Applications of Statistical Methods and Machine Learning in the Space Sciences

Editorial

16 March 2023

Editorial: Applications of statistical methods and machine learning in the space sciences

Bala Poduval

Karly M. Pitman

Olga Verkhoglyadova

and

Peter Wintoft

2,665 views

3 citations

Editors

Bala Poduval

University of New Hampshire

Karly Pitman

Space Science Institute (SSI)

Olga Verkhoglyadova

NASA Jet Propulsion Laboratory (JPL)

Peter Wintoft

Swedish Institute of Space Physics

Impact

A timeline diagram to present the problem formulation of our deep learning-based full-disk flare prediction model using bi-daily observations of full-disk line-of-sight magnetograms and prediction window of 24 h considered to label the magnetogram instances.

Methods

12 August 2022

Towards coupling full-disk and active region-based flare prediction for operational space weather forecasting

Chetraj Pandey

, 3 more and

Berkay Aydin

2,232 views

32 citations

Review

15 November 2022

Certification of machine learning algorithms for safe-life assessment of landing gear

Haroun El Mir

and

Suresh Perinpanayagam

4,189 views

5 citations

Original Research

23 June 2022

Towards the Identification and Classification of Solar Granulation Structures Using Semantic Segmentation

S. M. Díaz Castillo

, 2 more and

S. V. Berdyugina

4,232 views

4 citations

(A) Distribution of the waiting times between the detection of the geo-effective CME and the onset of the induced geomagnetic storm; the arithmetic mean (vertical dashed red line) and the best-fitting lognormal function (blue dashed curve) are also displayed. (B) E − Dst 2D distribution; the most probable Dst value and the lower Dst threshold are marked, as a function of E, as a black solid and red dashed line, respectively; the gray dashed area denotes the 68% probability interval of Dst per energy bin. (C) ΔtSYM-H−Δtρvb scatter-plot (blue open circles) fitted by a linear function (red dashed line) with correlation coefficient ρ = 0.83. Adapted from Figure 4 of Telloni et al. (2021).

Mini Review

03 June 2022

Statistical Methods Applied to Space Weather Science

Daniele Telloni

2,195 views

5 citations

Histogram of dBH/dt for all data between 1995 and 2019 (blue) and storm-only data (orange) for (A) the mid-latitude station NEW and (B) the high-latitude station YKC. (C) Correlation coefficient between different solar wind parameters and dBH/dt for all data (blue) and storm-only data (orange). The symbol indicates the average of all six stations, and bars represent the range of the individual stations.

Original Research

25 May 2022

Revisiting the Ground Magnetic Field Perturbations Challenge: A Machine Learning Perspective

Victor A. Pinto

, 5 more and

Hyunju K. Connor

3,153 views

17 citations

Histogram of spacecraft position in KSM coordinates for Saturn radius [R; (A) and (D)], local time [(B,E)] and latitude [(C,F)] for the 1-min interpolated MAG-only data used in the RNN approaches (left panels) and 10-min interpolated MAG and full particle data set used in the RF/SVM/LR approach. For the RNN approach the data is split across the overall data set (black line), training set (red), validation set (cyan) and test set (blue). For the RF approach, a validation set was not used but the time periods of the test and training sets were matched as closely as possible to the RNN data sets for the sake of comparison. The histograms are scaled relative to the total number of samples in each set (for the 1-min interpolated data: Ntrain = 504,338, Nval = 109,582, and Ntest = 109,259; for the 10-min interpolated data Ntrain = 103,665 and Ntest = 32,475).

Original Research

20 May 2022

Classification of Cassini’s Orbit Regions as Magnetosphere, Magnetosheath, and Solar Wind via Machine Learning

Kiley L. Yeakel

, 6 more and

Peter Kollmann

1,932 views

4 citations

Number of known outliers ranked in top N selections for the (A) astrophysics, (B) planetary, (C) Earth, and (D) FMNIST datasets.

Original Research

10 May 2022

Domain-Agnostic Outlier Ranking Algorithms—A Configurable Pipeline for Facilitating Outlier Detection in Scientific Datasets

Hannah R. Kerner

, 7 more and

Sakshum Kulshrestha

3,153 views

4 citations

Measurements from the four MMS spacecraft (MMS1, MMS2, MMS3, MMS4) on 11 July 2017 showing (left) the measured magnetic field B^=(Bx,By,Bz) from FGM (Russell et al., 2016) and (right) particle current density J^=(Jx,Jy,Jz) from FPI (Pollock et al., 2016) in GSE coordinates.

Original Research

09 May 2022

A New Three-Dimensional Empirical Reconstruction Model Using a Stochastic Optimization Method

Xun Zhu

, 4 more and

Roy B. Torbert

1,169 views

4 citations

LSTM and MLR density plots using the 2015 test set for all 4 stations. Subplots (A, C, E, G) show prediction results for the LSTM models of each station while subplots (B, D, F, H) show prediction results for the MLR models of each station. Further, we can see that the lines of best fit (solid red) between the LSTM predictions and observed data are much closer to the perfect prediction lines (dashed red) than those of the MLR predictions and the observed data with a 18.2% increase in the average Pearson correlation values.

Original Research

02 May 2022

Multi-Variate LSTM Prediction of Alaska Magnetometer Chain Utilizing a Coupled Model Approach

Matthew Blandin

, 6 more and

Shishir Priyadarshi

2,124 views

9 citations

Results of GM method application to HDR identification for a GIM on 27 October, 2002 (A) and on 19 July, 2001 (B). Corresponding results for the image processing approach are shown in (C,D). For 27 October, 2002, the GM method identifies 3 HDRs (A), while the image processing approach identifies 3 HDRs (C). For 19 July, 2001, the GM method identifies 5 HDRs (B), while the image processing approach identifies 6 HDRs (D).

Perspective

29 April 2022

Understanding Large-Scale Structure in Global Ionospheric Maps With Visual and Statistical Analyses

Olga Verkhoglyadova

, 1 more and

Jacob Kosberg

1,124 views

5 citations

(A,B) Ion spectral energy flux densities observed by THD in the magnetosphere and magnetosheath, respectively. The observation times are marked in the title of each subfigure. The red asterisks denote the maximum flux, and the green asterisks indicate the point at which the flux is one-tenth lower than the maximum flux.

Original Research

08 April 2022

Overshoot Structure Near the Earth’s Subsolar Magnetopause Generated by Magnetopause Motions

Xiaojian Song

, 6 more and

Xi Luo

1,391 views

0 citations

Review

14 March 2022

Recent Applications of Bayesian Methods to the Solar Corona

Iñigo Arregui

Solar coronal seismology is based on the remote diagnostics of physical conditions in the corona of the Sun by comparison between model predictions and observations of magnetohydrodynamic wave activity. Our lack of direct access to the physical systems of interest makes information incomplete and uncertain so our conclusions are at best probabilities. Bayesian inference is increasingly being employed in the area, following a general trend in the space sciences. In this paper, we first justify the use of a Bayesian probabilistic approach to seismology diagnostics of solar coronal plasmas. Then, we report on recent results that demonstrate its feasibility and advantage in applications to coronal loops, prominences and extended regions of the corona.

3,709 views

10 citations