Predicting space weather: Machine learning enhances GNSS signal stability

Traditional methods for detecting ionospheric scintillation rely on specialized and expensive ionospheric scintillation monitoring receivers (ISMRs). In contrast, the novel strategy introduced by the Hong Kong Polytechnic University researchers uses common geodetic GNSS receivers and a pre-trained machine learning decision tree algorithm to identify ionospheric amplitude scintillation events with high precision. This approach leverages the vast network of geodetic GNSS receivers and processes carrier-to-noise ratio (C/N0) and elevation angle data to compute an alternative scintillation index (S4c), which correlates well with the traditional S4 index used by ISMRs6. The machine learning algorithm enhances detection accuracy by distinguishing between the periodic nature of multipath effects and the irregularities of scintillation, resulting in a remarkable 99.9% detection accuracy.

The study published in Satellite Navigation uses machine learning to enhance the detection of ionospheric scintillation by employing a pre-trained machine learning decision tree algorithm. This algorithm processes the carrier-to-noise ratio (C/N0) and elevation angle data collected at 1-Hz intervals from common geodetic GNSS receivers. By leveraging the periodic nature of multipath effects, which differ from the irregularities of scintillation, the machine learning algorithm improves detection accuracy and reduces false alarms. Experimental results demonstrate that the decision tree algorithm achieves a remarkable 99.9% detection accuracy, surpassing traditional hard and semi-hard threshold methods.