skip to main content


Title: Automatic Spread‐F Detection Using Deep Learning
Abstract

Spread‐F (SF) is a feature that can be visually observed on ionograms when the ionosonde signals are significantly impacted by plasma irregularities in the ionosphere. Depending on the scale of the plasma irregularities, radio waves of different frequencies are impacted differently when the signals pass through the ionosphere. An automated method for detecting SF in ionograms is presented in this study. Through detecting the existence of SF in ionograms, we can help identify instances of plasma irregularities that are potentially affecting the high‐frequency radio‐wave systems. The ionogram images from Jicamarca observatory in Peru, during the years 2008–2019, are used in this study. Three machine learning approaches have been carried out: supervised learning using Support Vector Machines, and two neural network‐based learning methods: autoencoder and transfer learning. Of these three methods, the transfer learning approach, which uses convolutional neural network architectures, demonstrates the best performance. The best existing architecture that is suitable for this problem appears to be the ResNet50. With respect to the training epoch number, the ResNet50 showed the greatest change in the metric values for the key metrics that we were tracking. Furthermore, on a test set of 2050 ionograms, the model based on the ResNet50 architecture provides an accuracy of 89%, recall of 87%, precision of 95%, as well as Area Under the Curve of 96%. The work also provides a labeled data set of around 28,000 ionograms, which is extremely useful for the community for future machine learning studies.

 
more » « less
Award ID(s):
2028032
NSF-PAR ID:
10445351
Author(s) / Creator(s):
 ;  ;  ;  
Publisher / Repository:
DOI PREFIX: 10.1029
Date Published:
Journal Name:
Radio Science
Volume:
57
Issue:
5
ISSN:
0048-6604
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    The Super Dual Auroral Radar Network (SuperDARN) is a network of High Frequency (HF) radars that are typically used for monitoring plasma convection in the Earth's ionosphere. A majority of SuperDARN backscatter can broadly be divided into three categories: (a) ionospheric scatter due to reflections from plasma irregularities in the E and F regions of the ionosphere, (b) ground scatter caused by reflections from the ground/sea surface following reflection in the ionosphere, and (c) backscatter from meteor trails left by meteoroids as they enter the Earth's atmosphere. Due to the complex nature of HF propagation and mid‐latitude electrodynamics, it is often not straightforward to distinguish between different modes of backscatter observed by SuperDARN. In this study, we present a new two‐stage machine learning algorithm for identifying different backscatter modes in SuperDARN data. In the first stage, a neural network that “mimics” ray‐tracing is used to predict the probability of ionospheric and ground scatter occurring at a given location along with parameters like the elevation angles, reflection heights etc. The inputs to the network include parameters that control HF propagation, such as signal frequency, season, UT time, and geomagnetic activity levels. In the second stage, the output probabilities from the neural network and actual SuperDARN data are clustered together to determine the category of the backscatter. Our model can distinguish between meteor scatter, 1/2 hop E‐/F‐region ionospheric as well as ground/sea scatter. We validate our model by comparing predicted elevation angles with those measured at a SuperDARN radar.

     
    more » « less
  2. This paper describes our pathloss prediction system submitted to the ICASSP 2023 First Pathloss Radio Map Prediction Challenge. We describe the architecture of PMNet, a neural network we specifically designed for pathloss prediction. Moreover, to enhance the prediction performance, we apply several machine learning techniques, including data augmentation, fine-tuning, and optimization of the network architecture. Our system achieves an RMSE of 0.02569 on the provided RadioMap3Dseer dataset, and 0.0383 on the challenge test set, placing it in the 1st rank of the challenge. 
    more » « less
  3. This research work explores different machine learning techniques for recognizing the existence of rapport between two people engaged in a conversation, based on their facial expressions. First using artificially generated pairs of correlated data signals, a coupled gated recurrent unit (cGRU) neural network is developed to measure the extent of similarity between the temporal evolution of pairs of time-series signals. By pre-selecting their covariance values (between 0.1 and 1.0), pairs of coupled sequences are generated. Using the developed cGRU architecture, this covariance between the signals is successfully recovered. Using this and various other coupled architectures, tests for rapport (measured by the extent of mirroring and mimicking of behaviors) are conducted on real-life datasets. On fifty-nine (N = 59) pairs of interactants in an interview setting, a transformer based coupled architecture performs the best in determining the existence of rapport. To test for generalization, the models were applied on never-been-seen data collected 14 years prior, also to predict the existence of rapport. The coupled transformer model again performed the best for this transfer learning task, determining which pairs of interactants had rapport and which did not. The experiments and results demonstrate the advantages of coupled architectures for predicting an interactional process such as rapport, even in the presence of limited data. 
    more » « less
  4. Abstract

    This paper uses a regional simulation of plasma convective instability in the postsunset equatorial ionosphere together with a global atmosphere/ionosphere/plasmasphere GCM (WAM‐IPE) to forecast irregularities associated with equatorial spreadF(ESF) for 1–2 hr after sunset. First, the regional simulation is initialized and forced using ionosphere state parameters derived from campaign data from the Jicamarca Radio Observatory and from empirical models. The irregularities produced by these simulations are found to be quantitatively similar to those observed. Next, the aforementioned state parameters are replaced with parameters from WAM‐IPE, and the resulting departures between the simulated and observed irregularities are noted. In one of five cases, the forecast failed to accurately predict ESF irregularities due to the late reversal of the zonal thermospheric winds. In four of five cases, significant differences between the observed and predicted prereversal enhancement (PRE) of the background vertical drifts resulted in degraded forecast accuracy. This highlights the need for improved PRE forecasting in the global‐scale model.

     
    more » « less
  5. Abstract

    Profiles of the electron number density in the ionosphere are observed at the Arecibo Radio Observatory in Puerto Rico on a regular basis. Here, we report on recent observations showing anomalous irregularities in the density profiles at altitudes >~300 km. The irregularities occurred during a period of “mid-latitude spreadF,” a space-weather phenomenon relatively common at middle latitudes in summer months characterized by instability and electron density irregularities in the bottomside of the ionosphericFlayer. Remarkably, electron density irregularities extended well above the layer, through the ionization peak and into the topside which is regarded as being stable. Neither the neutral atmosphere nor the ionosphere is thought to be able to support turbulence locally at this altitude. A numerical simulation is used to illustrate how a combination of atmospheric and plasma dynamics driven at lower altitudes could explain the phenomenon.

     
    more » « less