the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 10 Jul 2026
Dataset of daily vertical displacements observed by GPS between 1994 and 2023 for hydrogeodetic studies over Europe
Abstract. Europe is currently the fastest-warming continent in the world, and it has experienced frequent and severe weather events, which have led to extensive droughts and floods, with consequences for ecosystems, health, economy, and other sectors. During the past two decades, these hydrological extremes have been quantified using Terrestrial Water Storage (TWS) changes obtained from the Gravity Recovery and Climate Experiment (GRACE) mission and its successor GRACE Follow-On. Unfortunately, GRACE/-FO-derived TWS changes do not have sufficient temporal and spatial resolutions for detailed analysis of sub-regional patterns or sub-monthly TWS changes over Europe, e.g., at the Eurostat NUTS 2 or 3 level. We suggest that both spatial and temporal resolutions could be enhanced in the future by using displacement time series observed at more than 6,000 permanent Global Positioning System (GPS) European stations. However, to turn this network into an observing system for TWS anomalies, GPS displacements must be carefully prepared in advance, and no useful dataset is available to our knowledge. Here we provide, for the first time, a quality-controlled dataset of long daily vertical displacement time series observed at 4,443 GPS antennas in Europe and surrounding regions between 1994 and 2023, after preprocessing and preselection to remove displacements seemingly unrelated to hydrospheric loading. We classify stations that pass our procedure as reference time series (benchmark datasets) with respect to hydrospheric changes. Three benchmark datasets are provided for use at different temporal scales: long-term (>1.1 years), seasonal (from 4 months to 1.4 years) and short-term (from 2 days to 5 months), even for the period of 8 years prior to GRACE (Klos and Bogusz, 2026). We show in this study that the displacements recorded by GPS stations included in the benchmark datasets (1) are to a great extent coherent with hydrological models, reflect accumulated precipitation records, and clearly reflect the influence of climate modes, (2) are mutually highly consistent on a regional scale and also consistent with the displacements determined by the InSAR (Interferometric Synthetic Aperture Radar) technique, (3) allow for the estimation of high-resolution TWS changes at all three temporal scales well, which matches closely with GRACE and ERA5-Land, potentially allowing for a better understanding of regional changes in the European hydrosphere.
- Preprint
(1836 KB) - Metadata XML
-
Supplement
(879 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2026-469', Vagner Ferreira, 08 Aug 2026
-
RC2: 'Comment on essd-2026-469', Laura Jensen, 28 Aug 2026
General comments:
The study by Klos et al. presents a European wide data set of GPS-derived vertical displacements derived by applying a benchmark algorithm to identify stations reflecting the short-term, seasonal and long-term variability of hydrological loadings. Generally, the approach seems to be solid and appropriate and the manuscript is clearly written and understandable. The authors take a great effort to validate their data set against independent other observations. The resulting benchmarked data set is indeed a valuable contribution for hydrogeodetic studies. While I find the study well suitable for a publication in ESSD, I have a few concerns that should be addressed for the final version of the manuscript (see below). Therefore, I recommend moderate revisions. Please find below my major comments.
- The authors apply the same benchmarking procedure as proposed in Klos et al. (2023), where GRACE and GLWS data were used as the benchmarking data sets. Here, they rely on another hydrological model, eCLM. However, I am wondering if the authors have evaluated the suitability of this model for representing TWS variability, or if there are any references that assess the quality of eCLM w.r.t. TWS. I assume that the benchmarking results heavily depend on the quality of the hydrological model data. Thus, I would also recommend to investigate to which extent the results are robust against using another hydrological model in the benchmarking procedure. This part of the manuscript should be extended.
- Section 3.1 discusses the results for 11 regions separately. Line 210 – 212 mentions that the regions are characterized by a similar amplitude of power-law noise and a similar spectral index, and refers to the Supplementary material for further information. However, I did not understand from that material either, how the region outlines are defined? There is only one sentence about it: “By combining both these estimates, we group the GPS stations regionally into 11 regions, …” How are the estimates combined, how is the grouping conducted?
- In Section 3.1 the reflection of droughts and floods in the three data sets are objectively described for each region. Periods of droughts and floods are marked in blue and orange in the figure, but it is not clear to me from which data source these blue/orange periods come? Is it from the literature or from an index data set? Please clarify this. Furthermore, I miss a critical discussion on the reasons why for some time periods the different data sets agree well and for some not. Specifically, it is often mentioned that some variations are not captured by SLR and DORIS, but no further conclusions are derived from that. In my opinion, the description of the individual regions could be shortened, in favor for a more extensive discussion of the overall conclusions to be drawn from this analysis.
- Line 478 – 481: “we clearly see” To be honest, I don’t clearly see the noise reduction from Fig 4. I suggest to add a direct comparison between the benchmarked and the original (full) data set for each of the components (short-term, seasonal, long-term), maybe in an additional figure. This could go along with my comment above, the usage of a different hydrological benchmarking data set. This would also further highlight the advantage of applying the benchmarking procedure and emphasize the quality of the new data set promoted in this study in comparison to using the full GPS station data.
- In the summary, I miss a few sentences about the limitations (or accuracy) of the data set.
Specific comments:
Line 25: I would I would avoid using references in the abstract. Half-sentence “even for the period of 8 years prior to GRACE” is kind of redundant to the period 1994 to 2023 mentioned four lines above. I suggest to remove it.
Line 71: “see Copernicus Climate Change Service” is a rather unspecific reference, please make it more specific to substantiate the statement.
Line 183/184: I don’t understand “summing the remaining components”. Isn’t it just Level 9 that is remaining for the long-term signal? Please clarify.
Line 189/190: I don’t understand this step. I thought that all stations with data coverage less than one year were already excluded in step 1? Why does the number of stations further reduce now?
Line 199: I was confused by the reference until I realized that this is the data publication belonging to this study. Maybe mention it in a sentence here, e.g.: “These data sets are made publicly available in Klos and Bogusz (2026).”
Figure 2: It would be nice if the map with the region outlines could be increased in size.
Line 448/449: Do you have any idea why for the long-term variability the correlations are so inhomogeneous? There are also many strong (probably significant) negative correlations right next to strong positive correlations. This phenomenon should be further discussed.
Line 510: “a heat wave” I wonder if it was just a single heat wave or more than one? If only one, please specify the year and provide a more specific reference.
Line 556 – 568: This paragraph is a bit unclear to me. “do not provide them separately” Do you mean in the publication of the data set? The following sentences about different regions: Iceland, etc, do they refer to the sub-periods or to the full period? The last sentence does not seem to be connected to the others. Furthermore, the paragraph ends quite abrupt. I suggest to integrate this paragraph into the previous paragraphs about the four periods and to end the section with a clear conclusion.
Line 561: “mismodelling of GIA effect”. I suggest to apply the same GIA model to the GRACE, GLWS and GPS data sets. Then the trend might still be affected by a suboptimal GIA model, but at least it could be excluded that the differences in the three data sets come from the GIA model.
Technical corrections:
Line 63: judgments –> judgements
Line 78: “incredible” sounds exaggerating, replace it by just “a” or a more objective word.
Line 80: a –> the
Line 127-129: Provide a reference for SoilGrids, introduce acronym ERA5 and provide proper reference, provide reference for Global Land Cover 2000 dataset.
Line 142: ones –> one
Line 156: how are the additional offsets estimated?
Line 203: why is the analysis preliminary?
Line 203: the inversion approach à an inversion approach
Line 226: This data is –> These data are
Line 289 – 292: “It is also worth noting…” This sentence is duplicated.
Line 294: Below, in the following subsections, –> In the following subsections
Line 297: Delete “now”
Lines 340 – 345: These sentences are a repetition of lines 333 – 340. Please remove and adjust the paragraph.
Line 467: perfect –> almost perfect
Line 493: Now, we –> We
Line 502: a uniform value –> a mostly uniform value
Line 529: observed –> observe
Line 549: The only difference –> The only major difference
Line 686: doi is incorrect, it must be “v1” instead of “v” at the end of the link.
Citation: https://doi.org/10.5194/essd-2026-469-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 276 | 101 | 36 | 413 | 46 | 28 | 31 |
- HTML: 276
- PDF: 101
- XML: 36
- Total: 413
- Supplement: 46
- BibTeX: 28
- EndNote: 31
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comments:
The paper entitled "Dataset of daily vertical displacements observed by GPS between 1994 and 2023 for hydrogeodetic studies over Europe" by Kłos et al. proposes a quality-controlled, three-tier (short-term, seasonal, long-term) benchmark dataset of daily GPS vertical displacements for 4,443 stations across Europe, derived through a rigorous pre-processing and wavelet-decomposition procedure, and validated against hydrological models, InSAR, SLR+DORIS, GRACE, and ERA5-Land. Overall, the dataset is valuable, as it could be used by the hydrogeodesy, remote sensing, and water-resources communities to derive high-resolution terrestrial water storage (TWS) estimates at sub-monthly and sub-regional scales that are currently unattainable from GRACE alone, and it extends the observational record by 8 years prior to the GRACE mission. However, to make the proposed dataset valuable for a wide range of users, the authors could, but are not obligated to, release the cleaned, non-decomposed daily displacement time series in addition to the three temporal-scale benchmark products to increase reusability beyond the authors' specific decomposition scheme. That said, a moderate revision is recommended, as the underlying dataset and validation are scientifically sound and represent a genuinely novel and useful contribution to ESSD. Below are some points the authors can consider in a revised version of their study.
Specific comments:
1. Sect. 2.1: Albeit the tidal (IERS2010, permanent tide retained) and non-tidal atmospheric/oceanic loading (GFZ/ECMWF-MPIOM) and GIA (ICE6G-D) corrections are clearly stated, and InSAR is explicitly tied to IGS14 (p.7, lines 247–249), the reference frame origin convention (CM, CF, or CE) used for the Love-number/Farrell (1972) forward modeling of hydrological, SLR+DORIS, GRACE, and GLWS displacements is never specified. Since NGL/IGS14 GPS coordinates are nominally CM-referenced while Farrell-type load Green's functions are conventionally formulated in CF, please confirm/state that all forward-modeled series share a common frame consistent with the GPS solution, or apply a degree-1 (geocenter) correction, otherwise a frame mismatch could bias the reported comparisons, especially the trend/long-term signals.
2. Step 1 (~line 150): Step 1 (~line 150) fixes the pre-processed set to 4,443 stations. Step 17 (~line 190) states that requiring >1 year of overlapping hydrological-model coverage "reduces the number of stations to 4,117." However, Figures 5 and 6 both report trends that are "estimated for 4,443 GPS locations." The trend maps should align with the two figure captions in Step 17. Furthermore, Step 1 (~line 150) needs clarification; it says, "It is worth mentioning that only 60 of these stations have a length of less than 3 years," which is unclear whether "these stations" refers to the 4,443 retained stations or the full 6,259 candidate set. Please specify the figures explicitly and revise them, i.e., 6259, 80%, 60, and 4443.
3. Step 6 (~line 170): The authors are encouraged, not obligated, to release the cleaned, non-decomposed daily displacement time series (post outlier/offset/loading correction, pre-wavelet-decomposition) referenced in Step 6, in addition to the three temporal-scale benchmark files. It could also contain the formal uncertainties for each epoch. I believe this would increase the dataset's reusability beyond the specific decomposition scheme proposed.
4. Lines ~290: Some sentences are slightly repeated. Where it reads "It is also worth noting that trends observed via GPS encompass all phenomena... reflect only hydrological loading" is just restated as "It is also worth noting that GPS-observed trends include local ground subsidence, while GRACE-derived and GLWS-predicted trends represent hydrospheric loading only." Keep one version only by merging them.
5. Lines 330–345: The sentence "After this drought, we see a decrease in GPS-observed displacements and a wet period from 2003 to 2007..." through "...but not fully captured by SLR and DORIS-based water storage" is almost a repetition of the preceding sentences, which read "After the drought, we see a decrease... All datasets remain consistent until 2007..." This might be a leftover editing, and please revise that.
6. Trend estimation / dataset files: The dataset does not report uncertainties (e.g., standard errors) for the GPS or GIA trend estimates in "Trend GPS & GIA.txt," nor for the displacement values in the three benchmark folders. Since trends are derived via least-squares, please add a column with formal trend uncertainties (ideally accounting for the noise model, e.g., power-law + white noise) or explicitly state in the readme why they are omitted. This is important for users to assess the significance of subsidence/uplift signals.
7. Data availability: No repository name, DOI, or file format/naming convention is provided in the text, as ESSD requires per journal policy. Furthermore, the DOI (10.18150/68WQXO) provided in Kłos & Bogusz (2026) is returning "DOI Not Found." Please check that.