東華大學圖書館 |

First-Principle Modeling and Machine Learning for Space Weather Forecasting.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	First-Principle Modeling and Machine Learning for Space Weather Forecasting./
作者:	Wang, Xiantong.
面頁冊數:	1 online resource (150 pages)
附註:	Source: Dissertations Abstracts International, Volume: 84-04, Section: B.
Contained By:	Dissertations Abstracts International84-04B.
標題:	Astrophysics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29730448click for full text (PQDT)
ISBN:	9798845469281

First-Principle Modeling and Machine Learning for Space Weather Forecasting.
Wang, Xiantong.

First-Principle Modeling and Machine Learning for Space Weather Forecasting. - 1 online resource (150 pages)

Source: Dissertations Abstracts International, Volume: 84-04, Section: B.

Thesis (Ph.D.)--University of Michigan, 2022.

Includes bibliographical references

Space weather is becoming a topic that has attracted increasing attention during the past few decades. The increase of human activities in space makes it critical to understand space weather events better. This dissertation applies a novel first-principle model to investigate the multi-scale physics in the Earth magnetosphere under strong solar wind driving conditions that have geomagnetic impacts and a machine learning model to perform solar flare forecasting related to the energy source of the space weather events.I perform a geomagnetic event simulation using a newly developed magnetohydrodynamic with adaptively embedded particle-in-cell (MHD-AEPIC) model, the first global geomagnetic storm simulation containing kinetic physics. I have developed effective criteria for identifying reconnection sites in the magnetotail and covering them with the PIC model. I compare the MHD-AEPIC simulation results with Hall MHD and ideal MHD simulations to study the impacts of kinetic reconnection at multiple physical scales. Three models produce very similar global scale features such as SYM-H, SuperMag Electrojet (SME) indexes, polar cap potentials, and field-aligned currents. They also produce good agreements with in-situ Geotail observations at the mesoscale. At the kinetic scale, the MHD-AEPIC simulation can produce a crescent shape distribution of the electron velocity space at the electron diffusion region, which agrees very well with Magnetospheric Multiscale (MMS) satellite observations. The MHD-AEPIC model compares well with observations at all scales, it works robustly, and the computational cost is acceptable due to the adaptive adjustment of the PIC domain.I investigate a kinetic physics mechanism in the magnetotail to induce sawtooth oscillations. The simulation results of our global MHD model with local kinetic physics show that when the total magnetic flux from the solar wind exceeds a threshold, sawtooth-like magnetospheric oscillations can be generated even without time-varying ionospheric outflow. The period of the oscillations varies from 1.5 to 3 hours, in good agreement with observations. The amplitude of the oscillations measured in the local Bz field also agrees well with observations at 8 RE distance from the center of Earth. The simulated oscillations cover a wide range of local times, similar to observations, although the distribution of magnitude as a function of local time is somewhat different from the observed distribution. This work suggests that kinetic reconnection physics in the magnetotail can be a major contributing factor to magnetospheric sawtooth oscillations.I implemented a deep learning network using Long-Short Term Memory (LSTM) to predict whether a solar active region (AR) will produce a flare of class Γ in the next 24 hours. The essence of using LSTM, a recurrent neural network, is its capability to capture temporal information of the data samples. The input features are time sequences of 20 magnetic parameters from the Space-weather HMI Active Region Patches (SHARPs). I analyze active regions from June 2010 to Dec 2018 and their associated flares identified in the GOES X-ray flare catalogs. The results (i) produce skill scores consistent with recently published results using LSTMs and are better than the previous results using single time input. (ii) The skill scores from the model show statistically significant variation when different years of data are chosen for training and testing. In particular, the years 2015 to 2018 have better True Skill Statistic (TSS) and Heidke Skill Scores (HSS) for predicting ≥ C medium flares than 2011 to 2014.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798845469281Subjects--Topical Terms:

535904
Astrophysics.
Subjects--Index Terms:

Space weatherIndex Terms--Genre/Form:

542853
Electronic books.

First-Principle Modeling and Machine Learning for Space Weather Forecasting.
LDR:05049nmm a2200409K 4500 001 2355717
005 20230523083132.5
006 m o d
007 cr mn ---uuuuu
008 241011s2022 xx obm 000 0 eng d
020 $a 9798845469281
035 $a (MiAaPQ)AAI29730448
035 $a (MiAaPQ)umichrackham004580
035 $a AAI29730448
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Wang, Xiantong. $3 3696157
245 1 0 $a First-Principle Modeling and Machine Learning for Space Weather Forecasting.
264 0 $c 2022
300 $a 1 online resource (150 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 84-04, Section: B.
500 $a Advisor: Toth, Gabor.
502 $a Thesis (Ph.D.)--University of Michigan, 2022.
504 $a Includes bibliographical references
520 $a Space weather is becoming a topic that has attracted increasing attention during the past few decades. The increase of human activities in space makes it critical to understand space weather events better. This dissertation applies a novel first-principle model to investigate the multi-scale physics in the Earth magnetosphere under strong solar wind driving conditions that have geomagnetic impacts and a machine learning model to perform solar flare forecasting related to the energy source of the space weather events.I perform a geomagnetic event simulation using a newly developed magnetohydrodynamic with adaptively embedded particle-in-cell (MHD-AEPIC) model, the first global geomagnetic storm simulation containing kinetic physics. I have developed effective criteria for identifying reconnection sites in the magnetotail and covering them with the PIC model. I compare the MHD-AEPIC simulation results with Hall MHD and ideal MHD simulations to study the impacts of kinetic reconnection at multiple physical scales. Three models produce very similar global scale features such as SYM-H, SuperMag Electrojet (SME) indexes, polar cap potentials, and field-aligned currents. They also produce good agreements with in-situ Geotail observations at the mesoscale. At the kinetic scale, the MHD-AEPIC simulation can produce a crescent shape distribution of the electron velocity space at the electron diffusion region, which agrees very well with Magnetospheric Multiscale (MMS) satellite observations. The MHD-AEPIC model compares well with observations at all scales, it works robustly, and the computational cost is acceptable due to the adaptive adjustment of the PIC domain.I investigate a kinetic physics mechanism in the magnetotail to induce sawtooth oscillations. The simulation results of our global MHD model with local kinetic physics show that when the total magnetic flux from the solar wind exceeds a threshold, sawtooth-like magnetospheric oscillations can be generated even without time-varying ionospheric outflow. The period of the oscillations varies from 1.5 to 3 hours, in good agreement with observations. The amplitude of the oscillations measured in the local Bz field also agrees well with observations at 8 RE distance from the center of Earth. The simulated oscillations cover a wide range of local times, similar to observations, although the distribution of magnitude as a function of local time is somewhat different from the observed distribution. This work suggests that kinetic reconnection physics in the magnetotail can be a major contributing factor to magnetospheric sawtooth oscillations.I implemented a deep learning network using Long-Short Term Memory (LSTM) to predict whether a solar active region (AR) will produce a flare of class Γ in the next 24 hours. The essence of using LSTM, a recurrent neural network, is its capability to capture temporal information of the data samples. The input features are time sequences of 20 magnetic parameters from the Space-weather HMI Active Region Patches (SHARPs). I analyze active regions from June 2010 to Dec 2018 and their associated flares identified in the GOES X-ray flare catalogs. The results (i) produce skill scores consistent with recently published results using LSTMs and are better than the previous results using single time input. (ii) The skill scores from the model show statistically significant variation when different years of data are chosen for training and testing. In particular, the years 2015 to 2018 have better True Skill Statistic (TSS) and Heidke Skill Scores (HSS) for predicting ≥ C medium flares than 2011 to 2014.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Astrophysics. $3 535904
650 4 $a Applied mathematics. $3 2122814
650 4 $a Astronomy. $3 517668
650 4 $a Computer science. $3 523869
653 $a Space weather
653 $a Numerical modeling
653 $a Machine learning
653 $a Solar flare prediction
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0596
690 $a 0984
690 $a 0606
690 $a 0364
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a University of Michigan. $b Climate and Space Sciences and Engineering. $3 3560685
773 0 $t Dissertations Abstracts International $g 84-04B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29730448 $z click for full text (PQDT)