Previous Abstract Return to Session F2 Next Abstract

Session F2: Atmospheric Effects on GNSS

GNSS Anomaly Detection and Direction of Arrival Estimates of Acoustic-Gravity Waves in the Ionosphere
Fiona Luhrmann and Jihye Park, Oregon State University
Date/Time: Wednesday, Sep. 18, 2:12 p.m.

Acoustic gravity waves (AGW) perturb the ionosphere as they propagate through. These perturbations can be detected with Global Navigation Satellite Systems (GNSS) carrier phase signal differences to measure the total electron content (TEC). Previous work has shown the success of a near real time (NRT) deep learning based anomaly detection algorithms to identify the occurrence of an AGW perturbation caused by an earthquake event. This research expands on this work to identify both the occurrence of an AGW caused perturbation, earthquake, or other source, and estimate the direction of arrival (DOA) of the wave. This will require a novel application of both anomaly detection and DOA algorithms in the application of ionospheric analysis. The launch of the Jet Propulsion Lab’s GUARDIAN project makes NRT TEC measurements possible providing the opportunity to apply an AGW automated detection and DOA tool as an extension to existing early warning systems.
Our previous research implemented a Long Short-Term Model (LSTM) deep learning neural network trained on typical ionospheric behavior. The LSTM method will indicate through error thresholding and error count when an anomalous ionospheric behavior has been detected. This anomalous behavior is compared against other neighboring flagged measurements to confirm a synchronous and anomalous wave event has occurred in a localized region. This algorithm has been shown to be successful in a simulated NRT data feed using GUARDIAN TEC measurements in detecting AGWs generated from the 2023 Kahramanmara? Earthquake Sequence in Türkiye.
DOA estimation has been extensively used and researched in civilian and military applications for tracking signal sources. Examples include the use of sonar, search and rescue, and seismology. Traditional DOA estimation methods are based on a physical array structure which can lead to limitations in possible applications. More recent DOA estimation methods have begun using deep learning neural networks to expand on traditional methods and applications. We propose a novel DOA method using the line-of-sight GNSS TEC measurements as our array structure to estimate the DOA of the AGW and its source(s). Now, in addition to AGW detection, information on the direction of the AGW can provide a spatial understanding of its propagation and source location.
The combination of the anomaly detection and DOA estimation algorithms will be trained and tested on the TEC measurements available online by GUARIDAN and the NASA Crustal Dynamics Data Information System (CDDIS). The GUARDIAN system tracks a minimum of 90 stations around the world with each station tracking at least two and at most four constellations leading to 60-100 links tracked every day per station. This large amount of data must be managed when training and testing. We rely on the assumption that the distribution of the data between training and test sets between near-by stations and within satellite constellations is not sufficiently different to warrant domain adaptation techniques. This assumption allows us to avoid training a model for each of the 90 plus stations in the GUARDIAN system. Instead, we can define a subset of stations evenly distributed across our experiment site on which we can train typical, regional ionospheric behavior prediction models. It is important to note that the regional aspect of this application is imperative. Typical ionospheric behavior is region-specific and therefore we must consider each station subset as a representative of typical ionospheric behavior for its region. This also allows scalability as more station data is added to the project over time. Of primary interest for this research will be the January 1st, 2024 magnitude 7.6 earthquake and tsunami at Noto Peninsula, Japan. Other point source events that induce AGW propagation into the ionosphere such as severe thunder and lightning storms are also of interest. We are limited in scope by the launch of GUARDIAN, therefore we focus on events in year 2023 and onwards. We intentionally leave the AGW source broad, as the anomaly detection and DOA algorithms are intended to be able to capture a variety of source events.
The goal of this research is to show a successful application of a NRT ionospheric anomaly detection and DOA estimate of an AGW caused by a natural hazard or other AGW producing event using the GUARDIAN TEC data. A successful experiment will confirm the use of NRT GNSS TEC measurements of the ionosphere to bolster existing early warning systems, especially those where GNSS coverage is more extensive than other monitoring infrastructure such as over open ocean or remote land regions. This application is intended to capture a wide variety of AGW source events without the use of a dense array of GNSS station data, proving to be a robust, scalable, and flexible tool to best perform in a dynamic environment.

Previous Abstract Return to Session F2 Next Abstract