Publications & Research

Building Atlas Grid on Peer‑Reviewed Science

The algorithms powering Atlas Grid are built on over a decade of published research in space weather forecasting, solar physics, and machine learning applications to heliophysics. Several of these methods are already deployed operationally by institutions including GFZ Potsdam and NOAA.

Dr. Fadil Inceoglu

Dr. Fadil Inceoglu

Dr. Fadil Inceoglu is a physicist and machine‑learning researcher specialising in solar and magnetospheric physics. He earned his PhD in 2015 from Aarhus University, where he investigated long‑term solar variability using cosmogenic nuclides. After positions at Aarhus University, the Leibniz Institute for Astrophysics Potsdam and GFZ Potsdam, he joined NOAA’s National Centers for Environmental Information and CIRES at the University of Colorado Boulder as a research scientist. His work focuses on developing AI algorithms for space‑weather forecasting, correcting satellite magnetometer data, and understanding the solar–terrestrial relationship.

LinkedIn Google Scholar

Operational Algorithms & Production Systems

Coronal Hole Detection (2022, ApJ)

Full citation: Inceoglu, F., Shprits, Y. Y., Heinemann, S. G., Bianco, S., 2022, Identification of Coronal Holes on AIA/SDO images using unsupervised Machine Learning, The Astrophysical Journal, 930:118.

Using unsupervised k‑means clustering on Solar Dynamics Observatory (SDO) images, this study automatically delineates coronal holes, the dark regions that spew high‑speed solar wind streams. The algorithm achieves performance comparable to complex convolutional neural networks while requiring minimal training data. Coronal hole maps generated by this method are now operational at GFZ Potsdam, providing real‑time inputs for high‑speed solar wind forecasting and feeding into OrbitAxiom's solar wind models.

Figure placeholder – to be added

Coronal hole detection overlay on SDO imagery.

DOI

XGBoost for GOES Magnetometer (2025, Frontiers AI)

Full citation: Inceoglu, F., Loto’aniu, P. T. M., 2025, Utilizing XGBoosts to correct arcjet contamination in magnetic field measurements from GOES missions, Frontiers in Artificial Intelligence, 8.

Magnetometer data from the GOES satellites are contaminated by arcjet thruster firings, which produce artificial disturbances. This paper uses XGBoost, a gradient‑boosted decision tree model, to learn the nonlinear relationship between clean GOES‑18 measurements and arcjet‑affected signals. The resulting correction algorithm outperforms the existing matrix‑based approach, especially for transient variations, and is planned to be deployed operationally to produce clean magnetic‑field data for space‑weather monitoring. OrbitAxiom ingests these measurements to ensure accurate geomagnetic indices.

Figure placeholder – to be added

Before/after correction plots showing contamination removal.

DOI

Solar Flare Transformers (2024, Frontiers)

Full citation: Donahue, K. P., Inceoglu, F., 2024, Forecasting Solar Flares with a Transformer Network, Frontiers in Astronomy & Space Science – Stellar and Solar Physics, Volume 10 – 2023, doi: 10.3389/fspas.2023.1298609.

Solar flares and the coronal mass ejections they accompany can cause satellite orbital decay and geomagnetic power outages. Building on transformer architectures, this work trains a time‑series model using parameters from NASA’s SHARPs data set to predict whether an active region will produce ≥C‑ or ≥M‑class flares within 24 hours. The attention‑based model outperforms earlier studies that used shorter data windows, providing longer lead times for operational forecasting. OrbitAxiom integrates these predictions to trigger alerts and updates ahead of energetic events.

Figure placeholder – to be added

Transformer architecture and prediction results.

DOI

AI Applications to Space Physics

CME/SEP Prediction (2018, ApJ)

Full citation: Inceoglu, F., Jeppesen, J. H., Kongstad, P., Marcano, N. J. H., Jacobsen, R. H., Karoff, C., 2018, Using machine learning methods to forecast if solar flares will be associated with CMEs and SEPs, The Astrophysical Journal, 861, 128.

Predicting which solar flares will spawn coronal mass ejections (CMEs) and solar energetic particles (SEPs) is critical for safeguarding satellites and power grids. This study applies support vector machines and multilayer perceptrons to NASA’s DONKI, GOES and SHARPs datasets and reports high skill scores (TSS≈0.92 and HSS≈0.92) for a 96‑hour forecast window. By distinguishing hazardous flares from benign ones, the model enables early warnings and targeted observations. These predictions feed into OrbitAxiom’s event prioritisation to protect assets during major eruptive events.

Figure placeholder – to be added

ROC curves and feature importance analysis.

DOI

AI for Galactic Cosmic Rays (2022, Sci Rep)

Full citation: Inceoglu, F., Pacini, A. A., Loto’aniu, P. T. M., 2022, Utilizing AI to unveil the nonlinear interplay of convection, drift, and diffusion on galactic cosmic ray modulation in the inner heliosphere, Scientific Reports, 12:20712.

Galactic cosmic rays (GCRs) are modulated by convection, drift, diffusion and adiabatic cooling as they travel through the heliosphere. Using Light Gradient Boosting Machines, this paper uncovers the nonlinear interplay among these processes and finds that heliospheric modulation cannot be explained by simple drift‑dominated minima versus diffusion‑dominated maxima. The work demonstrates AI’s ability not only to predict but to gain physical insight into complex systems, revealing dynamic behaviour across the solar cycle. OrbitAxiom incorporates these insights into its Space Weather models.

Figure placeholder – to be added

Neural network revealing physical relationships in GCR modulation.

DOI

Fundamental Solar & Heliospheric Physics

Solar Dynamo (2024, Sci Rep)

Full citation: Inceoglu, F., Arlt, R., 2024, Evaluating solar‑like behavior in turbulent alpha and Babcock–Leighton mechanisms using non‑kinematic nonlinear flux‑transport solar dynamos, Scientific Reports, 14:23425.

The Sun’s magnetic field emerges from dynamo processes combining turbulent α‑effects and Babcock–Leighton mechanisms. Simulating 30000 years of activity with a nonlinear flux‑transport dynamo, this study shows that the turbulent α mechanism reproduces the Schwabe cycle, quasi‑biennial oscillations and Gleissberg cycles and matches proxy data for grand minima and maxima. These insights inform long‑term space‑weather and space-climate forecasting and help OrbitAxiom estimate baseline solar activity.

Figure placeholder – to be added

Dynamo simulation results.

DOI

L1 Multi‑Spacecraft Coherence (2025, A&A)

Full citation: Pelkum Donahue, K., Inceoglu, F., 2025, Connecting solar wind turbulence to plasma parameters at L1 using multi‑spacecraft coherence, Astronomy & Astrophysics.

Monitoring the solar wind at the Lagrange‑1 point requires understanding how turbulence and plasma conditions vary between spacecraft. Analysing decades of ACE and Wind data, this paper computes the cross‑coherence of magnetic‑field and plasma fluctuations and uses k‑means clustering to relate coherence levels to variations in velocity and density. High coherence corresponds to stable plasma conditions, while low coherence indicates turbulent variations, guiding instrument calibration and future missions such as SWFO‑L1. OrbitAxiom leverages this multi‑spacecraft analysis to improve its interplanetary shock arrival time statistics.

Figure placeholder – to be added

Coherence analysis between spacecraft.

DOI

Solar QBO in Magnetosphere (2023, Sci Rep)

Full citation: Inceoglu, F., Loto’aniu, P. T. M., 2023, Detection of solar QBO‑like signals in earth’s magnetic field from multi‑GOES mission data, Scientific Reports, 13:19460.

The Sun exhibits quasi‑biennial oscillations (QBOs) with periods of 1.3–1.6 years. By analysing multi‑decadal GOES magnetometer data, this paper reports the first detection of QBO‑like signals in Earth’s magnetosphere, implying that solar QBOs propagate through the solar wind into the geospace environment. These findings reveal new aspects of Sun–Earth coupling and underscore the need to include QBO‑driven variations in space‑weather models. OrbitAxiom incorporates these periodic signals to refine long‑term drag forecasts and magnetospheric indices.

Figure placeholder – to be added

QBO signal detection in geomagnetic data.

DOI

Complete Publications List

A chronological list of Dr. Inceoglu’s peer‑reviewed publications is provided below, grouped by year from most recent to earliest.

2025

  • Pelkum Donahue, K., Inceoglu, F., 2025, Connecting solar wind turbulence to plasma parameters at L1 using multi‑spacecraft coherence, Astronomy & Astrophysics. DOI
  • Inceoglu, F., Loto’aniu P. T. M., 2025, Occurrence characteristics and amplitude‑frequency relationship of the Pc5 ULF waves from 3 decades of GOES data, Scientific Reports, 15, Article number: 3666. DOI
  • Inceoglu, F., Loto'aniu, P. T. M., 2025, Utilizing XGBoosts to correct arcjet contamination in magnetic‑field measurements from GOES missions, Frontiers in Artificial Intelligence, 8. DOI

2024

  • Inceoglu, F., Arlt, R., 2024, Evaluating solar‑like behaviour in turbulent alpha and Babcock–Leighton mechanisms using non‑kinematic nonlinear flux‑transport solar dynamos, Scientific Reports, 14:23425. DOI
  • Loto’aniu P. T. M., Inceoglu, F., 2024, The distribution of Pc5 ultralow‑frequency waves at geostationary orbit, The Astrophysical Journal, 969, 91. DOI
  • Inceoglu, F., 2024, Exploring solar dynamo behaviour using an annually resolved carbon‑14 compilation during multiple grand solar minima, Scientific Reports, 14:5617. DOI
  • Donahue, K. P., Inceoglu, F., 2024, Forecasting solar flares with a transformer network, Frontiers in Astronomy & Space Science – Stellar and Solar Physics, Volume 10 – 2023. DOI

2023

  • Loto’aniu, P. T. M., Davis, A., Jarvis, A., Grotenhuis, M., Rich, F. J., Califf, S., Inceoglu, F., Pacini, A., Singer, H. J., 2023, Initial on‑orbit results from the GOES‑18 spacecraft science magnetometer, Space Science Reviews, 219, 84. DOI
  • Inceoglu, F., Loto’aniu, P. T. M., 2023, Detection of solar QBO‑like signals in Earth’s magnetic field from multi‑GOES mission data, Scientific Reports, 13:19460. DOI
  • Loto’aniu, P. T. M., Mulligan, P., Johnson, L., Biesecker, D., Steenburgh, R., Pizzo, V., Inceoglu, F., Rodriguez, H., 2023, Solar sail missions for sub‑L1 sampling of the interplanetary magnetic field and plasma on the Sun–Earth line, Bulletin of the American Astronomical Society, 55(3). DOI

2022

  • Inceoglu, F., Pacini, A. A., Loto’aniu, P. T. M., 2022, Utilizing AI to unveil the nonlinear interplay of convection, drift, and diffusion on galactic cosmic‑ray modulation in the inner heliosphere, Scientific Reports, 12:20712. DOI
  • Inceoglu, F., Shprits, Y. Y., Heinemann, S. G., Bianco, S., 2022, Identification of coronal holes on AIA/SDO images using unsupervised machine learning, The Astrophysical Journal, 930:118. DOI
  • Inceoglu, F., Howe, R., Loto’aniu, P. T. M., 2022, Causal interaction between the subsurface rotation‑rate residuals and radial magnetic field at different timescales, The Astrophysical Journal, 925(2), 170. DOI

2021

  • Inceoglu, F., Loto’aniu, P. T. M., 2021, Using unsupervised and supervised machine‑learning methods to correct offset anomalies in the GOES‑16 magnetometer data, Space Weather, 19(12), e02892. DOI
  • Inceoglu, F., Howe, R., Loto’aniu, P. T. M., 2021, The QBO‑type signals in the subsurface flow fields during solar cycles 23 and 24, The Astrophysical Journal, 920:49. DOI

2020

  • Inceoglu, F., Marcano, N. J. H., Jacobsen, R. H., Karoff, C., 2020, A general overview for localizing short gamma‑ray bursts with a CubeSat mega‑constellation, Frontiers in Astronomy & Space Science – High‑Energy and Astroparticle Physics, 7, 573060. DOI

2019

  • Inceoglu, F., Simoniello, R., Arlt, R., Rempel, M., 2019, Constraining nonlinear dynamo models using quasi‑biennial oscillations from sunspot area data, Astronomy & Astrophysics, 625, A117. DOI
  • Jeppesen, J. H., Jacobsen, R. H., Inceoglu, F., Toftegaard, T. S., 2019, A cloud detection algorithm for satellite imagery based on deep learning, Remote Sensing of Environment, 229, 247–259. DOI

2018

  • Inceoglu, F., Jeppesen, J. H., Kongstad, P., Marcano, N. J. H., Jacobsen, R. H., Karoff, C., 2018, Using machine‑learning methods to forecast if solar flares will be associated with CMEs and SEPs, The Astrophysical Journal, 861, 128. DOI

2017

  • Inceoglu, F., Arlt, R., Rempel, M., 2017, The nature of grand minima and maxima from fully non‑linear flux‑transport dynamos, The Astrophysical Journal, 848, 93. DOI
  • Inceoglu, F., Simoniello, R., Knudsen, M. F., Karoff, C., 2017, Hemispheric progression of solar cycles in solar magnetic‑field data and its relation to the solar dynamo models, Astronomy & Astrophysics, 601, A51. DOI

2016

  • Karoff, C., Knudsen, M. F., De Cat, P., Bonanno, A., Fogtmann‑Schulz, A., Fu, J., Frasca, A., Inceoglu, F., Olsen, J., Zhang, Y., Hou, Y., Wang, Y., Shi, J., Zhang, W., 2016, Observational evidence for enhanced magnetic activity of superflare stars, Nature Communications, 7, 11058. DOI
  • Inceoglu, F., Knudsen, M. F., Olsen, J., Karoff, C., Herren, P.‑A., Schwikowski, M., Aldahan, A., Possnert, G., 2016, A continuous ice‑core ¹⁰Be record from Mongolian mid‑latitudes: influences of solar variability and local climate, Earth and Planetary Science Letters, 437, 47–56. DOI
  • Inceoglu, F., Simoniello, R., Knudsen, M. F., Karoff, C., Olsen, J., Turck‑Chiéze, S., 2016, On the current solar magnetic activity in the light of its behaviour during the Holocene, Solar Physics, 291, 303–315. DOI

2015

  • Inceoglu, F., Simoniello, R., Knudsen, M. F., Karoff, C., Olsen, J., Turck‑Chiéze, S., Jacobsen, B. H., 2015, Grand solar minima and maxima deduced from ¹⁰Be and ¹⁴C: magnetic dynamo configuration and polarity reversal, Astronomy & Astrophysics, 577, A20. DOI
  • Karoff, C., Inceoglu, F., Knudsen, M. F., Olsen, J., Fogtmann‑Schulz, A., 2015, The Lost Sunspot Cycle: New support from ¹⁰Be measurements, Astronomy & Astrophysics, 575, A77. DOI

2014

  • Inceoglu, F., Knudsen, M. F., Karoff, C., Olsen, J., 2014, Reconstruction of sub‑decadal changes in sunspot numbers based on the NGRIP ¹⁰Be record, Solar Physics, 289, 4377–4392. DOI
  • Inceoglu, F., Knudsen, M. F., Karoff, C., Olsen, J., 2014, Modelling the relationship between neutron counting rates and sunspot numbers using the hysteresis effect, Solar Physics, 289, 1387–1402. DOI