are applicable to any system that has large amounts of data availability. With the increasing amount of space weather data, we are able to make use of these tools and techniques to analyse, model and forecast the complex systems of space weather. As such, this session is for contributions that employ these state of the art tools that have been developed in system science.

Poster Viewing
From Thursday morning to Friday noon

Talks
Friday December 1, 09:45 - 11:00, Delvaux
Friday December 1, 11:45 - 13:00, Delvaux

Click here to toggle abstract display in the schedule

Talks : Time schedule

Friday December 1, 09:45 - 11:00, Delvaux

09:45	Understanding performance of neural network models for short-term predictions applied to geomagnetic indices	Wintoft, P et al.	Invited Oral
	Peter Wintoft, Magnus Wik
	Swedish Institute of Space Physics
	Many processes in the solar-terrestrial chain relevant for space weather are nonlinear with temporal dynamics (memory). Different nonlinear system identification approaches, such as support vector machines, neural network models, and NARMAX models, have successfully been applied to the analysis and prediction in space weather. They perform very well compared to numerical MHD models (Rast\"{a}tter et al., 2013) and in the prediction mode they are computationally light-weight which means that large number of events can be analysed and they can be implemented for real-time operation. In this work we specifically focus on neural networks. A great challenge is the collection and preprocessing of data from which the mathematical relations can be extracted and tested. This includes studying the distribution of data and identifying the tails (extremes), a task that becomes complex in multi-dimensional space (e.g. solar wind plasma and magnetic fields), and especially also when the temporal evolution needs to be considered. For example, a point in an one-dimensional case may lie well inside from the tail, but when embedded in state-space the point can become extreme. The extremes are important to study because they are often related to extreme effects but also because they are the most difficult from a modelling perspective. To some degree tackle the difficulties at the extremes we use an ensemble of models from which the median prediction is used. The data set size and distribution put limits on the accuracy and operational range of the models. The models performs best in the bulk of the distribution and can not extrapolate outside the observed distribution and it is important to understand when this happens and possibly look at simpler alternative models. We will describe the latest developments along that described above applied to the short-term predictions of geomagnetic indices \emph{Kp}, \emph{Dst}, \emph{AE}, and local geomagnetic $dB/dt$. We will also compare with various simple relations for the driving phase of the storm such as $V B_{\bot} \sin^4 \theta/2$ and $R_\mathrm{quick}$ (Borovsky and Birn, 2013). To be able to perform a more detailed analysis of the storm events we apply a time series wavelet analysis in order to determine range, length, and phases of individual storms. Finally, we describe the verification carried out be able to asses the performance. This work has in part been supported by ESA SSA Space Weather ESC contract No 4000113185/15/D/MRP and European Union's Horizon 2020 grant agreement No 637302 (PROGRESS).
10:10	Automatically worked stages for estimation of the level of expected radiation hazards from SEP events	Dorman, L et al.	Oral
	Lev Dorman[1,2] and Lev Pustil’nik[1]
	[1]Israel Cosmic Ray and Space Weather Center, affiliated to Tel Aviv University, Israel Space Agency, and Shamir Research Institute; [2]Institute of Terrestrial Magnetism, Ionosphere and Radiowave Propagation RAN of N.V. Pushkov (IZMIRAN), Moscow, RU-142190, Russia
	In the last years thanks to formation of NMDB became possible to have on-line through Internet one-minute cosmic ray (CR) data from many neutron monitors and muon telescopes (in high energy region) as well as from several spacecrafts (in very low energy region). To avoid damage of electronics and negative effects for people health is necessary on-line forecast expected fluency of energetic particles and radiation hazards. It was shown by myself and colleagues in many original research papers that this possible to do by using the first 20-30 minutes of CR data on the basis of coupling functions, spectrographic method, and by solving inverse problem, and then calculate expected results on radiation hazards for many hours of great Solar Energetic Particle (SEP) event. Usually it takes a lot of time (at least, several months). But for really protection of satellites, aircrafts, and people from dangerous radiation hazards all this must be made automatically, including the formation of corresponding alerts on the expected level of radiation hazard for different objects. We describe several automatically worked stages and obtain corresponding algorithms. The first stage works continue, collecting from Internet all available one minute data on CR variations (corrected on meteorological and geomagnetic effects). The second stage also works continue according to automatically working program "SEP-Start" - supposed, developed and checked in the Israel Cosmic Ray and Space Weather Center. Using of this program on many CR stations and on satellites allowed to determine automatically the beginning of SEP event (it can be different at different stations caused to anisotropy at beginning of SEP). If the second stage gives positive result, starts to work automatically the third stage according to program "SEP-Coupling" – using method of coupling functions and spectrographic method for transformation obtained at different altitudes and cutoff rigidities data on CR intensity variations to the space and calculation CR energy spectrum and angle distribution out of the Earth’s atmosphere and magnetosphere, directly in the interplanetary space near the Earth. After obtaining results by third stage starts to work automatically the fourth stage according to program "SEP-Inverse Problem", and it is determined source function, time of ejection SEP into interplanetary space, and diffusion coefficient of propagation in dependence of SEP energy and distance from the Sun. After obtaining results by fourth stage starts to work automatically the fifth stage according to program "SEP-Direct Problem", and it is determined by found at fourth stage parameters the time variations of primary SEP in dependence of particles energy in interplanetary space near the Earth for many hours ahead, up to few days (on the basis of only 20-30 minutes of SEP beginning). On the basis of information, obtained in the fifth stage, it is easy to calculate by known coupling functions and cutoff rigidities expected time variations of SEP intensity in SPACE- and AIR-CRAFTS at different trajectories, and compare the beginning part with available observations and estimate the quality of forecasting (sixth stage, program “SEP-Forecasting”). If the forecasted radiation hazard is expected dangerous for different objects, will be immediately send corresponding Alerts (seventh stage, program “SEP-Alerts”). With time by obtaining new data, forecasting Alerts became more and more exactly. We prepared all algorithms to realize this program automatically. We consider also how to extend the program for more complicated cases, including anisotropic SEP events. Keywords: great SEP events, radiation hazards, automatically forecasting, algorithms
10:30	Empirical modeling of the plasmasphere dynamics using neural networks	Zhelavskaya, I et al.	Oral
	Irina Zhelavskaya[1,2],Yuri Shprits[1,2,3],Maria Spasojevic[4]
	[1]Helmholtz Centre Potsdam, GFZ German Research Centre For Geosciences, Potsdam, Germany,[2]University of Potsdam, Potsdam, Germany,[3]University of California, Los Angeles, Los Angeles, CA, USA,[4]Hansen Experimental Physics Laboratory, Stanford University, CA, USA
	We propose a new empirical model for reconstructing the global dynamics of the cold plasma density distribution based only on solar wind data and geomagnetic indices. Utilizing the density database obtained using the NURD (Neural-network-based Upper hybrid Resonance Determination) algorithm for the period of October 1, 2012 - July 1, 2016, in conjunction with solar wind data and geomagnetic indices, we develop a neural network model that is capable of globally reconstructing the dynamics of the cold plasma density distribution for 2 ≤ L ≤ 6 and all local times. We validate and test the model by measuring its performance on independent datasets withheld from the training set and by comparing the model predicted global evolution with global images of He+ distribution in the Earth’s plasmasphere from the IMAGE Extreme UltraViolet (EUV) instrument. We identify the parameters that best quantify the plasmasphere dynamics by training and comparing multiple neural networks with different combinations of input parameters (geomagnetic indices, solar wind data, and different durations of their time history). We demonstrate results of both local and global plasma density reconstruction. This study illustrates how global dynamics can be reconstructed from local in-situ observations by using machine learning techniques.
10:45	Variation of geomagnetic responses to the solar wind input: Half-year increase during declining phases and 2009 specialty	Yamauchi, M et al.	Oral
	M. Yamauchi[1] and B. Olsthoorn[2]
	[1]Swedish Institute of Space Physics (IRF), Kiruna; [2]Master course student, Department of Physics, Stockholm University
	Variations of the Sun-Earth coupling efficiency (AL, AU, ASY-D, and SYM-H indices for the same level of solar wind electromagnetic energy input determined by Akasofu's energy coupling function "epsilon") since 1981 were examined using NASA/OMNI 5-min data. The seasonal variation was removed by averaging the indices for given ranges of the epsilon values over every three months around equinoxes and solstices. (1) For small to moderate epsilon, the AL and AU responses to the same epsilon shortly increased beyond the fluctuation level for about half a year for both equinoxes and solstices during 1983, 1994, 2003, and late 2015, all during early declining phase of solar cycles, although the timing does not necessarily the same as the peak solar wind energy input; (2) Except these singular periods, this Sun-Earth coupling efficiency for small to moderate epsilon continuously decreased over the past three decades until 2009, and then started to recover afterword. (3) The short increase during the singular periods and long-term trend with 2009 low are also found in the mid-altitude ASY-D index, but are not clear in SYM-H; The long term change in the Sun-Earth coupling efficiency raises a possibility that it can be related to the strength of the solar cycle. If this is true, the strength of the solar cycle 25 will somewhat recover from current solar cycle 24.

Friday December 1, 11:45 - 13:00, Delvaux

11:45	Global Solar Magnetic Maps and Forecasting Space Weather with ADAPT	Henney, C et al.	Oral
	Carl J. Henney[1], Kathleen Shurkin[2], C. Nick Arge[3]
	[1]AFRL, KAFB, NM, USA; [2]ISR, Boston College, Chestnut Hill, MA, USA; [3]NASA Goddard, Greenbelt, MD, USA
	Progress toward the forecasting of key space weather parameters, up to 7 days in advance, using SIFT (Solar Indices Forecasting Tool) with the ADAPT (Air Force Data Assimilative Photospheric flux Transport) model will be presented. The SIFT forecasting method reviewed here is outlined in Henney et al. 2012 and Henney et al. 2015. The new method utilizes the solar near-side magnetic field distribution estimated with the ADAPT flux transport model as input to the SIFT empirical models that predict typical input parameters to ionospheric and thermospheric models; e.g., selected bands (between 0.1 to 175 nm) of solar soft X-ray (XUV), far ultraviolet (FUV), and extreme ultraviolet (EUV) irradiance, along with observed F10.7 (solar 10.7 cm, 2.8 GHz, radio flux), the sunspot number (SSN), and the Mg II core-to-wing ratio. Input to the ADAPT model includes photospheric magnetograms from the NISP (NSO Integrated Synoptic Program) ground-based instruments, GONG and VSM. The ADAPT flux transport model evolves an ensemble of realizations, using relatively well-understood transport processes during periods for which there are no observations, and updates the ensemble using data assimilation methods that rigorously take into account model and observational uncertainties. We have updated the ADAPT model to utilize line-of-sight and vector magnetograms, along with helioseismic far-side detections, from the Helioseismic and Magnetic Imager (HMI) to create global radial field distribution maps. We also plan to incorporate data from the Polarimetric and Helioseismic Imager (PHI) on Solar Orbiter into ADAPT to provide uniquely vital full-disk vector magnetogram input from the far-side and higher latitude polar regions. ADAPT model development is supported primarily by AFRL, with additional support from NASA.
12:05	Use of systems based models for the forecasting of space weather	Walker, S et al.	Oral
	S. N. Walker[1], T. Arber[2], K. Bennett[2], M. Liemohn[3], B. van der Holst[3], P. Wintoft[4], N. Y. Ganushkina[5], and M. A. Balikhin[1]
	[1]Automatic Control and Systems Engineering, University of Sheffield, Sheffield, UK; [2]Dept Physics, University of Warwick, Coventry, UK; [3]Climate and Space Sciences Engineering, University of Michigan, Michigan, USA; [4]Swedish Institute of Space Physics, Lund, Sweden; [5]Finnish Meteorological Institute, Helsinki, Finland
	Changes in the solar wind induce a response in the magnetosphere through a complex series of processes and interactions. The underlying philosophy used within numerical codes is to understand and model each process individually before conjugating them to model the overall system. The diversity and scales of the processes involved in such a chain is a major drawback for this kind of approach. Systems science proves a second, alternate route for modelling such process chains. These methodologies, which are data driven, study the evolution of a system as a whole, based on a set of system inputs. In contrast to other data driven methodologies, system science techniques can be used to probe the underlying physics and may also be used for the validation of numerical and analytical models.
12:25	Evidence on second-order phase transition of the magnetosphere around magnetic storms	Balasis, G et al.	Oral
	Georgios Balasis[1], Ioannis A. Daglis[2,1], Yiannis Contoyiannis[3], Stelios M. Potirakis[3], Constantinos Papadimitriou[1], Nikolaos S. Melis[4], Omiros Giannakis[1], Athanassios Papaioannou[1], Anastasios Anastasiadis[1], and Charalampos Kontoes[1]
	[1]Institute for Astronomy, Astrophysics, Space Applications and Remote Sensing, National Observatory of Athens, Greece; [2]Section of Astrophysics, Astronomy and Mechanics, Department of Physics, National and Kapodistrian University of Athens, Greece; [3]Department of Electronics Engineering, Piraeus University of Applied Sciences (TEI of Piraeus), Greece; [4]Institute of Geodynamics, National Observatory of Athens, Greece
	Criticality has been proposed as a suitable framework to study the nonlinear system of the Earth's magnetosphere. In particular, it has been suggested that phase transitions may shed light on the mechanisms of magnetospheric dynamics. The magnetic field variations observed by the mid-latitude HellENIc GeoMagnetic Array (ENIGMA) with respect to the two strongest magnetic storms of the current solar cycle are analysed using the method of critical fluctuations (MCF). The application of MCF to the ENIGMA time series reveals the existence of criticality in the range of 8 to 45 hours prior to the onset of the two storms. We show that the underlying dynamical processes in the magnetosphere prior to intense magnetic storms present dynamics analogous to those of thermal systems undergoing second-order phase transition. We discuss the consequence of this feature and the possible implications of intermittent criticality appearance on space weather forecasting efforts.
12:45	Tracking and characterizing the evolution of active regions with SDO/HMI	Attie, R et al.	Oral
	Raphael Attie[1], Barbara Thompson[1], Veronique Delouille[2]
	[1]NASA GSFC; [2]Royal Observatory of Belgium
	Measurements of solar photospheric flows by algorithms applied to imaging data can be considered part of a larger effort to automatically extract solar properties using advanced image processing methods. Because solar plasma flows exist at several scales, some of which are not immediately apparent on solar image series, sophisticated and automated tracking procedures are needed. I will present here a new framework made of two core components that analyzes (i) the horizontal plasma flows and (ii) magnetic field evolution, and that is used to characterize the evolution of active regions at unprecedented accuracy: (i) The so-called “Balltracking” algorithm - originally developed by Potts et al. (2004) to track granules in MDI images - that I developed further over the last decade. It now uses the full-cadence data series from SDO/HMI for characterizing the plasma flows during the emergence of active regions (ARs). This method derives horizontal flow fields in the Euler reference frame where granules are observed and can provide error maps and maps of the supergranular boundaries. (ii) The “Magnetic Balltracking” algorithm that tracks magnetic elements in the Lagrange reference frame from quiet sun magnetograms (Attie & Innes 2015, Attie et al. 2016). I will demonstrate its capacity to detect and track emerging magnetic flux in ARs from their earliest emergence phase -- when sunspots are not yet visible in continuum images -- onwards. Used together on SDO/HMI data, “Balltracking” and “Magnetic Balltracking” define a solar features tracking framework that helps addressing the following questions: (i) What are the properties of large-scale photospheric flows prior to and after the emergence of ARs? (ii) What is the contributions of the horizontal flows regarding the magnetic energy transport into and out of the corona? (iii) What properties of flows, magnetic field, and energy transfer add to our ability to forecast flares?

Posters

1	Long Term Variation of Latitudinal Distribution of Coronal Holes	Chargeishvili, B et al.	e-Poster
	Bidzina Chargeishvili[1], Darejan Japaridze[1], Tengiz Mdzinarishvili[1], Bidzina Shergelashvili[2,1,3]
	[1]Abastumani Astrophysical Observatory at Ilia State University, University St. 2, Tbilisi, Georgia; [2]Space Research Institute, Austrian Academy of Sciences, Schmiedlstrasse 6, 8042 Graz, Austria; [3]Combinatorial Optimization and Decision Support, KU Leuven campus Kortrijk, E. Sabbelaan 53, 8500 Kortrijk, Belgium
	The evolution of the latitudinal distribution of activity of coronal holes is studied using SOHO/EIT 171 Å data from May 1996 to February 2017. The obtained butterfly-like structure diagram shows the strong difference of evolutionary shapes of a latitudinal distribution of non-polar and polar coronal holes. The distribution of non-polar coronal holes has a clearly expressed oscillatory form. The distribution of polar coronal holes has a form of recurrent soft hills. Among the hills appear fibrous vertical patterns that are dominant for nonpolar coronal holes. Different origin of polar and non-polar coronal holes is discussed. Characteristic periodicities for all patterns are of the order of year and the Synodic rotation period. The oscillation shape of coronal hole distribution is explained by periodic overlapping of open and closed structures of magnetic structures, which are caused by annual variations in the position angle of disk center with respect to line-of-sight. North-South asymmetry is expressed for the entire period of study. Both polar and non-polar zones are more active in the Southern hemisphere. Distribution and other features of activity of coronal holes show a strong correlation with long-term data of the general magnetic field and daily hemispheric sunspots numbers. A correlation between the variation of CH distribution, solar wind, and interplanetary magnetic field is considered.
2	Three-dimensional data assimilation and reanalysis of radiation belt electrons	Cervantes villa, J et al.	p-Poster
	Juan Sebastian Cervantes-Villa[1,2], Yuri Shprits[1,2,3], Adam Kellerman[3], Alexander Drozdov[3]
	[1]GFZ German Research Centre for Geosciences, Potsdam, Germany; [2]Institute of Physics and Astronomy, University of Potsdam, Potsdam, Germany; [3]Earth, Planetary, and Space Sciences, University of California, Los Angeles, California, USA
	Satellite observations are often incomplete and inaccurate and may have only limited spatial coverage. However, through data assimilation they can be blended with information from physics-based models, in order to fill gaps and lead to a better understanding of the underlying dynamical processes. Data assimilation methods have been extensively used to analyze and predict meteorological, oceanographic, and climate processes. With the advent of space-borne observational data and the development of more sophisticated space-physics models, dynamical processes in the Earth’s radiation belts can be analyzed and assessed using data assimilation methods. In this study, reanalysis of radiation belt electrons in the Earth’s magnetosphere is achieved through data assimilation of Van Allen Probes mission and Geostationary Operational Environmental Satellite with the 3D Versatile Electron Radiation Belt code using a split-operator Kalman filter technique. The forecast capability of the data assimilative code is assessed. Results are statistically validated for several field models and boundary conditions. Sensitivity of the reanalysis electron flux to available spacecraft data is also studied.
3	Proton Prediction using Deep learning	Yang, S et al.	p-Poster
	SeungBum Yang[1], TaeYoung Kim[1], JangSeok Choi[2], DoHyun Kim[1], SoYeon Kang[1], MyungJin Choi[1]
	[1]InSpace.co.,ltd; [2]Korea Space Weather Center (KSWC) Radio Research Agency
	We recently developed a model for proton prediction using deep learning technology which is adopted in various fields, and describe input data specification for proton prediction and processing process of preprocessing software and deep learning model for proton prediction. The proton prediction of the deep learning base predict the proton as polymorphism, various scientific data gathered in various kinds of space environments are input and the proton is predicted. The data used for deep learning was 31,141 in the period of SOHO MDI magnetogram 1996~2010, 15200 in the period of Continuum 1996~2010, and 5,110 data from 2010 in 1996~2017. Constructed learning data by purifying data with different forms(Image or Text), executing matching in chronological order and we have designed a combined deep learning model using Convolution Layer of CNN which is deep learning method specialized in image and LSTM(Long-short term memory) technique specialized for time series analysis. As a result of calculating accuracy and using proton prediction using PE(Percent Error) method widely used for time series accuracy analysis for accuracy analysis of results, it is possible to predict the distribution and pattern similarly. In this paper, we try to develop a technique of deep learning from data preprocessing for deep learning based proton prediction, and an overall process of calculating prediction accuracy. Using this model, a binary classification method and a regression model for predicting the amount of protons were applied so that the presence or absence of a sun event could be determined.
4	Solar Demon flare and dimming statistics from AIA observations 2010-2017	Kraaikamp, E et al.	p-Poster
	Emil Kraaikamp, Cis Verbeeck
	Royal Observatory of Belgium
	Solar Demon has been covering the majority of Solar Cycle 24 from June 2010 up to now, employing SDO/AIA images to perform near real-time detections and to create science catalogs of solar eruptive events (flares, dimmings and EUV waves). In total, over 10 000 flares and 4 700 dimming events have been detected in the AIA 9.4 and 21.1 nm channels respectively. The science quality catalog provides a wealth of information on event location, intensity, duration, and related parameters. Here we present the statistical distribution of these parameters for the flare and dimming events detected by Solar Demon.
5	Forecasting the AE indices using machine learning	Wik, M et al.	p-Poster
	Magnus Wik, Peter Wintoft
	Swedish Institute of Space Physics
	The AE indices are a measure of the auroral electrojet activity and geomagnetic substorms. These indices are often used in studies related to space weather effects, such as e.g. geomagnetically induced currents (GIC), and in radiation belt models. We predict the AE, AL and AU indices, using neural networks, driven by solar wind data B, By, Bz, plasma density and velocity. In this study we used ACE level 2 data. Due to the high time resolution of the indices, which is not realistic to be captured by any model, the AE indices and solar wind data were resampled to 5 minutes. The networks were trained using 10 years of data, spanning 18 years in total from 1998 to 2015, and tested against 4 years of data. Since there is also a UT and seasonal dependence, we also added sine and cosine functions of UT and day-of-year. We performed several model studies, to analyse the importance of input parameters, network topology, lead time and time delays. The models are verified using standard forecast verification techniques. This work has been supported by the European Union’s Horizon 2020 grant agreement No 637302 (PROGRESS).
6	A Comparative Performance Study of Machine Learning Algorithms for Space Weather Forecasting	Nanouris, N et al.	p-Poster
	Nikolaos Nanouris, Panos Boumis
	Institute for Astronomy, Astrophysics, Space Applications and Remote Sensing, National Observatory of Athens, 15236 Penteli, Greece
	In this study we evaluate the efficiency of numerous supervised machine learning classifiers addressing the feasibility of predicting events occurring in a fixed temporal interval with a strong emphasis to solar flares. The assessment was accomplished via robust metrics which are less vulnerable to the random chance and the number of instances building the sample (e.g., statistics on the receiver operating characteristic curves and skill scores such as HSS and TSS) in terms of a 10-fold cross-validation procedure. Two datasets were used toward this direction. The first case concerns the prediction of the most energetic solar flares, namely the M- and X-type (positive class), with respect to the weaker B- and C-type ones (negative class), based on data provided by the GOES satellite from 2010 to 2016. In order to find physical paths driving these events, the 13 standard SHARP variables, derived from vector magnetograms of the active regions, were adopted as potential predictors of flaring activity within a 24-hour interval before the occurrences. An attempt to classify independently all four types of flares is also performed and compared to the results of similar earlier works. The second case deals with the optimization of operations at Helmos observatory, Greece, situated at a noticeable altitude of 2340 m with strongly fast-changing and irregular weather conditions which render the observing procedure uncertain. The approach was based on 6 meteorological parameters obtained from the sole station situated at this distinguished location. Still in a binary form, our response variable is considered positive as long as the successful observing time exceeds a selected fraction of each night, while negative if this time is shorter. The data spanned from 2014 to 2017 during the intense (summer-to-fall) 4-month operating period of the observatory, regarding a 6-hour interval before the observing process. Among 60 algorithms, tested through the WEKA software, 4 proved to be extremely competitive; although powerful, neither managed to achieve evenly high success rates for both the two classes. In order to provide a unified and handful approach, we suggest a hybrid scheme which includes a combination of two algorithms and its probabilistic treatment in the Bayesian framework with objective predictive probabilities close to 90%.
7	Using the Local Ensemble Transform Kalman Filter (LETKF) For Upper Atmosphere Modelling	Angling, M et al.	p-Poster
	Sean Elvidge, Matthew J. Angling
	University of Birmingham, Birmingham, UK
	This paper describes the local ensemble transform Kalman filter (LETKF) and presents the initial results of the Advanced European electron density (Ne) Assimilation System (AENeAS). AENeAS is a new physics-based data assimilation model of the ionosphere/thermosphere. The model assimilates electron density virtual height profiles from ionosondes and TEC measurements from GNSS receivers using the LETKF. The LETKF is an ensemble Kalman filter variant which combines the ensemble transform Kalman filter (ETKF) with the local ensemble Kalman filter (LEKF). The localization in the LETKF allows the analysis to be performed around each model grid point and completely in parallel. The LETKF results are equivalent to the LEKF results but are calculated in a more efficient manner, similar to the ETKF. Like any Kalman filter the LETKF requires a background model. AENeAS currently uses the Thermosphere Ionosphere Electrodynamics General Circulation Model (TIE-GCM). However, the maximum altitude modelled by TIE-GCM is between 500 – 700 km (depending on solar conditions). Since AENeAS is designed to provide ionospheric parameters such as the total electron content (TEC) an NeQuick topside is fitted above these heights.
8	Space radiation study based on cascades simulations in geospace	Tezari, A et al.	p-Poster
	Pavlos Paschalis[1], Helen Mavromichalaki[1], Anastasia Tezari[1,2], Maria Gerontidou
	[1]Nuclear and Particle Physics Department. Faculty of Physics, National and Kapodistrian University of Athens, 15784 Athens Greece; [2]Medical School, National and Kapodistrian University of Athens, 11527 Athens Greece
	Space radiation is ionizing, such as solar energetic particles (SEP) and galactic cosmic rays (GCR), and non-ionizing (UV-radiation). Ionizing radiation has many biological and technological effects, and therefore, exposure to space radiation may place astronauts and aviation crews at significant risk. DYASTIMA (Dynamic Atmospheric Shower Tracking Interactive Model Application) is a new Geant4 software application that simulates the atmospheric showers of secondary particles of galactic or solar origin in the atmospheric layers, providing all the necessary information about the cascade. DYASTIMA-R is an extension of DYASTIMA and may be used for the calculation of aircrew and spacecrew dose rate and equivalent dose rate for various types of particles at different atmospheric altitudes. This method is based on Monte Carlo simulations with the usage of phantoms and allows the calculation of the dose in flight scenarios, characterized by different altitudes, different geographic latitudes and different solar and cosmic ray activity, and at the same time the experimentation with different shielding materials. Preliminary results are presented.
9	Operational control of near-Earth’s radiation conditions by space weather services at SMDC MSU	Bobrovnikov, S et al.	p-Poster
	S. Bobrovnikov, V.Kalegaev, V.Barinova, N.Kuznetsov, I. Myagkova, D. Nguyen, Yu. Shugay, N. Vlasova
	Skobeltsyn Institute of Nuclear Physics, Moscow State University, Moscow, 119991, Russia
	Space Monitoring Data Center (SMDC) of Moscow State University provides mission support for Russian satellites and give operational analysis and forecasting of radiation conditions in space. SMDC Web-sites (http://smdc.sinp.msu.ru/ and http://swx.sinp.msu.ru/) give access to current data about the level of solar activity, geomagnetic and radiation state of Earth’s magnetosphere and heliosphere in the near-real time. The scientific models of space environment factors have been converted to operational engineering services. They are implemented as space weather Web-applications that provide forecasts of geomagnetic and radiation condition at given satellite orbits. Radiation dose and SEE rate control are of special importance in a practical satellite operation. Satellites are always under the influence of high-energy particle fluxes during their orbital flight. The three main sources of particle fluxes: the Earth’s radiation belts, the galactic cosmic rays, and the solar energetic particles (SEP), are taken into account by SMDC operational services to estimate the radiation dose caused by high-energy particles to a satellite at LEO orbits taking into account the geomagnetic cut-off depending on geomagnetic activity level. Complex analysis of the data “from Sun to Earth” obtained during disturbed period in the end of February 2014 has been carried out on the base of SMDC data and facilities. Research supported by RSF Grant No 16-17-00098.
10	A self-consistent method for deriving polar ionospheric convection from eigenanalysis of SuperDARN radar data	Shore, R et al.	p-Poster
	Robert Shore, Mervyn Freeman, Gareth Chisham
	British Antarctic Survey, UK
	We present the results of applying a meteorological analysis method called Empirical Orthogonal Functions (EOF) to month-long samples of polar ionospheric plasma velocity data from SuperDARN. The EOF method is used to characterise and separate contributions to the variability of plasma motion in the northern polar ionosphere. EOFs decompose the noisy and sparse SuperDARN data into a small number of independent spatio-temporal basis functions, for which no a priori specification of source geometry is required. We use these basis functions to infill where data are missing. This infill only converges when it reinforces patterns present in the original data, thus providing a self-consistent description of the plasma velocity at the original temporal resolution of the SuperDARN data set. The leading modes of the EOF decomposition are found to be the two-cell Dungey-cycle convection pattern associated with IMF Bz, a single-cell perturbation to it associated with IMF By, and other modes. The relative importance of these modes (i.e., relative contribution to the total variance) is found to vary with season. These results are compared against previous results from an EOF analysis of equivalent currents from ground-based magnetometer data. Our study is a proof-of-concept, aimed to test the methodology which will subsequently be applied to SuperDARN data spanning a full solar cycle.
11	Training a new generation of Space Weather experts in Machine Learning	Amaya, J et al.	p-Poster
	Jorge Amaya, Diego Gonzalez-Herrero, Giovanni Lapenta
	Mathematics Department, KU Leuven, Belgium
	One of the greatest challenges of Space Weather is forecasting solar wind conditions at 1AU from remote observations of the Sun. The power of classical weather forecasting is derived from the close integration of simulations and measurements using data assimilation. But contrary to weather forecasting, in Space Weather it is currently impossible to take constant measurements of the solar wind conditions at different points in space. Space Weather has to rely on different approaches to connect the Sun to the Earth. Operational services today favor two main approaches: physics based modeling of the solar wind using MHD simulations, and empirical models derived from statistical analysis of historical records. The former method is complex, requires the use of large computational systems, and is sensitive to initial and boundary conditions; the later is simple and fast, but is not based in physical laws and requires human intervention for the correction of anomalies. One alternative to these two systems is the use of Artificial Intelligence (AI). The field of Machine Learning (ML), a sub domain of AI, has gained an enormous push in recent years. ML works in a similar way as the empirical models discussed before, but requires no human intervention for the selection and fitting of the different model coefficients. ML can produce empirical models that correlate two sets of physical data, autonomously and automatically without the intervention of human experts. ML is also based on very simple mathematical principles, which makes it easy to understand and to implement on a computer. Teaching how to use this new technology is one of the goals of the Centre for mathematical Plasma-Astrophysics at KU Leuven. Using data gathered from space missions, we are teaching our Space Weather course students how to properly build and run a ML algorithm for forecasting. We build the projects by emphasizing the importance of data analysis (treatment and reduction), model definition, training evaluation, and validation. In this presentation we will show how the ML concepts have been introduced to our students and we will show how in time such models have evolved in complexity leading to more accurate results. We are convinced that ML will be an integral part of Space Weather in the near future and we are training the new generation of data analysts in a a technology that can be applied to Space Weather, but also to different fields of research and industry.
12	Robust Nonlinear Predictive Model Identification for Kp index Forecasting	Wei, H et al.	p-Poster
	Yuanlin Gu, Hua-Liang Wei, Simon N. Walker and Michael A. Balikhin
	Department of Automatic Control and Systems Engineering, University of Sheffield, S1 3JD, Sheffield, UK
	Forecasting of Kp index is important for understanding the dynamic relationship between the magnetosphere and solar wind. This study presents 3-hours ahead prediction for Kp index using NARMAX models built with a novel robust model structure detection method through a multi-model learning approach. Numerical results show that the models with robust structure can produce very good Kp forecast performance and provide transparent and compact representations of the relationship between Kp index and solar wind parameters and magnetic field indices, namely, solar wind velocity (V), southward interplanetary magnetic field (Bs), solar wind rectified electric field (VBs), and dynamic flow pressure (P). It is expected that robustness and conciseness of the models can highly benefit the space weather forecast tasks.
13	Hidden factors in Solar-Terrestrial Connection as reason of natural limitation on forecasting efficiency of space weather impacts	Pustilnik, L et al.	p-Poster
	Pustil’nik Lev A.
	Israel Space Weather and Cosmic Ray Center of Tel Aviv University, Shamir Research Institute and Israel Space Agency
	Solar-Terrestrial Connection (STC) caused by complex and dynamic process inside the Sun. First of all it is non-stable dynamo process in the Sun; the second one is non-stability of the global atmospheric circulation in the Earth. These reasons lead from time to time to drastically change of the STC manifestations as whole. Partly it leads to change of the phase relations between different components of the Earth environment responses (EER) on solar activity (SA). The dynamo mechanism what control SA, includes few basically elements (differential rotation, solar convection, toroidal and poloidal magnetic fields, global circulation, helicity of turbulence in convective zone, and more). All this elements are connected one to another with direct and feedbacks causal-reason relations between them. Most from these elements are hidden for direct observation and cannot be measured and taking into account in forecasting methods of SA. In result solar cycle has dynamically change in amplitude, in period and in phase behavior up to phenomena of strange attractor. This non-stationarity leads to different phase pattern of main manifestations of solar activity and its non-stability (sunspots number, flares frequency (with different phase patterns for different amplitudes), coronal holes, chromosphere emission and more). When this non-stability overlaps on non-stability of the global atmospheric and oceanic circulation, it leads to complex and non-stable response of the earth weather and climate on solar activity. We discuss this non-universal character of STC and its sequences for identification of solar variability in atmospheric, agriculture response and wheat price dynamics. We propose list of necessary conditions and possible scenarios of SCT taking in account the non-universality of SCT. Other group of hidden factors prevented reliable forecasting of solar flares impact is uncertainty of magnetic structure topology above region of flare energy release. This factor limits our ability to predict stability of magnetic trap saved hot plasma and accelerated particles. This part of the magnetic structure is responsible for runaway mode of post-flare coronal mass ejection and solar cosmic ray protons into solar wind. The reason of this uncertainty is difficulties with restoration of magnetic field topology above active region with taking in account of its helicity and fine structure. These hidden factors naturally cause of low efficiency of used forecasting methods and high level of false alerts in practice. In this situation short time forecasting of possible impact based on preliminary registration of flare itself, its amplitude and location combined with first minutes data on diffused high energy solar cosmic ray allow to forecast future radiation impact (delayed relative X-ray emission on hour-hours), and estimate expected radiation fluency and period of maximal radiation.
14	Development of MLT electron flux models	Boynton, R et al.	p-Poster
	M. A. Balikhin, S. N. Walker
	University of Sheffield
	The low energy radiation belt electrons can cause surface charging, which can interfere with satellite electronic systems. As such, forecasts of the low energy electrons can aid in the mitigation of these adverse effects. At Geosynchronous Earth Orbit, the fluxes of electrons with energies up to several hundred keV can vary widely in Magnetic Local Time. This spatial variation makes it challenging to deduce a data based model, since the spacecraft measuring the electron flux can never remain fixed at a specific MLT. This study aims to develop a set of Nonlinear AutoRegressive Moving Average eXogenous (NARMAX) models for different MLT employing electron flux data from the GOES 13 and 15 spacecraft as the outputs, and solar wind data as the inputs. This work has been supported by the European Union's Horizon 2020 grant agreement No 637302 (PROGRESS).
15	Increasing the horizon of the Sheffield GEO radiation belt electron flux forecasts.	Walker, S et al.	p-Poster
	S. N. Walker[1], T. Arber[2], K. Bennett[2], M. Liemohn[3], B. van der Holst[3], P. Wintoft[4], N. Y. Ganushkina[5], and M. A. Balikhin[1]
	[1]Automatic Control and Systems Engineering, University of Sheffield, Sheffield, UK; [2]Dept Physics, University of Warwick, Coventry, UK; [3]Climate and Space Sciences Engineering, University of Michigan, Michigan, USA; [4]Swedish Institute of Space Physics, Lund, Sweden [5]Finnish Meteorological Institute, Helsinki, Finland
	Accurate and timely forecasts for the electron environment at GEO are required in order to mitigate adverse effects and protect hardware assets. The greater the lead time, the more mitigation options are available. The Sheffield GEO electron flux forecast model is currently driven by measurements of the solar wind parameters at L1, providing a lead time of between 16 and 24 hours (depending upon particle energy). In this presentation we discuss early results based of driving the Sheffield model using forecasts of the solar wind at L1 resulting from the AWSOM/SWIFT model, developed within the Horizon 2020 project PROGRESS. Since these solar wind forecasts are based on observations of the Sun, the forecasting time horizon is further increased by 24-48 hours.
16	Forecasting the photospheric magnetic field using machine learning	Nikolic, L et al.	p-Poster
	Ljubomir Nikolic[1], Julio J. Valdes[2]
	[1]Canadian Hazards Information Service, Natural Resources Canada, Ottawa, Canada; [2]National Research Council Canada, Ottawa, Canada
	The Sun’s magnetic field is a key component that controls solar outputs and influences space weather conditions. Therefore, modelling of the solar magnetic field represents an important research topic. Modelling and forecasting the photospheric magnetic field is particularly important for forecasts such as the solar wind, flares and solar irradiance forecast. The Canadian Space Weather Forecast Center is trying to enhance its forecast capabilities using advanced numerical techniques. To support these efforts, we employ machine learning to investigate the potential of this numerical approach in forecasting the photospheric magnetic field. We use Global Oscillation Network Group (GONG) synoptic maps to train the model and to derive the field. More than 7000 maps from 2007-2016 are used for training and testing of the model. The obtained results show a good agreement between observed and derived synoptic maps. Furthermore, we demonstrate the applicability of the machine learning approach and modelled maps by deriving the F10.7 radio flux, and by forecasting the global coronal magnetic field using the potential-field source-surface model.