| 31 Mar 2026
MORE, a new convection-permitting reanalysis dataset over Italy and Alpine region: Validation and application in weather, climate and hydrology
Abstract. This study presents a new convection-permitting reanalysis dataset over Italy and Alpine region, produced through dynamical downscaling of ERA5 reanalysis with the non-hydrostatic mesoscale model MOLOCH. MORE (MOloch-downscaled ERA5 REanalysis) is a very high spatial resolution (~1.7 km) gridded dataset, covering the 1990–present period. The dataset includes hourly outputs of a wide range of variables at the surface and on pressure-levels.
MORE validation follows a multiscale framework applied to precipitation and near-surface air temperature using dense and quality-controlled observational datasets. MORE is benchmarked against other convection-permitting products and coarser- resolution reanalyses. Results show that MORE realistically reproduces spatial and temporal variability and improves the simulation of wet-hour frequency, precipitation intensity, sub-daily extremes, particularly during convective regimes, and key climate indicators such as the number of tropical nights, although a systematic cold bias is present in temperature.
As an application example, the May 2023 Emilia-Romagna (northern Italy) catastrophic flood is analyzed. MORE successfully reproduces the meteorological evolution of the two heavy rainfall events, providing added value in the representation of the mesoscale features resulting in localized precipitation extremes. In cascade, hydrological simulations driven by MORE data improve the representation of catchment-scale discharge, and soil moisture dynamics.
Overall, MORE represents the highest-resolution convection-permitting reanalysis currently available for Italy and the Alpine region. Its comprehensive set of variables at hourly resolution makes it a valuable reference for hydrometeorological studies, climate change adaptation, and climate services in regions with complex terrain and high exposure to extremes. The dataset is openly available at DOI: https://doi.org/10.5281/zenodo.18470948 (Stocchi, P. 2026) and will be periodically updated to ensure long-term accessibility, reliability and completeness.
The manuscript presents a new hindcast over Italy and the Alpine region, obtained through dynamical downscaling of ERA5 using the MOLOCH model. I use the term hindcast to distinguish this product from conventional reanalyses. I am impressed by the breadth and quality of the work carried out by the authors. The datasets employed meet the standards required for state-of-the-art climate science, the dynamical downscaling methodology is well-established and thoroughly tested, and the numerical model itself has a solid track record in the literature. The authors are to be commended for ensuring the dataset is fully traceable and openly accessible. The paper is dense and rich in results. Perhaps, more so than is typical for a single publication. Upon reading it, I was left with the impression that the manuscript could sustain more than one paper. That said, I do not question the authors' decision to present their hindcast alongside extensive evaluation and illustrative applications, as this provides a comprehensive reference for the community.
In my view, the primary limitation of the manuscript in its current form lies in the clarity of its communication, as detailed in the comments below. I therefore recommend publication of the manuscript, provided the authors adequately address the remarks that follow.
Comments:
1. Motivations and originality. In the Introduction, you note that several other deterministic reanalyses/hindcasts already exist over the same domain, at resolutions that are not substantially different from that of MORE. Yet in the Conclusions, you characterise MORE as a major achievement. Intuitively, I would associate a truly major breakthrough with a product offering a qualitatively new capability e.g. such as a convection-permitting ensemble hindcast, to give one example. This raises a question that I believe deserves more explicit attention: what motivated the development of yet another deterministic hindcast over this domain?
I would encourage the authors to revisit and strengthen the Introduction accordingly. Several angles could be worth exploring: Is the novelty primarily associated with the use of MOLOCH, which may be better suited than other models to specific weather regimes or orographic settings relevant to the Alpine region? Is there value in contributing an additional member to a growing multi-model ensemble of regional reanalyses, thereby benefiting the broader community? Is there coordination or an established dialogue with other hindcast producers over this domain? Or is the primary driver a more practical one: namely, the need for an in-house product to support bias adjustment of regional climate projections?
A clear and candid articulation of the authors' motivations in the Introduction would greatly benefit the reader, and would at the same time allow a more precise assessment of the originality and contribution of this work.
2. Verification scope. There is a clear imbalance between the number of variables produced and delivered by MORE and the two variables selected for verification. I acknowledge that soil properties are implicitly evaluated through the application sections. In general, I agree that 2-metre temperature (t2m) and total precipitation (tp) are likely to be the most widely used variables among the target user community, and that they benefit from relatively dense observational networks, making observation-based verification straightforward. However, in its current form, the manuscript does not clearly articulate the rationale behind this choice, nor does it explain why upper-level atmospheric variables are excluded from the verification exercise. A comparison between ERA5 and MORE at upper levels would, for instance, have been an informative addition. I am not asking the authors to perform such additional analysis, but rather to provide a transparent justification, either in the Introduction or in the Methods/Evaluation section, for the decision to focus exclusively on surface variables. Where relevant, the authors could also direct readers to existing literature in which MOLOCH has been evaluated against other variables (e.g. wind, pressure, upper-level temperature), thereby contextualising the verification choices made here and reassuring readers that the model's performance for those variables has been assessed elsewhere.
3. Communication of evaluation results. I have some uncertainty regarding the overarching motivation behind the evaluation as presented. In places, the text and its interpretations read as a sanity check of the MOLOCH downscaling. This is a legitimate goal, but the suitability of MOLOCH for dynamical downscaling is already well supported by prior literature. I would argue that the most important message of the evaluation is currently underemphasised, and that the text would benefit from revision to remove repetitions (e.g. the concepts in lines 711-714 are repeated several times across the manuscript) and improve readability.
In my view, the central contribution of the evaluation should be to give prospective users a concrete sense of the accuracy and precision of MORE for the verified variables at the temporal aggregations considered. Specifically, I would encourage the authors to address the following questions explicitly: What is the typical uncertainty of daily precipitation estimates from MORE, expressed, for instance, as a relative error? How does this change at hourly aggregation? What uncertainty should a user expect when working with 2-metre temperature? The answers to these questions are arguably the most practically valuable outcome of the evaluation, yet they are not sufficiently prominent in the current manuscript.
Equally important is guidance on the appropriate use of MORE: for which types of analysis is the product recommended? Are there specific regions, seasons, weather regimes, or applications for which MORE is known to perform less reliably, and where its use should therefore be approached with caution? This information, both the quantified uncertainty estimates and the user guidance, should be given a much more prominent role in the paper, and should be reflected clearly in the Conclusions and the Abstract.
4. Lines 113-119. The statement on "ensemble approaches" is not clear. Consider moving this paragraph to Conclusions?
5. Lines 211 and following paragraphs. The manuscript would benefit from a more explicit statement of the objectives underlying the two distinct strands of the evaluation. What insights do the authors expect to gain from comparing MORE against observations? And what is the purpose of comparing MORE against other reanalyses? I believe the answers are implicit in the text, but leaving them for the reader to infer is an unnecessary source of ambiguity. Once again, I would encourage the authors to state these motivations clearly and upfront. For instance, framing the comparison against observations as an assessment of the accuracy and precision of MORE, and framing the comparison against other reanalyses as a sanity check confirming that the downscaling procedure does not introduce gross inconsistencies relative to established products.
On a related point, there is a degree of internal tension in the manuscript: the authors state that they do not intend to rank the reanalyses or hindcasts (line 308) under comparison, yet the structure and language of the evaluation section largely does exactly that. I would ask the authors to resolve this inconsistency. If the comparison with other reanalyses is intended as a sanity check rather than a performance ranking, then the criteria for a satisfactory outcome should be made explicit. What would a successful sanity check look like? What level of agreement, or disagreement, with other reanalyses would the authors consider acceptable, and on what basis? Providing clear answers to these questions would considerably improve the interpretability of this section for the reader.
6. If the comparison with other reanalyses is intended as a sanity check, upscaling MORE to the coarser reference grid would be a more methodologically appropriate basis for the intercomparison. Evaluating global reanalyses on the native high-resolution grid of a convection-permitting product places them in an inherently unfavourable position (lines 398-402). That said, such a comparison may carry practical value by illustrating quantitatively what is lost by forgoing dynamical downscaling, and may discourage the uncritical use of global reanalyses for applications requiring high spatial detail. If this is part of the authors' intent, I would encourage them to state it explicitly.
Line 251. How do you remap? Bilinear interpolation?
Figure 4. Check the P99.9. Should it be P99?
Figure 13. The use of dots renders the figure somewhat difficult to interpret. The authors may wish to consider replacing dots with lines, which would sacrifice some precision in the representation of individual values but would likely improve the readability of the distributional tails considerably.
Figure 14. Change MOLOCH-ISAC into MORE.