TY - JOUR A1 - Adolfs, Marjolijn A1 - Hoque, Mohammed Mainul A1 - Shprits, Yuri Y. T1 - Storm-time relative total electron content modelling using machine learning techniques JF - Remote sensing N2 - Accurately predicting total electron content (TEC) during geomagnetic storms is still a challenging task for ionospheric models. In this work, a neural-network (NN)-based model is proposed which predicts relative TEC with respect to the preceding 27-day median TEC, during storm time for the European region (with longitudes 30 degrees W-50 degrees E and latitudes 32.5 degrees N-70 degrees N). The 27-day median TEC (referred to as median TEC), latitude, longitude, universal time, storm time, solar radio flux index F10.7, global storm index SYM-H and geomagnetic activity index Hp30 are used as inputs and the output of the network is the relative TEC. The relative TEC can be converted to the actual TEC knowing the median TEC. The median TEC is calculated at each grid point over the European region considering data from the last 27 days before the storm using global ionosphere maps (GIMs) from international GNSS service (IGS) sources. A storm event is defined when the storm time disturbance index Dst drops below 50 nanotesla. The model was trained with storm-time relative TEC data from the time period of 1998 until 2019 (2015 is excluded) and contains 365 storms. Unseen storm data from 33 storm events during 2015 and 2020 were used to test the model. The UQRG GIMs were used because of their high temporal resolution (15 min) compared to other products from different analysis centers. The NN-based model predictions show the seasonal behavior of the storms including positive and negative storm phases during winter and summer, respectively, and show a mixture of both phases during equinoxes. The model's performance was also compared with the Neustrelitz TEC model (NTCM) and the NN-based quiet-time TEC model, both developed at the German Aerospace Agency (DLR). The storm model has a root mean squared error (RMSE) of 3.38 TEC units (TECU), which is an improvement by 1.87 TECU compared to the NTCM, where an RMSE of 5.25 TECU was found. This improvement corresponds to a performance increase by 35.6%. The storm-time model outperforms the quiet-time model by 1.34 TECU, which corresponds to a performance increase by 28.4% from 4.72 to 3.38 TECU. The quiet-time model was trained with Carrington averaged TEC and, therefore, is ideal to be used as an input instead of the GIM derived 27-day median. We found an improvement by 0.8 TECU which corresponds to a performance increase by 17% from 4.72 to 3.92 TECU for the storm-time model using the quiet-time-model predicted TEC as an input compared to solely using the quiet-time model. KW - ionosphere KW - relative total electron content KW - geomagnetic storms KW - neural KW - networks KW - NTCM KW - European storm-time model Y1 - 2022 U6 - https://doi.org/10.3390/rs14236155 SN - 2072-4292 VL - 14 IS - 23 PB - MDPI CY - Basel ER - TY - JOUR A1 - Del Corpo, Alfredo A1 - Vellante, Massimo A1 - Zhelavskaya, Irina A1 - Shprits, Yuri Y. A1 - Heilig, Balazs A1 - Reda, Jan A1 - Pietropaolo, Ermanno A1 - Lichtenberger, Janos T1 - Study of the average ion mass of the dayside magnetospheric plasma JF - Journal of geophysical research : Space physics N2 - The investigation of heavy ions dynamics and properties in the Earth's magnetosphere is still an important field of research as they play an important role in several space weather aspects. We present a statistical survey of the average ion mass in the dayside magnetosphere made comparing plasma mass density with electron number density measurements and focusing on both spatial and geomagnetic activity dependence. Field line resonance frequency observations across the European quasi-Meridional Magnetometer Array, are used to infer the equatorial plasma mass density in the range of magnetic L-shells 1.6-6.2. The electron number density is derived from local electric field measurements made on Van Allen Probes using the Neural-network-based Upper-hybrid Resonance Determination algorithm. The analysis is conducted separately for the plasmasphere and the plasmatrough during favorable periods for which both the plasma parameters are observed simultaneously. We found that throughout the plasmasphere the average ion mass is similar or equal to 1 amu for a wide range of geomagnetic activity conditions, suggesting that the plasma mainly consist of hydrogen ions, without regard to the level of geomagnetic activity. Conversely, the plasmatrough is characterized by a variable composition, highlighting a heavy ion mass loading that increases with increasing levels of geomagnetic disturbance. During the most disturbed conditions, the average radial structure shows a broad maximum around 3-4 Earth radii, probably correlated with the accumulation of oxygen ions near the plasmapause. Those ions are mostly observed in the post-dawn and pre-dusk longitudinal sectors. KW - magnetospheric average ion mass KW - magnetospheric plasma spatial KW - distribution KW - oxygen torus KW - geomagnetic activity dependence KW - field line KW - resonances Y1 - 2022 U6 - https://doi.org/10.1029/2022JA030605 SN - 2169-9380 VL - 127 IS - 10 PB - American Geophysical Union CY - Washington, DC ER - TY - JOUR A1 - Drozdov, Alexander A1 - Allison, Hayley J. A1 - Shprits, Yuri Y. A1 - Usanova, Maria E. A1 - Saikin, Anthony A1 - Wang, Dedong T1 - Depletions of Multi-MeV Electrons and their association to Minima in Phase Space Density JF - Geophysical research letters N2 - Fast-localized electron loss, resulting from interactions with electromagnetic ion cyclotron (EMIC) waves, can produce deepening minima in phase space density (PSD) radial profiles. Here, we perform a statistical analysis of local PSD minima to quantify how readily these are associated with radiation belt depletions. The statistics of PSD minima observed over a year are compared to the Versatile Electron Radiation Belts (VERB) simulations, both including and excluding EMIC waves. The observed minima distribution can only be achieved in the simulation including EMIC waves, indicating their importance in the dynamics of the radiation belts. By analyzing electron flux depletions in conjunction with the observed PSD minima, we show that, in the heart of the outer radiation belt (L* < 5), on average, 53% of multi-MeV electron depletions are associated with PSD minima, demonstrating that fast localized loss by interactions with EMIC waves are a common and crucial process for ultra-relativistic electron populations. KW - radiation belts KW - EMIC KW - VERB KW - PSD Y1 - 2022 U6 - https://doi.org/10.1029/2021GL097620 SN - 0094-8276 SN - 1944-8007 VL - 49 IS - 8 PB - American Geophysical Union CY - Washington ER - TY - JOUR A1 - Guo, Yingjie A1 - Ni, Binbin A1 - Fu, Song A1 - Wang, Dedong A1 - Shprits, Yuri Y. A1 - Zhelavskaya, Irina A1 - Feng, Minghang A1 - Guo, Deyu T1 - Identification of controlling geomagnetic and solar wind factors for magnetospheric chorus intensity using feature selection techniques JF - Journal of geophysical research : A, Space physics N2 - Using over-5-year EMFISIS wave measurements from Van Allen Probes, we present a detailed survey to identify the controlling factors among the geomagnetic indices and solar wind parameters for the 1-min root mean square amplitudes of lower band chorus (LBC) and upper band chorus (UBC). A set of important features are automatically determined by feature selection techniques, namely, Random Forest and Maximum Relevancy Minimum Redundancy. Our analysis results indicate the AE index with zero-time-delay dominates the intensity evolution of LBC and UBC, consistent with the evidence that chorus waves prefer to occur and amplify during enhanced substorm periods. Regarding solar wind parameters, solar wind speed and IMF B-z are identified as the controlling factors for chorus wave intensity. Using the combination of all these important features, a predictive neural network model of chorus wave intensity is established to reconstruct the temporal variations of chorus wave intensity, for which application of Random Forest produces the overall best performance. Plain Language Summary Whistler mode chorus waves are electromagnetic waves observed in the low-density region near the geomagnetic equator outside the plasmapause. The dynamics of Earth's radiation belts are largely influenced by chorus waves owing to their dual contributions to both radiation belt electron acceleration and loss. In this study, we use feature selection techniques to identify the controlling geomagnetic and solar wind factors for magnetospheric chorus waves. Feature selection techniques implement the processes which can select the features most influential to the output. In this study, the inputs are geomagnetic indices and solar wind parameters and the output is the chorus wave intensity. The results indicate that AE index with zerotime delay dominates the chorus wave intensity. Furthermore, solar wind speed and IMF B-z are identified as the most important solar wind drivers for chorus wave intensity. On basis of the combination of all these important geomagnetic and solar wind controlling factors, we develop a neural network model of chorus wave intensity, and find that the model with the inputs identified using the Random Forest method produces the overall best performance. Y1 - 2021 U6 - https://doi.org/10.1029/2021JA029926 SN - 2169-9380 SN - 2169-9402 VL - 127 IS - 1 PB - Wiley CY - Hoboken, NJ ER - TY - JOUR A1 - Haas, Bernhard A1 - Shprits, Yuri Y. A1 - Allison, Hayley A1 - Wutzig, Michael A1 - Wang, Dedong T1 - Which parameter controls ring current electron dynamics JF - Frontiers in astronomy and space sciences N2 - Predicting the electron population of Earth's ring current during geomagnetic storms still remains a challenging task. In this work, we investigate the sensitivity of 10 keV ring current electrons to different driving processes, parameterised by the Kp index, during several moderate and intense storms. Results are validated against measurements from the Van Allen Probes satellites. Perturbing the Kp index allows us to identify the most dominant processes for moderate and intense storms respectively. We find that during moderate storms (Kp < 6) the drift velocities mostly control the behaviour of low energy electrons, while loss from wave-particle interactions is the most critical parameter for quantifying the evolution of intense storms (Kp > 6). Perturbations of the Kp index used to drive the boundary conditions at GEO and set the plasmapause location only show a minimal effect on simulation results over a limited L range. It is further shown that the flux at L & SIM; 3 is more sensitive to changes in the Kp index compared to higher L shells, making it a good proxy for validating the source-loss balance of a ring current model. KW - ring current KW - magnetosphere KW - electron lifetimes KW - electrons KW - van allen probes (RBSP) KW - ring current model KW - verb Y1 - 2022 U6 - https://doi.org/10.3389/fspas.2022.911002 SN - 2296-987X VL - 9 PB - Frontiers Media CY - Lausanne ER - TY - JOUR A1 - Inceoglu, Fadil A1 - Shprits, Yuri Y. A1 - Heinemann, Stephan G. A1 - Bianco, Stefano T1 - Identification of coronal holes on AIA/SDO images using unsupervised machine learning JF - The astrophysical journal : an international review of spectroscopy and astronomical physics N2 - Through its magnetic activity, the Sun governs the conditions in Earth's vicinity, creating space weather events, which have drastic effects on our space- and ground-based technology. One of the most important solar magnetic features creating the space weather is the solar wind that originates from the coronal holes (CHs). The identification of the CHs on the Sun as one of the source regions of the solar wind is therefore crucial to achieve predictive capabilities. In this study, we used an unsupervised machine-learning method, k-means, to pixel-wise cluster the passband images of the Sun taken by the Atmospheric Imaging Assembly on the Solar Dynamics Observatory in 171, 193, and 211 angstrom in different combinations. Our results show that the pixel-wise k-means clustering together with systematic pre- and postprocessing steps provides compatible results with those from complex methods, such as convolutional neural networks. More importantly, our study shows that there is a need for a CH database where a consensus about the CH boundaries is reached by observers independently. This database then can be used as the "ground truth," when using a supervised method or just to evaluate the goodness of the models. Y1 - 2022 U6 - https://doi.org/10.3847/1538-4357/ac5f43 SN - 1538-4357 VL - 930 IS - 2 PB - IOP Publ. Ltd. CY - Bristol ER - TY - JOUR A1 - Landis, Daji August A1 - Saikin, Anthony A1 - Zhelavskaya, Irina A1 - Drozdov, Alexander A1 - Aseev, Nikita A1 - Shprits, Yuri Y. A1 - Pfitzer, Maximilian F. A1 - Smirnov, Artem G. T1 - NARX Neural Network Derivations of the Outer Boundary Radiation Belt Electron Flux JF - Space Weather: the international journal of research and applications N2 - We present two new empirical models of radiation belt electron flux at geostationary orbit. GOES-15 measurements of 0.8 MeV electrons were used to train a Nonlinear Autoregressive with Exogenous input (NARX) neural network for both modeling GOES-15 flux values and an upper boundary condition scaling factor (BF). The GOES-15 flux model utilizes an input and feedback delay of 2 and 2 time steps (i.e., 5 min time steps) with the most efficient number of hidden layers set to 10. Magnetic local time, Dst, Kp, solar wind dynamic pressure, AE, and solar wind velocity were found to perform as predicative indicators of GOES-15 flux and therefore were used as the exogenous inputs. The NARX-derived upper boundary condition scaling factor was used in conjunction with the Versatile Electron Radiation Belt (VERB) code to produce reconstructions of the radiation belts during the period of July-November 1990, independent of in-situ observations. Here, Kp was chosen as the sole exogenous input to be more compatible with the VERB code. This Combined Release and Radiation Effects Satellite-era reconstruction showcases the potential to use these neural network-derived boundary conditions as a method of hindcasting the historical radiation belts. This study serves as a companion paper to another recently published study on reconstructing the radiation belts during Solar Cycles 17-24 (Saikin et al., 2021, ), for which the results featured in this paper were used. KW - radiation belts KW - forecasting (1922, 4315, 7924, 7964) KW - machine learning (0555) Y1 - 2022 U6 - https://doi.org/10.1029/2021SW002774 SN - 1542-7390 VL - 20 IS - 5 PB - American Geophysical Union CY - Washington ER - TY - JOUR A1 - Long, Minyi A1 - Ni, Binbin A1 - Cao, Xing A1 - Gu, Xudong A1 - Kollmann, Peter A1 - Luo, Qiong A1 - Zhou, Ruoxian A1 - Guo, Yingjie A1 - Guo, Deyu A1 - Shprits, Yuri Y. T1 - Losses of radiation belt energetic particles by encounters with four of the inner Moons of Jupiter JF - Journal of geophysical research, Planets N2 - Based on an improved model of the moon absorption of Jovian radiation belt particles, we investigate quantitatively and comprehensively the absorption probabilities and particle lifetimes due to encounters with four of the inner moons of Jupiter (Amalthea, Thebe, Io, and Europa) inside L < 10. Our results demonstrate that the resultant average lifetimes of energetic protons and electrons vary dramatically between similar to 0.1 days and well above 1,000 days, showing a strong dependence on the particle equatorial pitch angle, kinetic energy and moon orbit. The average lifetimes of energetic protons and electrons against moon absorption are shortest for Io (i.e., similar to 0.1-10 days) and longest for Thebe (i.e., up to thousands of days), with the lifetimes in between for Europa and Amalthea. Due to the diploe tilt angle absorption effect, the average lifetimes of energetic protons and electrons vary markedly below and above alpha eq ${\alpha }_{\mathrm{e}\mathrm{q}}$ = 67 degrees. Overall, the average electron lifetimes exhibit weak pitch angle dependence, but the average proton lifetimes are strongly dependent on equatorial pitch angle. The average lifetimes of energetic protons decrease monotonically and substantially with the kinetic energy, but the average lifetimes of energetic electrons are roughly constant at energies