the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Aug 2026
Evaluating European Ground Motion Service against Local InSAR Analysis along Key Swedish Corridors
Abstract. Ground motion poses a significant threat to the safety and functionality of transport infrastructure, particularly in regions affected by landslides, subsidence, freeze–thaw processes, and large-scale tunnelling. Interferometric Synthetic Aperture Radar (InSAR) enables millimeter-level displacement monitoring over large areas and is used to support infrastructure risk assessment. Here, we evaluate the usability of two InSAR approaches, (i) Persistent Scatterer Interferometry (PSI) using displacement products from the European Ground Motion Service (EGMS), and (ii) a combined Persistent- and Distributed-Scatterer (PS+DS) phase-linking workflow across key Swedish transport corridors. Rather than a strictly controlled methodological comparison, we evaluate how differences between operational InSAR solutions translate into practical performance for infrastructure monitoring. Results show that PS+DS consistently yields five times as many measurement points while maintaining RMSE values within a similar range to EGMS (mostly below 6 mm in the Line Of Sight direction). The EGMS exhibits narrower RMSE distributions, but PS+DS achieves comparable median RMSE of the measurement points across most land-cover types and corridor environments. Both datasets capture similar displacement patterns along all corridors, with measurement points from both approaches concentrating within the linear corridor geometry. Overall, the five-fold increase in PS+DS point density offers infrastructure operators a practical pathway to finer-scale ground motion monitoring for railway maintenance and risk assessment, while demonstrating how the choice of operational InSAR processing strategy shapes the information available for Earth-observation-based infrastructure applications.
- Preprint
(2775 KB) - Metadata XML
-
Supplement
(3416 KB) - BibTeX
- EndNote
Status: open (until 29 Oct 2026)
-
RC1: 'Comment on egusphere-2026-3758', Mahdi Motagh, 09 Sep 2026
reply
-
CC1: 'Reply on RC1', Saeid Aminjafari, 10 Sep 2026
reply
Dear Prof. Motagh, thank you so much for your constructive feedback and important comments on our manuscript. We will apply all your comments in our revised version.
Citation: https://doi.org/10.5194/egusphere-2026-3758-CC1 -
AC1: 'Reply on RC1', Saeid Aminjafari, 18 Sep 2026
reply
We have uploaded our responses as a supplement, and applied your suggestions in the revised version.
-
CC1: 'Reply on RC1', Saeid Aminjafari, 10 Sep 2026
reply
-
RC2: 'Comment on egusphere-2026-3758', Andreas Piter, 07 Oct 2026
reply
Dear Saeid, dear co-authors,
thank you for submitting your interesting work. Ground motion services have received a lot of attention which highlights the importance of its validation which you are carrying out in this paper.This paper analyses and compares the results from the EGMS with a local processing using SARvey software for the specific case of linear infrastructure monitoring. Several metrics are considered for the comparison and are computed for 5 study areas in Sweden with different characteristics.
I agree with Mahdi's comments stating that the paper focuses much on consistency between the methods while differences are not sufficiently highlighted.
In addition, I have some comments on the methodological part of the paper and hope that my comments help you improve the paper.Best, Andreas
Processing:
In section 2.5, you state that you are excluding the winter acquisitions from the local processing. Does it mean that you remove the winter images from the SLC stack before starting the processing or do you remove it after SARvey processing from the final displacement time series? Both lead to particular challenges which need to be considered:
If you remove before the processing, this will result in longer temporal baselines of roughly half a year. Do the interferograms maintain coherence during that period? As you are using an SBAS interferogram network, the estimated linear displacement velocities will most likely be underestimated due to the winter gaps.
If you exclude the winter images after the SARvey processing, the results are heavily affected by the low coherent interferograms from winter acquisitions. This would result in less coherent pixels, unwrapping errors and will propagate to the whole displacement time series during interferogram network inversion. I would highly recommend to avoid this processing strategy. SARvey has not been developed nor been tested for cases with regular data gaps in winter.
Please clarify which of the two approaches you have taken and how you handle these challenges. Please also report how it is dealt with in EGMS.
The number of selected pixels is one of the main features of the comparison. You have chosen to select pixels following the approach of Mirzaee et al. (2023) which selects DS pixels after Phase Linking and PS pixels by combining Top Eigenvalue Percentage (TEP) and Amplitude Dispersion (ADI). Even though the selection of PS pixels using TEP+ADI is possible in SARvey, it is not recommended and will not be supported in a future version of SARvey. As we have demonstrated in our paper (Piter et al., 2024), TEP does not reflect the phase quality of the pixels. Hence, TEP+ADI yields an unreliable selection of pixels which has an immediate impact on your comparison with EGMS. Please reprocess without these pixels and instead rely on the Temporal Phase Coherence (TPC) approach implemented in SARvey for selecting PS pixels. From my experience, pixels below a TPC of 0.7 are more likely to be unreliable.
Overall goal of the paper:
As far as I understand, your goal is to validate the quality of the EGMS. For the comparison, you have chosen to process the same areas yourself (local processing). Which of the two methods do you consider to be more accurate and more precise - and how would you actually know? As long as there is no independent reference data (e.g. leveling, GNSS), your analysis will be a relative comparison. This is a common issue in the validation of InSAR methodology. Nevertheless, the overall goal of the paper could be one of the following:
- The local processing offers many advantages which arise from the user's flexibility during the processing. So, how far do I get with the out-of-the-shelf EGMS products which do not require any processing by the user? And what additional information do I achieve when adding a local processing?
- Does the one-(is-supposed-to)-fit-all methodology used in EGMS provide reasonable displacement results? Can it be reproduced using a different methodology? To which degree do they differ and what could be the reason?
Methodological differences EGMS vs. SARvey:
You are stating that the methods are different and cannot be compared to each other. While this is true, please provide the reader with the main concept of each method and discuss the main methodological differences. This would help the reader to understand the differences in the results and sets the experiments into context.
Displacement time series:
In your comparison, you mainly address the mean velocities. Please also provide a comparison based on the displacement time series and show a few representative time series from both EGMS and SARvey.
Metrics:
From my point of view, the selection of quality metrics could be discussed in more detail. Why did you choose the given metrics? What do they tell and, importantly, what are their limitations?
I would argue, that the distance of the points to the centreline of the infrastructure does not tell anything about the quality of the results. It is affected by the low geolocation accuracy of the pixels in Sentinel-1 images and the selection of pixels. The respective results (Fig. 6) do not show any significant difference between the datasets. I would recommend to remove this section from the analysis.
The comparison between the EGMS and SARvey based on the RMSE of the displacement time series is not yet fair. As far as I know, the pixels in the EGMS are selected if their RMSE is below 4mm (see also Fig. 4). Please consider applying the same selection approach for the pixels from SARvey to make the results comparable. Moreover, please check whether EGMS applies any temporal smoothing of the displacement time series. If so, please also filter the time series retrieved with SARvey to make it comparable. No temporal filtering is applied in SARvey which would lead to higher RMSE values compared to the (probably filtered) EGMS results.It is important for the comparison to bring the displacement results to the same local reference point. It is visible in Fig. 4b and d that they do not have the same reference point.
In my opinion, the selected study areas are relatively small. Particularly in the urban area you are subsetting the points to small corridors which makes it hard to interpret the displacement signal. Is the displacement really related to the excavation? Or is it part of a larger displacement phenomenon? I have looked into the EGMS and there are several displacement signals present in Gothenburg. Please show your displacement results in the context of its neighbourhood.
You are highlighting that "the underlying quality metrics are not identical across processing chain" in Section 2.4. Nevertheless, you apply the same coherence thresholds to both methods even though they correspond to totally different characteristics of the data. I would recommend to avoid using the "same" thresholds and instead use only the RMSE w.r.t. a displacement model for selecting the pixels at the end.
Vegetation:
You are discussing vegetated areas in multiple locations in the paper (e.g. "Moreover, Li et al. (2026) recommend treating InSAR velocities below 5 mm yr⁻¹ with particular care given InSAR's difficulties in vegetated terrain"). In my eyes, this discussion seems to be inappropriate as you are not dealing with vegetation. InSAR should not be used over vegetated areas due to decorrelation - that is clear. However, you are here working with infrastructures (railway, highway, buildings). Please remove the discussion related to vegetation.Citation:
I find the citation of the paper Aziz Zanjani (2024) misleading in two sentences:
- "These localized tunnel-corridor signals resemble construction-related subsidence patterns reported, by Aziz Zanjani et al. (2024), which showed that InSAR can capture localized settlement from large-scale construction when combined with contextual information such as construction timelines and geotechnical data."
However, the paper was not about tunnel-corridor signals.
- "Consistently, Aziz Zanjani et al. (2024) report that at national and local scales PS+DS yields significantly higher point counts along man-made infrastructure, while PSI provides a sparser but more conservative displacement field."
Please consider that the paper does not discuss any national scale processing.
Please adjust the sentences accordingly.
Reproducibility:
Please provide more details on the data and processing to make it reproducible. Which EGMS version/update/time span has been used? What is specific to the EGMS methodology that is used for Sweden? Please also highlight that your findings are not generalizable to the rest of Europe due to different processing strategies used for different areas in the EGMS.
Please provide details on the processing parameters used for SARvey (e.g. look into our paper about SARvey: Piter et al., 2026).Citation: https://doi.org/10.5194/egusphere-2026-3758-RC2
Model code and software
EGMS-vs-PS-DS-Phase-linking Saeid Aminjafari https://doi.org/10.5281/zenodo.19958548
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 222 | 95 | 82 | 399 | 61 | 94 | 80 |
- HTML: 222
- PDF: 95
- XML: 82
- Total: 399
- Supplement: 61
- BibTeX: 94
- EndNote: 80
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This is a very important and timely study evaluating the usability of publicly available EGMS data versus more advanced and adapted InSAR processing approaches for linear infrastructure monitoring. The authors have selected five case studies in Sweden and investigate how differences in the assumptions and processing strategies adopted by different InSAR approaches affect the resulting deformation estimates and their suitability for an operational railway-monitoring application.
I consider this to be a valuable and critical assessment that deserves publication. In my view, however, the manuscript still needs a revision as it does not fully communicate the importance and richness of the results the authors have obtained.
Major Comments
For example, in Fig. 2c, uplift/stability appears to be detected at the northern end of the railway, whereas the PS+DS results in Fig. 2d indicate subsidence at approximately the same location. A similar inconsistency appears in the southern part of the study area. This raises an important question: which of these deformation estimates is more reliable, and what is the possible reason for the discrepancy?
When examining the results presented in the Supplementary Material, I find several other locations where similarly interesting discrepancies occur. Since one of the main objectives of the study is to objectively assess and compare the different processing approaches, I strongly recommend that the authors make better use of these differences. Rather than focusing primarily on the consistency of the results, the manuscript would be considerably more valuable if it explicitly highlighted where the approaches disagree, investigated the possible causes of these discrepancies, and discussed their implications for operational infrastructure monitoring.
Since the paper aims to evaluate the applicability of the different InSAR products for operational railway monitoring, the results from all five cases are important for demonstrating the robustness and generality of the conclusions. I therefore recommend moving at least the key results from Cases 3–5 into the main text, while retaining additional details in the Supplementary Material if necessary.
Minor Comment
-Line 100: The authors state:
“The interferometric phase also contains contributions from topography, atmospheric delays, and noise, which different InSAR methods model and separate to isolate the ground-motion signal.”
Given that InSAR is a mature technique with a history of approximately three decades, I suggest citing some of the foundational/original publications on interferometric SAR at this point, rather than relying primarily on recent literature in 2024.
Piter, A., et al. (2026). SARvey-survey with SAR: A new Open-Source InSAR Research Software. Environmental Modelling & Software, 107004, https://doi.org/10.1016/j.envsoft.2026.107004
I hope the authors find my comments useful in improving the quality of the manuscript