Specification and forecast ionospheric parameters, such as ionospheric electron density (Ne), have been an important topic in space weather and ionosphere research. Neural networks (NNs) emerge as a powerful modeling tool for Ne prediction. However, heavy manual attention costs time to determine the optimal NN structures. In this work, we propose to use neural architecture search (NAS), an automatic machine learning method, to address this problem of NN models. NAS aims to find the optimal network structure through the alternated optimization of the hyperparameters and the corresponding network parameters. A total of 16-year data from Millstone Hill incoherent scatter radar (ISR) are used for NN models. One single-layer NN (SLNN) model and one deep NN (DNN) model are trained with NAS, namely SLNN-NAS and DNN-NAS, for Ne prediction and compared with their counterparts without NAS from previous studies, denoted as SLNN and DNN. Our results show that SLNN-NAS and DNN-NAS outperformed SLNN and DNN, respectively. NN models can reveal more finer details than the empirical ionospheric model developed using traditional data fitting approaches. DNN-NAS yields the best prediction accuracy measured by quantitative metrics and rankings of daily pattern prediction. The limited improvement of NAS is likely due to the network complexity and the limitation of fully connected NN without a memory mechanism.

Volcanic eruptions provide broad spectral forcing to the atmosphere and many previous studies have examined the IT disturbances caused by volcanic eruptions through both observations and modeling. Understanding the primary mechanisms that are relevant to explain the variety in waveform characteristics is still an important open question for the community. In this study, Global Navigation Satellite System (GNSS) Total Electron Content (TEC) data are analyzed and compared to simulations performed by the Global Ionosphere-Thermosphere Model with Local Mesh Refinement (GITM-R) for the first phase of the 2015 Calbuco eruption that occurred on 22 April. A simplified source representation and spectral acoustic-gravity wave (AGW) propagation model are used to specify the perturbation at the lower boundary of GITM-R at 100 km altitude. This modeling specification shows a good agreement with GNSS observations for some waveform characteristics such as travel/onset times and relative magnitudes. Most notably, GITM-R is able to reproduce the significance of AGWs as a function of radial distance from the vent, showing acoustic dominant forcing in the near field (<500 km) and gravity dominant forcing in the far-field (>500 km). The estimated apparent phase speeds from GITM-R simulations are consistent with observations with ~10% difference from observation for both acoustic wave packets and a trailing gravity mode. Relevance of the simplifications made in the lower atmosphere are then discussed and test changes to the assumed propagation structure, from direct propagation to ground-coupled propagation, show some improvement to the data-model comparison, especially the second acoustic wave-packet

The impacts of solar eclipses on the ionosphere-thermosphere system particularly the composition, density, and transport are studied using numerical simulation and subsequent model-data comparison. We introduce a model of a solar eclipse mask (shadow) at Extreme Ultra Violet (EUV) wavelengths that computes the corresponding shadowing as a function of space, time, and wavelength of the input solar image. The current model includes interfaces for Solar Dynamics Observatory (SDO) and Geostationary Operational Environmental Satellites (GOES) EUV telescopes providing solar images at nine different wavelengths. We show the significance of the EUV eclipse shadow spatial variability and that it varies significantly with wavelength owing to the highly variable solar coronal emissions. We demonstrate geometrical differences between the EUV eclipse shadow compared to a geometrically symmetric simplification revealing changes in occultation vary $\pm$20\%. The EUV eclipse mask is validated with in-situ solar flux measurements by the PROBA2/LYRA instrument suite showing the model captures the morphology and amplitudes of transient variability while the modeled gradients are slower. The effects of spatially EUV eclipse masks are investigated with Global Ionosphere Thermosphere Model (GITM) for the 21 August 2017 eclipse. The results reveal that the modeled EUV eclipse mask, in comparison with the geometrically symmetric approximation, causes changes in the Total Electron Content (TEC) in order of $\pm$20\%, 5-20\% in F-region plasma drift, and 20-30\% in F-region neutral winds.

In this study, a new high-latitude empirical model is introduced, named for Auroral energy Spectrum and High-Latitude Electric field variabilitY (ASHLEY). This model aims to improve specifications of soft electron precipitations and electric field variability that are not well represented in existing high-latitude empirical models. ASHLEY consists of three components, ASHLEY-A, ASHLEY-E and ASHLEY-Evar, which are developed based on the electron precipitation and bulk ion drift measurements from the Defense Meteorological Satellite Program (DMSP) satellites during the most recent solar cycle. On the one hand, unlike most existing high-latitude electron precipitation models, which have assumptions about the energy spectrum of incident electrons, the electron precipitation component of ASHLEY, ASHLEY-A, provides the differential energy fluxes in the 19 DMSP energy channels under different geophysical conditions without making any assumptions about the energy spectrum. It has been found that the relaxation of spectral assumptions significantly improves soft electron precipitation specifications with respect to a Maxwellian spectrum (up to several orders of magnitude). On the other hand, ASHLEY provides consistent mean electric field and electric field variability under different geophysical conditions by ASHLEY-E and ASHLEY-Evar components, respectively. This is different from most existing electric field models which only focus on the large-scale mean electric field and ignore the electric field variability. Furthermore, the consistency between the electric field and electron precipitation is better taken into account in ASHLEY.

Solar eclipses cause profound effects on ionosphere-thermosphere dynamics due to the abatement of solar Extreme Ultra Violet (EUV) irradiance. The reduced EUV flux cause relative reduction of ionospheric plasma density and temperature, and well as it reduces thermospheric temperature, and alters neutral winds. Numerical simulations are used to understand and characterize the ionosphere-thermosphere response to solar eclipses and to compare the model results with observations. The models have traditionally implemented simplified solar eclipses, assuming spherically symmetric models with the maximum eclipse (obscuration) set to ~15%. We present a realistic model of solar eclipses, using Solar Dynamic Observatory (SDO) Atmospheric Imaging Assembly (AIA) images of the solar corona. This model computes the eclipse occultation factors as a function of geolocation and time for a chosen SDO AIA wavelength. The model includes an interface to retrieve raw high-resolution SDO AIA, the model includes horizon computation for a smooth and accurate transition at the terminators. The model is 100% pythonic, featuring parallel execution. We present observations and numerical simulations of the ionosphere-thermosphere system bolstering the importance of the accurate EUV eclipse description. We present use 21 August 2017, and 10 June 2021 solar eclipses as examples to show the effects of realistic EUV flux and transient gradients within the penumbra, and compare it with simulations using symmetric penumbra. We integrated the EUV penumbra in the Global Ionosphere Thermosphere Model (GITM), and show that the difference between EUV and symmetric eclipse amounts to as much as plus-minus 1 TECu.

In this study, field-aligned currents (FACs) obtained from the Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE) dataset have been used to specify high-latitude electric potential in the Global Ionosphere Thermosphere Model (GITM). The advantages and challenges of the FAC-driven simulation are investigated based on a series of numerical experiments and data-model comparisons for the 2013 St Patrick’s Day geomagnetic storm. It is found that the cross-track ion drift measured by the Defense Meteorological Satellite Program (DMSP) satellites can be well reproduced in the FAC-driven simulation when the electron precipitation pattern obtained from Assimilative Mapping of Ionospheric Electrodynamics (AMIE) technique is used in GITM. It is also found that properly including the neutral wind dynamo is very important when using FACs to derive the high-latitude electric field. Without the neutral wind dynamo, the cross-polar-cap potential and hemispheric integrated Joule heating could be underestimated by more than 20%. Moreover, the FAC-driven simulation is able to well reproduce the ionospheric response to the geomagnetic storm in the American sector. However, the FAC-driven simulation yields relatively larger data-model discrepancies compared to the AMIE-driven GITM simulation. This may result from inaccurate Joule heating estimations in the FAC-driven simulation caused by the inconsistency between the FAC and electron precipitation patterns. This study indicates that the FAC-driven technique could be a useful tool for studying the coupled ionosphere and thermosphere system provided that the FACs and electron precipitation patterns can be accurately specified.