Analysis and Predictive Modeling of Solar Coronal Holes Using Computer Vision and ARIMA-LSTM Networks (2405.09802v3)

Abstract: In the era of space exploration, coronal holes on the sun play a significant role due to their impact on satellites and aircraft through their open magnetic fields and increased solar wind emissions. This study employs computer vision techniques to detect coronal hole regions and estimate their sizes using imagery from the Solar Dynamics Observatory (SDO). Additionally, we utilize hybrid time series prediction model, specifically combination of Long Short-Term Memory (LSTM) networks and ARIMA, to analyze trends in the area of coronal holes and predict their areas across various solar regions over a span of seven days. By examining time series data, we aim to identify patterns in coronal hole behavior and understand their potential effects on space weather.

• The paper develops a framework that automates solar coronal hole detection via computer vision and employs LSTM networks for accurate area prediction.
• The methodology applies multi-filter binarization and advanced anti-noising techniques to enhance detection accuracy in AIA 193 images.
• Results validated over a seven-day period show the model’s effectiveness in forecasting solar events, aiding proactive space weather management.

Analysis and Predictive Modeling of Solar Coronal Holes Using Computer Vision and ARIMA-LSTM Networks

Introduction

This paper explores the analysis and predictive modeling of solar coronal holes, leveraging computer vision for detection and Long Short-Term Memory (LSTM) networks for predicting their area. Coronal holes—regions on the sun with open magnetic field lines and cooler temperatures—are significant due to their impact on space weather through influencing solar wind emissions. The research addresses the automation of coronal hole detection and forecasts their size over time, improving capabilities in anticipating and preparing for space weather events.

Methodology

The methodology comprises two main phases: detection and prediction.

Coronal Hole Detection using Computer Vision

Computer vision techniques are crucial in this methodology to automate the detection process, thereby reducing human error and increasing efficiency. The use of multi-filters and binarization process is applied to the original AIA 193 images to highlight coronal holes with improved accuracy.

Figure 1: A depiction of the multi-filters computer vision based detection applied to an original AIA 193 image~\cite{Nasa}.

Middle regions of the sun, located around the solar equator, are of special interest due to their direct influence on Earth's geomagnetic environment. Images are processed to binary format to distinguish coronal holes against other solar features. Following this detection, anti-noising techniques such as non-local means and Laplacian of Gaussian methods are applied to refine the images and further enhance detection accuracy.

Time Series Prediction using Deep Learning

Following detection, LSTM networks are employed for the prediction of coronal hole sizes. LSTM networks, with their ability to handle long-term dependencies in sequential data, are well-suited for this task. The deep learning model uses historical data to make future predictions about coronal hole sizes.

Results

The model is evaluated using data spanning several years, revealing the LSTM's capacity to accurately predict coronal hole areas over a seven-day period. The predictions align closely with observed data, demonstrating the model's effectiveness.

Figure 2: Comparison of actual and predicted middle area values for a 7-day period from July 31, 2021, to August 06, 2021, obtained after training the LSTM on data from January 06, 2011, to July 30, 2021.

Discussion

While the model is effective in the middle regions of the sun, future research could explore enhancements in other regions like the poles, which also significantly impact space weather. Extending the dataset to include multiple solar cycles and improving prediction models could provide even more valuable insights into solar activity.

Conclusion

This paper develops a robust framework for the detection and prediction of coronal hole areas using computer vision and LSTM networks. By improving our ability to forecast solar events, this research aids in the proactive management of space weather impacts on Earth's technological infrastructure. Continued advancements in this field are imperative to ensure the readiness and resilience of critical systems against the effects of solar activities.