Abstract
This paper proposes a methodology for automatic, accurate, and early detection of amplitude ionospheric scintillation events, based on machine learning algorithms, applied on big sets of 50 Hz postcorrelation data provided by a global navigation satellite system receiver. Experimental results on real data show that this approach can considerably improve traditional methods, reaching a detection accuracy of 98%, very close to human-driven manual classification. Moreover, the detection responsiveness is enhanced, enabling early scintillation alerts.
Highlights
The propagation through the atmosphere has a significant influence on radio signals broadcast by satellites toward the Earth
The purpose is twofold: on one side, observation of the signals themselves, which are a source of information for understanding and modeling the upper layers of the atmosphere [8]; on the other side, the signals can be used as detectors and triggers to raise warning and take countermeasures for Global Navigation Satellite System (GNSS)-based operations
The work presented in this paper aims at proposing an alternative method for the detection of amplitude scintillation based on machine learning
Summary
The work presented in this paper aims at proposing an alternative method for the detection of amplitude scintillation based on machine learning The scope of this approach is multifold as follows: 1) to propose an alternative to the use of traditional scintillation indices, the performance of which may depend on algorithmic choices, such as detrending and average operations; 2) to use only common GNSS stand-alone receivers observables; 3) to be able to understand the presence of the scintillation event including the transient time before and after its strongest phase, providing an early run-time alert; 4) to provide an automatic method, resembling manual observation of the observables, while keeping the cost low in terms of human effort and enabling run-time detection; 5) to reduce the rate of false alarms due to the ambiguity between scintillation and other events, such as multipath, that may affect the assessment of the classical amplitude scintillation index, without the need of prefiltering data; 6) to reduce the missed detection caused by a priori filtering of data at low-elevation angle, often implemented to hard-cut multipath effect; 7) to use computationally efficient machine learning algorithms, such decision tree.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
