Author ORCID Identifier
Date of Award
Doctor of Philosophy in Computing Sciences - (Ph.D.)
Jason T. L. Wang
Chase Qishi Wu
This dissertation addresses multiple crucial problems in space weather and climate, presenting new machine learning-based data analytics algorithms and models for tackling the problems.
First, the dissertation presents two new approaches to predicting solar flares. One approach, called DeepSun, predicts solar flares by utilizing a machine-learning-as-a-service (MLaaS) platform. The DeepSun system provides a friendly interface for Web users and an application programming interface (API) for remote programming users. It adopts an ensemble learning method that employs several machine learning algorithms to perform multiclass flare prediction. The other approach, named SolarFlareNet, forecasts the occurrence of solar flares within the next 24 to 72 hours by using a deep learning-based transformer model. This model is implemented into a fully operational near real-time flare forecasting system accessible on the Web.
Second, the dissertation presents a deep learning method, specifically a bidirectional long short-term memory (biLSTM) network, to predict if a solar active region (AR) would produce a solar energetic particle (SEP) event given that (i) the AR will produce an M- or X-class flare and a coronal mass ejection (CME) associated with the flare, or (ii) the AR will produce an M- or X-class flare regardless of whether or not the flare is associated with a CME. Experimental results demonstrate the superiority of the biLSTM network over related machine learning algorithms and its feasibility for SEP prediction.
Third, the dissertation presents multiple algorithms and models to forecast geomagnetic indices, which are used by geospace scientists to measure space storms and their activities. The algorithms and models include a graph neural network combined with bidirectional long short-term memory for predicting the SYM-H index, a transformer-based model for predicting the Kp index, and a hybrid model combining multi-head attention layers and long short-term memory with a convolutional neural network for predicting the disturbance storm time (Dst) index. These algorithms and models incorporate Bayesian inference into their learning frameworks, capable of quantifying both aleatoric (data) uncertainty and epistemic (model) uncertainty when predicting future indices.
Finally, the dissertation presents a method, named TSlnet, to reconstruct total solar irradiance (TSI). A minor change in solar irradiance can have a significant impact on the Earth's climate and atmosphere. As a result, studying and measuring solar irradiance is crucial in understanding climate change and solar variability. TSlnet reconstructs total solar irradiance by leveraging deep learning for short and long periods of time that span beyond the current physical models' data availability. It can be used to reconstruct TSI for more than 9,000 years.
All the algorithms and models presented in this dissertation are implemented into open-source software tools using Jupyter notebooks with GitHub, which are publicly available on the Web. These tools are Binder enabled and have Zenodo archive for download. The tools are integrated into a machine learning (ML) enhanced cyberinfrastructure that contains ML software and databases for advancing space weather research and education.
Abduallah, Yasser, "Machine learning-based data analytics for understanding space weather and climate" (2022). Dissertations. 1721.