the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Mar 2026
EMO-1: an improved version of the high-resolution multi-variable gridded meteorological dataset for Europe
Abstract. High-quality, gridded meteorological datasets are essential for continental-scale hydrological modelling. This paper introduces EMO-1, an advanced version of the European Meteorological Observations (EMO) dataset, developed to support the operational European Flood Awareness System (EFAS) of the Copernicus Emergency Management Service. EMO-1 provides daily and 6-hourly meteorological fields across Europe at a high spatial resolution of 1 arc-minute (~1.5 km) covering the period from 1990 to 2024. The dataset represents a substantial upgrade over its predecessor, EMO-5, integrating observations from 47 data providers and increasing significantly the number of stations used for interpolation. It harmonizes heterogeneous data from in-situ stations and integrates "virtual stations" from high-resolution regional grids and ERA5-Land to minimize data gaps in regions with very low station density. The dataset covers eight variables: daily and 6-hourly total precipitation, minimum and maximum air temperature, 6-hourly average air temperature, daily mean wind speed, solar radiation, and water vapor pressure. A rigorous, two-tier quality control system is applied to filter erroneous observations before processing. For interpolation EMO-1 uses an open-source implementation of the Angular Distance Weighting (ADW) scheme. Cross-validation results demonstrate that ADW maintains interpolation skill comparable to other deterministic methods while offering superior computational efficiency and reproducibility. EMO-1 prioritizes station density and timeliness to serve operational forecasting needs, although the resulting spatial-temporal heterogeneity makes it less suitable for long-term trend analysis. The dataset is openly available under a Creative Commons Attribution 4.0 license via the Joint Research Centre Data Catalogue, with foreseen annual updates and accessible open-source interpolation software to foster transparency and community collaboration.
- Preprint
(4292 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2025-723', Tomasz Berezowski, 25 Jun 2026
-
RC2: 'Comment on essd-2025-723', Michael Matiu, 21 Sep 2026
Salamon et al. present EMO-1, an updated version of the EMO-5 dataset, which is European gridded dataset of daily and sub-daily surface meteorology. The creation of such a dataset in merging many different sources is a huge effort, for which the authors are to be commended. Also for their plan to provide yearly updates! However, while the harmonization and QC of the in-situ data seem state-of-the-art, the creation of the grids is less motivated, explained, and validated than is required for a data paper.
Major points:
-
As the other reviewer has noted, validation is lacking. Internal validation is difficult to understand since details on the k-fold are missing; also I would expect some spatial and temporal assessment of the errors. External validation is difficult, since there are no comparable products available. However, a comparison to the standard reference of European climate grids, E-OBS, is definitely required, which would also help future users in choosing the right product.
-
The authors do not motivate why they chose 1km as spatial resolution. I don’t see a particular benefit of increasing resolutions just “because we can”, since the distribution of stations has likely much larger inter-station distance than 1km (and the authors correctly state that the distance between stations is the most important factor for creating grids). Besides, it would be useful to know the inter-station distance for each variable, and also how it varies over space.
-
Related to the resolution issue. ADW is a very simplistic interpolation scheme, and while it might work well for precipitation, which is the main variable for the authors, it certainly has drawbacks for the other variables. For instance at 1km resolution, temperature and radiation will be strongly affected by topography, and the simple elevation correction (6K/km) is not sufficient in areas of complex topography. Which is why many products in mountain areas such the Alps, Scandes, or Carpathians use more sophisticated approaches to deal with elevation and temperature. Moreover, for radiation, terrain shading is not accounted for, but becomes relevant at these spatial resolutions. As a continental product, I do not expect the same sophistication as for more regional products, but this issue is a consequence of increasing resolutions without adapting the interpolation technique; and thus needs to be discussed as a limitation. Even better would be a quantitative comparison to existing high-resolution products such as are available in the Alps or Scandes.
-
The merging of ERA5L with stations is more tricky and not discussed at all. It is known that ERA5L can have significant biases depending on variables, season, and location, so treating them as virtual stations might introduce errors in the EMO-1 grids. Moreover, it is not explained how the mismatch between areal averaged meteorology as provided by ERA5L and point-wise information as provided by the in-situ stations has been dealt with.
Minor points:
-
Appendix: It seems that many data providers only have data from 2020 onwards. Why is that? I assume there are data also before.
-
Table 1, daily tp: are you sure it is 6-6 UTC? From what I remember, many provide it also as 0-0 UTC.
-
L253: Repeated stratified k-fold: please provide more details.
-
L255: CSI is a metric for binary variables. How was it applied to precip?
-
Fig 3: Seems to mix virtual and in-situ stations. I would suggest distinguishing between in-situ and virtual stations. Also it seems that virtual stations have only been used for precip and not for the other variables? Why are there stations in the sea (e.g., around the UK)? A sequential colour scale might be more useful to identify length by color.
-
Why use virtual stations with a cutoff on lon/lat? To me, it seems that station density is sometimes equal in other regions as well.
-
The authors often provide very vague methodological descriptions, for instance related to the climate zones, thresholds, expert judgement, etc. Either the author’s improve the description or provide the source code in some open repository, so the steps can be retraced.
-
There is a kind of mismatch between the maps of data availability and the time series such as Fig4. From Fig4, I would expect the maps to be more clustered depending on data provider.
-
While the author’s developed the dataset with mainly one particular hydrological application in mind, it would be useful, if they could think of other potential use cases, which surely exist.
Citation: https://doi.org/10.5194/essd-2025-723-RC2 -
Data sets
EMO: A high-resolution multi-variable gridded meteorological data set for Europe Goncalo Gomes, et al. https://data.jrc.ec.europa.eu/dataset/0bd84be4-cec8-4180-97a6-8b3adaac4d26
Model code and software
pyg2p Goncalo Gomes, et al. https://github.com/ec-jrc/pyg2p
gridding Goncalo Gomes, et al. https://github.com/ec-jrc/lisflood-utilities?tab=readme-ov-file#gridding
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 782 | 388 | 120 | 1,290 | 140 | 127 |
- HTML: 782
- PDF: 388
- XML: 120
- Total: 1,290
- BibTeX: 140
- EndNote: 127
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The presented manuscript contains information concerning development of a sub-daily grided meteorological product for Europe. The product itself is interesting, and certainly will find some applications. However, the manuscript contains a serious shortcoming which has to corrected before publication.
My major concern is that the manuscript lacks validation. There is a sort of validation presented in Table 4 and appendix B, which gives an overview of what errors can be expected for 56 rainy days, however, this cannot be used as a proper, unbiased, fair validation of the dataset. Since only rainy days were chosen, the error values are clearly biased towards rainy conditions. 56 days is also a too short record to conclude about the performance in the 1990-2024 period. Nonetheless, Authors clearly state in the manuscript that those errors were used to justify the selection of the interpolation algorithm (’We assessed the three interpolation methods applying the repeated stratified k-fold cross validation method’) i.e. not to validate entire dataset. Authors also argue that the precipitation was the most important variable in the dataset, hence the results in the main part of the manuscript consider only precipitation. Maybe precipitation is the most important for the Authors use case, but in general I do not agree with that, as other users may have different use cases, especially that the potential evapotranspiration (derived from remaining variables in the dataset) is often greater in some parts of Europe than precipitation. In this scope, the sentence ’Cross-validation results demonstrate that ADW maintains interpolation skill comparable to other deterministic methods while offering superior computational efficiency and reproducibility’ in the abstract is misleading as it suggest a fair validation was done.
To summarize, the manuscript has to be updated with the following points before publication:
Another point is that Authors are mixing points from gridded reanalysis products with in situ observations. Please better justify this approach, especially in scope of homogeneity of the data from different source types. Please include experimental results to support this.
Finally, the study lacks comparisons with other datasets, such as E-OBS, or different national products. While this point is not obligatory, inclusion of such a comparison would greatly increase the manuscript impact and reduce a workload of future users (who would have to perform such analysis on their own to find out which dataset to use). Authors state that: ’In a future study we aim to provide a comparison of the new EMO-1 dataset with other gridded meteorological datasets such as the […]’, however, such an analysis should be performed now - it should be a part of the data descriptor paper.
Minor points:
Please include x and y axis labels in all relevant plots. Remember about units.
Please include legend titles where applicable. Remember about units.
Figures A9-A16 - remove letter (a) - (h) from the plotting area. Remove letters a) - h) from the plot caption.
Figure 3 and figures A9-A16 use sequential color scale - now the figures are extremely difficult to parse. Remember about units in the legend. The best if you would use a perceptually uniform sequential color scale.
Tables in appendix B - all are missing units.
Acknowledgements - usually meteorological organizations require mentioning the data source in the acknowledgements - please make sure you comply.