the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A comparison of modeled daytime E-regions from E-PROBED and PyIRI with ionosonde observations
Abstract. While the F-region is the primary focus of many ionospheric models because it contains the peak electron density, the E-region is an important region for ionospheric conductivities and high-frequency radio propagation. This study analyzes modeled E-regions from the newly developed PyIRI and E-PROBED models. A long-term comparison of E-region predictions from E-PROBED and PyIRI with ionosonde observations is performed for three sites spanning low- (Fortaleza, Brazil), mid-(El Arenosillo, Spain), and high-latitudes (Gakona, Alaska). Modeled foE and hmE trends are compared against a combination of manually-scaled and automatically-scaled ionograms using ARTIST-5 for the period 2009–2024 for El Arenosillo and Gakona, and 2015–2024 for Fortaleza. Measured and modeled virtual heights are compared for a subset of the ionograms through the use of a numerical ray-tracer. Overall, the models showed reasonable agreement with the ionosonde observations, with solar cycle, seasonal, and diurnal trends well captured for foE. E-PROBED generally overestimates foE with Mean Absolute Relative Errors (MRAEs) peaking around 70 % at dusk, while PyIRI showed close agreement with ionosonde foE resulting in MRAE peaks around 10 %. The hmE predictions showed weaker agreement, with a 15–20 km overestimate from E-PROBED when compared against auto-scaled ionograms, and a constant hmE prediction of 110 km for all times from PyIRI. However, manually-scaled hmE estimates show close agreement with E-PROBED predictions, indicating that great care must be taken when using auto-scaled hmE. Modeled virtual heights derived from E-PROBED and PyIRI show reasonable agreement with ionosonde observations, providing confidence in altitude-integrated electron density profiles. A slight bias exists between the modeled and measured virtual heights, and the direction of the bias reverses for manual- versus auto-scaled ionograms, demonstrating that auto-scaled uncertainties are also present in the virtual height observations. Overall, these results indicate that E-PROBED and PyIRI provide reasonable E-region estimates and may be used for practical applications that require modeled E-region parameters.
- Preprint
(9179 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2025-3731', Anonymous Referee #1, 26 Sep 2025
reply
-
AC1: 'Reply on RC1', Daniel Emmons, 10 Oct 2025
reply
We sincerely thank you for your careful review and thoughtful feedback; we are grateful for your time and effort. Your review has helped refine this manuscript and we address each of your specific comments below:
- The Mean Relative Absolute Error (MRAE) values, for instance in Figure 2, are currently presented on a logarithmic scale. While such representation emphasizes the dynamic range, a linear scale would provide a more direct appreciation of error magnitudes and may be more accessible to the general readership. This reviewer recommends the authors to consider revising the figure accordingly
- We originally used the logarithmic scale such that a single colorbar could be used for both the PyIRI and E-PROBED MRAE subfigures. However, we agree that a linear scale would be more accessible for readers, and we have changed the scale for each of the MRAE figures from logarithmic to linear. A separate colorbar is now used for each subfigure.
- A salient result is the significant offset between E-PROBED and ionosonde foE near dusk. This reviewer suggests the authors to provide additional physical interpretation of this phenomenon. Possible contributing factors may include ionospheric tilts, GNSS-RO retrieval geometry, or specific assumptions intrinsic to the model. A brief discussion would provide valuable context
- This significant offset is certainly worthy of additional discussion in the manuscript. As you suggest, ionospheric tilts and other causes of horizontal density gradients near dusk such as the Appleton Anomaly with pre-reversal enhancement can impact GNSS-RO derived electron density profile estimates. As shown in Wu et al., (2023), high inclination (cross-latitude) sensors such as COSMIC-1 are more sensitive to ionospheric inhomogeneities. Since E-PROBED was derived from COSMIC-1 observations, and there are known horizontal density gradients during dusk, it stands to reason that the RO derived EDPs are likely impacted by the inhomogeneities, resulting in foE overestimates by E-PROBED near dusk. This explanation has been added to the paragraph discussing ionospheric tilts in the Discussion section.
- A central theme of the manuscript concerns the discrepancy between auto-scaled and manually scaled ionograms. In some instances (e.g., EA036 in 2009), the divergence is substantial not only relative to manually scaled values but also with respect to the modal behavior of the dataset. This raises a fundamental question regarding data reliability: in cases of large offsets, which dataset should be deemed more trustworthy? If the manually scaled values are regarded as the reference truth, the validity of long-term comparisons may be undermined, given the sparse availability of such data, largely limited to 2009. The authors acknowledge this limitation, but a more explicit statement on how future studies might reconcile these inconsistencies would strengthen the manuscript
- We completely agree that the discrepancy between manual and auto-scaled ionograms is problematic for comparing long-term hmE trends. While the foE trends between manual and auto-scaled ionograms agree, the large differences in hmE trends caused us to move the majority of the hmE figures to the Appendix along with a caution that the results must be taken with great care. Ideally, a future study with a long-term collection of manually-scaled ionograms could be performed to reanalyze the hmE trends and remove these inconsistencies. We have added explicit statements on this topic to Section 3.1 around Figure 3 as well as Section 4 (Discussion).
- In figures where foE or hmE are displayed separately for ionosondes, PyIRI, and E-PROBED (e.g., Figure 1), it would be advantageous to extract the representative median curves from each panel and combine them into an additional comparative panel. Such an approach would greatly facilitate direct comparison and reduce the cognitive load on the reader
- Thank you for the suggestion, this addition certainly makes the comparison easier for readers. We added the median curves for each of the trends (ionosonde, PyIRI, and E-PROBED) to each of the subfigures for the yearly, seasonal, and diurnal figures. The two reference trends for each subfigure are semi-transparent but visible to compare against the primary dataset and trend.
- While the manuscript adequately presents the numerical performance of both models, additional emphasis on their conceptual distinctions, namely, PyIRI as an ionosonde-driven semi-empirical model versus E-PROBED as a GNSS-RO-based climatological model, would better contextualize the observed biases and delineate the respective domains of applicability. In addition, the essential differences between PyIRI and the conventional IRI (Fortran) remain insufficiently explained. Since many readers may not be familiar with these distinctions, this reviewer suggests the authors to provide a more detailed introduction to both models
- Thank you for pointing out this omission. We have added a more detailed introduction to both models in the Introduction section, including details on the differences between PyIRI and IRI. Some of the E-PROBED introduction was moved from the Methodology section and placed in the Introduction section to improve the flow of the document.
- The comparison of results across low-, mid-, and high-latitude stations is an important strength of this work. Nevertheless, further discussion on latitude-dependent ionospheric drivers would affect the results. For example, auroral electron precipitation at high latitudes, sporadic E contamination at mid latitudes, and the equatorial electrojet at low latitudes may differentially affect the observed discrepancies between E-PROBED and ionosondes. Furthermore, GNSS-RO retrieval geometry differs across latitudes, since most GNSS satellites are not polar orbiting, potentially reducing accuracy at higher latitudes
- The contamination of both the ionosonde and GNSS-RO observations caused by ionospheric irregularities is certainly worthy of discussion as the various irregularities will produce differential uncertainties in the datasets for the different ionosonde sites and measurement techniques. The GNSS-RO retrieval geometry also increases the likelihood of encountering ionospheric irregularities due to the integrated nature of the observations that traverse large horizontal distances. We added a paragraph in the Discussion section to address these issues.
Once again, we thank the referee for the thoughtful and detailed review.
Daniel Emmons
daniel.emmons@afit.edu
Citation: https://doi.org/10.5194/egusphere-2025-3731-AC1
-
AC1: 'Reply on RC1', Daniel Emmons, 10 Oct 2025
reply
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
2,517 | 40 | 13 | 2,570 | 94 | 106 |
- HTML: 2,517
- PDF: 40
- XML: 13
- Total: 2,570
- BibTeX: 94
- EndNote: 106
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Reviewer Report for Emmons et al.
The present manuscript investigates the reproducibility of the ionospheric E-region as simulated by two recently developed models, PyIRI and E-PROBED, through comparative validation against ionosonde observations. The analysis encompasses three geographically distinct stations representative of different latitude regimes: Fortaleza, Brazil (low latitude); El Arenocillo, Spain (mid latitude); and Gakona, United States (high latitude). Long-term observational datasets spanning the interval from 2009 to 2024 were employed as the reference standard. The evaluation was conducted with respect to three fundamental parameters of the E-region: the critical frequency (foE), peak height (hmE), and virtual height.
The results indicate that both models are capable of reproducing solar cycle, seasonal, and diurnal variations in foE, with PyIRI exhibiting particularly strong consistency with observational data. In contrast, E-PROBED demonstrates a systematic overestimation of foE, most notably during dusk hours where the discrepancy becomes most pronounced. With regard to hmE, PyIRI is constrained by its constant altitude at 110 km, which makes it difficult to capture temporal or spatial variability of E-region. E-PROBED exhibits a mean overestimation of approximately 15 km, although comparison with manually scaled ionograms suggests relatively good performance. For virtual height, both models exhibit general agreement with ionosonde data, confirming the feasibility for estimating integrated electron density profiles within the E-region.
Overall, this study convincingly demonstrates both the strengths and limitations of PyIRI and E-PROBED, while simultaneously identifying issues related to automated scaling uncertainties and avenues for further model refinement. Clarification of certain aspects, particularly those related to figure presentation, the interpretation of dusk offsets, and the treatment of scaling discrepancies, would further enhance the clarity and readability of the manuscript. I therefore recommend publication subject to minor revisions after addressing the specific comments below: