the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Aug 2026
Forecasting the 12 November 2019 Venice Acqua Alta event: the role of baroclinic compounding effects
Abstract. The northern Adriatic Sea is known to be threatened by extreme sea level events. Venice is exposed to these events, locally known as Acqua Alta, which cause the flooding of the historical city centre. Despite its relevance, a comprehensive analysis of extreme Acqua Alta events that considers all possible compounding drivers, including baroclinic contributions, is, to our knowledge, still missing. In this work, numerical results based on the Copernicus Marine Mediterranean Analysis and Forecasting System are analyzed to better understand the physical mechanisms underlying the 12 November 2019 Acqua Alta event and to assess individual processes playing a role in the event dynamics.
The novel finding of this study is the identification of the contribution of baroclinic dynamics. We estimate a net baroclinic contribution of 7 cm (4 %) to the 175 cm sea level peak. We show for the first time that internal seiches along the northern Adriatic slope contribute to the dynamics of the Acqua Alta event. Moreover, we identify a northern Adriatic density front and quantify its contribution to the time-mean positive northward sea level slope.
- Preprint
(2724 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 16 Oct 2026)
- RC1: 'Comment on egusphere-2026-4734', Anonymous Referee #1, 09 Sep 2026 reply
-
RC2: 'Comment on egusphere-2026-4734', Anonymous Referee #2, 17 Sep 2026
reply
The manuscript addresses an important contribution to extreme sea levels that has not been considered in previous sea level extreme studies in the Adriatic. The manuscript is well structured and written, but lacks in-depth analysis of the quoted baroclinic sea level contribution.
That is, at first, process-level explanation of internal seiches and further exploitation of model results. Why the seiche has the period of 12 h? What is spatial structure of the seiche? Why the seiche is excited (by the barotropic seiche of 21.5 h, or)? Some background basic-level (at least) conceptual modelling, not just quoting other study(ies). At present, the text is not convincig on this.
Further, I also lack a wider discussion - are your results applicable to other similar basins with extreme sea levels? Are there similar studies? There are many coastal regions where such a process (if here proved) may occur and contribute to extreme sea levels? Or not? You should dig a bit into literature to increase the potential impact of your results.
I also agree with other comments of the other reviewer.
Citation: https://doi.org/10.5194/egusphere-2026-4734-RC2 -
RC3: 'Comment on egusphere-2026-4734', Anonymous Referee #3, 21 Sep 2026
reply
Referee’s comments on manuscript egusphere-2026-4734 by Anna Chiara Goglio et al.
My initial reading of the manuscript left a favourable impression. The study addresses several interesting aspects of physical oceanography, particularly the role of baroclinic processes during atmospheric forcing events, within the highly practical context of improving flood prediction in the Venice Lagoon. The manuscript is generally fluent, well structured, and pleasant to read.
A more detailed examination, however, raised several questions concerning both the forecasting methodology and the physical-oceanographic interpretation of the results.
1. Forecasting methodology and operational relevance
1.1. The principal result is a 7 cm reduction in the peak sea-level forecast error during a single event with a maximum water level of 175 cm: the error decreases from 25 cm in the barotropic simulation to 18 cm in the baroclinic simulation. The operational significance and scientific robustness of this improvement have not yet been demonstrated sufficiently. The authors should clarify whether a 7 cm improvement would materially affect flood-warning decisions or protective measures.
1.2. The MOSE barrier system is intended to protect Venice and its lagoon from exceptionally high water levels, and operational attention has therefore shifted toward supporting timely decisions concerning barrier activation. The manuscript refers to a critical sea-level threshold of 140 cm, corresponding to the flooding of approximately 60% of the historic city centre. However, the practical importance of achieving the highest possible forecast accuracy after the barrier-activation threshold has already been exceeded remains unclear. The authors should explain more explicitly how improved forecasts above this threshold inform operational decision-making.
1.3. The comparison of two forecast errors from a single event does not provide a statistically robust assessment of forecast improvement. Lionello et al. (2021) identified 25 events during 1936–2019 in which sea level at Venice reached or exceeded 140 cm. I was unable to find consistent and comparable forecast-performance metrics for these events in the manuscript. A broader event-based evaluation, using clearly defined and consistent verification metrics, would substantially strengthen the conclusions.
1.4. In addition to the baroclinic processes examined in the manuscript, coastal sea level may be influenced by wave setup and sub-hourly oscillations associated with meteotsunamis. The magnitude of these contributions may exceed the reported 7 cm improvement. Although representing such processes would require additional effort and the coupling of circulation and wave models at relatively fine spatial and temporal scales, their potential contribution should at least be discussed in a forecasting-oriented study.
1.5. The use of a purely barotropic model configuration, such as MedFS-BT, may appear outdated in the context of modern operational forecasting, although barotropic models remain valuable for process studies, sensitivity experiments, and benchmarking. The authors should therefore explain more clearly the present operational role of MedFS-BT and justify its use as the principal baseline for evaluating the baroclinic configuration.
1.6. Conditions in the Venice region are forecast using the regional Med MFC model at a horizontal resolution of approximately 4 km. It is unclear why higher-resolution downstream or nested models were not employed, particularly given the complex geometry and dynamics of the Venice Lagoon. The authors should discuss the availability of such modelling systems and explain whether collaboration with the teams operating them was considered.
2. Physical oceanography
2.1. The analysis of oceanographic processes during extreme high-water events is interesting and constitutes one of the strengths of the manuscript. The results are based on experiments with an advanced implementation of the NEMO model; however, the rationale for several parameter choices requires further explanation. Many circulation models, including those applied in the North Sea and Baltic Sea, use second-order turbulence closures based on turbulent kinetic energy and its dissipation rate. It is therefore unclear why the authors selected the comparatively traditional Pacanowski–Philander Richardson-number-dependent mixing scheme. The treatment of bottom friction also requires justification. Because bottom friction is highly important for storm-surge dynamics, the authors should explain their use of a spatially constant drag coefficient and discuss the potential consequences of neglecting spatial variations in bottom roughness.
2.2. The discussion of internal seiches is an interesting component of the study. However, the model results are not compared with observations, and the basis for the estimated period of approximately 12 h remains unclear. The manuscript would benefit from an accompanying resonance or normal-mode analysis. Moreover, the process-oriented investigation need not be restricted to the selected high-water interval. The authors could instead identify and analyse periods during which observations show distinct internal-wave or seiche signatures and then assess whether the model reproduces these signals.
2.3. The discussion of frontal dynamics during periods of high sea level is also of interest, but it relies exclusively on model output. The analysis would be considerably strengthened by comparison with observations. In particular, the authors should investigate whether suitable satellite or other remote-sensing products are available through the Copernicus Marine Service or related data repositories.
Overall, the manuscript addresses an important topic and contains potentially valuable results. However, the operational significance and statistical robustness of the reported forecast improvement require stronger substantiation, while several modelling choices and physical interpretations need further justification and observational support.
Citation: https://doi.org/10.5194/egusphere-2026-4734-RC3 -
RC4: 'Comment on egusphere-2026-4734', Anonymous Referee #4, 23 Sep 2026
reply
Referee's report on the manuscript 'Forecasting the 12 November 2019 Venice Acqua Alta event: the role of baroclinic compounding effects' (egusphere-2026-4734) by A. C. Goglio et al.
General comments
The authors use baroclinic and barotropic versions of the MedFS operational system driven by the ECMWF atmospheric forcing in order to reproduce sea-level variability off Venice over the three-day interval (10-12 November 2019) that culminated in one of the highest sea levels observed there. They show that both versions of the model underestimate the maximum sea-level height, with the baroclinic simulation surpassing the barotropic simulation by 7 cm. Consequently, they conclude that the baroclinic effects have contributed, albeit moderately, to the sea-level maximum. The manuscript is well written, the data set is adequate, and the modeling results are interesting. Therefore, I recommend that the manuscript be published after the points raised below are addressed.
Specific comments
Line 91: Spatial resolution of the ECMWF forcing (0.1 degree) is too low: it has been shown before that a much higher resolution (0.01 degree) is needed if the atmospheric mesoscale depression - which caused a meteotsunami and therefore contributed almost 30 cm to the Venice flood of 12 November 2019 - is to be reproduced. It does not seem logical that attention is paid in the present manuscript to the relatively small baroclinic effect and that at the same time a much larger effect is not taken into account.
Line 110: Previous studies have shown that during the three-day interval considered here, basin-wide surface seiches were of no importance and therefore the interval was influenced by tides (throughout the interval) and by a storm surge (at the end of the interval). A meteotsunami was also important on 12 November 2019, but - as already mentioned - it is not included in the present simulations due to the low spatial resolution of atmospheric forcing.
Lines 198-200 (and later in the paper): The baroclinic current component defined on these lines is actually the vertically sheared current. The sheared current may occur in the homogeneous sea and therefore it is not necessarily related to the baroclinic effects.
Lines 212-213 (and later in the paper): Figure 5 shows that baroclinic waves are formed over the first part of the three-day interval, probably due to the interaction of tidal currents with basin bathymetry and the consequent rising and lowering of isopycnals. It is however not possible to state that the baroclinic waves represent the baroclinic seiches without showing that they could propagate to a suitable reflection point (Adriatic coast, Adriatic shelf break, Otranto Strait?) and then return to the North Adriatic so as to form the standing waves (i.e., seiches). More probably, the baroclinic waves represent the free waves that are generated over the sloping bottom and that have periods determined by tidal forcing.
Lines 224-225: A comparison of Fig. 5g and Fig. 6b shows that over the first two days of the three-day interval considered, the baroclinic waves are much more pronounced in the presence of tidal forcing than in the absence of tidal forcing. It is only during the third day that the vertically sheared currents are similar in the two figures. The sheared currents coincide with the formation of storm surge and are most probably driven by the sirocco wind generating the surge.
Lines 295-339: The discussion and conclusion sections would benefit from a simple summary of the dynamics revealed by the manuscript: readers like to get an idea that is easily remembered. Figure 7 strongly suggests that the Po River plume was well developed in November 2019, whereas Term 2 dominating the right-hand side of Equation (4) implies a few centimeters of sea-level difference between the plume and the surrounding area. Is it then possible that the baroclinic effect discussed by the authors was primarily due to the Po River plume being advected by the sirocco wind that caused the storm surge? I encourage the authors to provide a simple verbalization of their main finding, if possible.
Technical corrections
Line 25: It is stated that the average depth of the Adriatic Sea is 126 m, but this is the average depth of the Adriatic shelf area - the average depth of the whole Adriatic is much larger.
Figures 4 and 5: If I see it correctly, the maximum sea-level height modeled by MedFS for the station I-TG is 93 cm according to Fig. 4a and 80 cm according to Fig. 5a. Did I get something wrong?
Citation: https://doi.org/10.5194/egusphere-2026-4734-RC4
Data sets
Observed sea level data at the ISMAR tide-gauge ISMAR-CNR, Institute of Marine Sciences, Venice, Italy https://www.comune.venezia.it/it/content/3-piattaforma-ismar-cnr
Copernicus Marine Analysis and Forecasting System product Copernicus Marine Service https://doi.org/10.48670/mds-00359
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 150 | 75 | 65 | 290 | 60 | 58 |
- HTML: 150
- PDF: 75
- XML: 65
- Total: 290
- BibTeX: 60
- EndNote: 58
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The paper addresses an important event and is generally well written. The modelling setup is interesting, and I appreciate the effort to separate the sea-level response into different components and to discuss the underlying dynamics. That said, I have several major concerns about the strength of the attribution and the consistency of some of the quantitative arguments. In its current form, I do not think the manuscript fully supports its main conclusions.
Main comments:
1. Attribution of the 7 cm difference
The manuscript appears to interpret the roughly 7 cm difference between MedFS and MedFS-BT at the event peak as the baroclinic contribution. I do not think this is sufficiently justified.
The two configurations differ in more than baroclinicity alone; they also involve changes in wave coupling, wind stress, bottom friction, and initialisation strategy. These differences are all potentially relevant at the centimetre level. As a result, the 7 cm signal cannot be attributed uniquely to baroclinic processes based on the present design.
At minimum, this needs to be stated much more cautiously, and the discussion should acknowledge the confounding model differences explicitly.
2. Relationship between the 7 cm and 3 cm estimates
The paper reports a 7 cm peak difference, while the front analysis gives a mean contribution of about 2.9 cm. The connection between these two numbers is not clear.
It is not obvious whether the offset applied in Sect. 2 removes the mean baroclinic setup or only part of it, and the reader is left unsure how the 3 cm front contribution should be related to the 7 cm peak difference. This point needs to be explained much more clearly.
3. Evidence for the internal seiche interpretation
The internal seiche interpretation is plausible, but the evidence is not yet convincing enough. The period appears to be inferred mainly from the model fields, and a 12-hour signal is close to the main tidal constituents. A spectral analysis, or at least a more direct diagnostic, would help considerably.
The relation between the identified oscillation, the filter bands used in the decomposition, and the “tides–seiches” component should also be clarified.
4. Link between the diagnosed processes and the peak sea level
The manuscript identifies baroclinic currents, internal-wave-like variability, and a mean density-front contribution, but it does not clearly show how these combine to explain the 7 cm peak difference at Venice.
A more explicit budget, or at least a clearer estimate of the sea-level signal associated with the internal seiche, would strengthen the paper substantially.
5. Scope and framing
This is presented as a forecasting study, but it is based on a single event and a single initialisation. It reads more like a case study or hindcast analysis than a forecast evaluation.
The framing should be toned down accordingly, especially where forecast skill or predictability is implied.
6. Robustness
Because the analysis is based on one event only, it is difficult to judge how robust the reported differences are to initial conditions or modest changes in setup. I do not think an ensemble is essential, but the paper should be clearer about the limits of the analysis.