Jakarta, INTI - It is crucial for solar storms to be predicted accurately because they can disrupt satellites, navigation systems, radio communications, and even the Earth's electricity grid. A solar storm is a temporary disturbance in Earth's magnetic field (magnetosphere) caused by solar activity.
This phenomenon releases enormous waves of energy, charged particles, and plasma clouds into the solar system, including toward Earth. One of the main triggers of solar storms is a Coronal Mass Ejection (CME), a massive ejection of plasma and magnetic fields from the Sun's atmosphere. When a CME heads toward Earth, researchers need to know exactly when the material will arrive so that mitigation can be implemented early.
Tiar Dani, an Associate Researcher at the Research Center for Space (PRA) of the National Research and Innovation Agency (BRIN), explained an illustration to help understand this research.
"Imagine a bullet fired into water. Although it initially travels very fast, its speed will decrease due to the water's drag. A similar phenomenon occurs with CMEs. As they travel from the Sun toward Earth, a CME doesn't travel in an empty vacuum. It must penetrate a continuous stream of charged particles from the Sun, known as the solar wind. This interaction with the solar wind produces drag, causing the CME's speed to change during its journey," he said.
Tiar explained that predicting the arrival time of the CMEs is a major challenge for researchers. This is because the solar wind is highly dynamic and changing, making it difficult to accurately model using conventional physics approaches alone.
Combining Physics and AI
To overcome this challenge, Tiar and his team conducted research using an approach that combines the laws of physics and artificial intelligence (AI).
The physics model used is the Drag-Based Model (DBM), which describes the effect of solar wind drag on the movement of high-energy particles from a CME event. This model was then combined with Random Forest-based Artificial Intelligence (AI) technology trained using historical data from various CME events previously observed over two solar cycles.
"Through the learning process from this data, the AI can estimate the amount of resistance each CME would experience during its journey to Earth. In other words, this system not only understands the physics underlying CME movement but also learns from past occurrences to produce more accurate predictions of when the CME will arrive," Tiar explained.
He continued, explaining that the hybrid physics-AI DBM model was able to predict the CME transit time with an average error of around 8.7 hours. In the field of space weather, this achievement is highly competitive and demonstrates a significant improvement in predictive capability. The model produced more accurate results than the standard DBM model. This is because the standard DBM model, by default, assumes that the solar wind resistance to the movement of high-energy particles from CME is constant.
This success is an important step toward a more reliable space weather early warning system. Like having a high-tech "alarm clock," researchers can now sound the alarm before a solar storm hits Earth. With faster and more accurate information, satellite operators, communications service providers, and critical infrastructure managers can take anticipatory measures to minimize the impact.
"Through this research, the foundation for utilizing AI in the space sector in Indonesia will be even stronger. Going forward, this innovation will be a key part of our ongoing Research in AI for Space initiative. The goal is to create a fully autonomous, resilient, and continuously scalable space weather early warning framework to protect future technological infrastructure," he said.
Conclusion
BRIN is developing a hybrid model based on physics and artificial intelligence (AI) to more accurately predict the arrival time of solar storms to Earth. This model combines a Drag-Based Model (DBM), which explains the influence of solar wind resistance on the movement of Coronal Mass Ejections (CMEs), with a Random Forest-based AI trained using historical data from two solar activity cycles. This innovation is expected to be the foundation of an independent and reliable space weather early warning system to protect satellites, communications, navigation, and critical infrastructure from the impacts of solar storms.
Read more: Amazon Announces Additional USD 13 Billion Investment in India for AI and Cloud