Analysis of ionospheric structure influences on residual ionospheric errors in GNSS radio occultation bending angles based on ray tracing simulations

Abstract

The Global Navigation Satellite System (GNSS) radio occultation (RO) technique is widely used to observe the atmosphere for applications such as numerical weather prediction and global climate monitoring. The ionosphere is a major error source to RO at upper stratospheric altitudes and a linear dual-frequency bending angle correction is commonly used to remove the first-order ionospheric effect. However, the residual higher-order ionospheric error (RIE) can still be significant so that it needs to be further mitigated for high accuracy applications, especially from 30 km altitude upward where the RIE is most relevant compared to the decreasing magnitude of the atmospheric bending angle. In a previous study we quantified RIEs using an ensemble of about 700 quasi-realistic end-to-end simulated RO events, finding typical RIEs at the 0.1 to 0.5 rad noise level, but were left with 26 exceptional events with 5 anomalous RIEs at the 1 to 10 rad level that remained unexplained. In this study, we focused on investigating the causes of the high RIE of these exceptional events, employing detailed along-raypath analyses of atmospheric and ionospheric refractivities, impact parameter changes, and bending angles and RIEs under asymmetric and symmetric ionospheric structures. We found that the main causes of the high RIEs are a combination of physics-based effects, where asymmetric ionospheric conditions play the 10 primary role, more than the ionization level driven by solar activity, and technical ray tracer effects due to occasions of imperfect smoothness in ionospheric refractivity model derivatives. We also found that along-ray impact parameter variations of more than 10 to 20 m are well possible due to ionospheric asymmetries, and depending on prevailing horizontal refractivity gradients are positive or negative relative to the initial impact parameter at the GNSS transmitter. Furthermore, mesospheric RIEs are 15 found generally higher than upper stratospheric ones, likely due to being closer in tangent point heights to the ionospheric E layer peaking near 105 km, which increases RIE vulnerability. In future we will further improve the along-ray modeling system to fully isolate technical from physics-based effects and to use it beyond this work for additional GNSS RO signal propagation studies.

Convolutional Neural Networks Combined with Machine Vision for Mechanical Compressor Defect Detection

“…The ratio R shows that, after the deep separable convolution, the corresponding calculation is much smaller than that of the standard convolution method. In the deep CNN model, the gradient explosion and gradient disappearance can be solved by introducing residual learning method [22][23][24]. In the deep separable convolution, the addition of multiple point-by-point convolutions makes the computation intensive.…”

Systematic Ionospheric Residual Errors in GNSS Radio Occultation: Theory for Spherically Stratified Media

“…(1996) studied the accuracy of the bending angle correction method and suggested the additional correction of systematic residual errors by numerical simulations given a priori knowledge of the ionospheric electron density. Over the past decade, an increased number of studies have assessed the systematic residual errors in bending angle either by simulation experiments using ray tracing (Coleman & Forte, 2017; Danzer et al., 2013, 2015; Li et al., 2020; Liu et al., 2013, 2015, 2018; Mannucci et al., 2011; Qu et al., 2015) or by theoretical considerations and empirical analyses using the relation between refractivity and bending angle under the assumption of local spherical symmetry (Angling et al., 2018; Danzer et al., 2020, 2021; Fan et al., 2017; Healy & Culverwell, 2015; Liu et al., 2020). On the latter account, Healy and Culverwell (2015) derived and suggested a simple addition to the standard correction using the already measured L1 and L2 bending angles, thereby reducing the sensitivity of residual ionospheric correction to a priori knowledge, as well as reducing algorithm complexity.…”