the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Mar 2026
Analysis and modeling of coastal hazards to the desalination plants in the Baltic Sea – disentangling the impacts of a coastal storm in summer 2023
Abstract. We present an analysis of coastal hazards associated with the passage of an extreme storm named Hans in the Baltic Sea in August 2023. The storm resulted in disturbance of drinking water production at the desalination plants on Gotland, the largest island of the Baltic Sea. The limited ground water resources combined with increased demand during the warm tourist season lead to recurring seasonal water stress on Gotland. Thus, drinking water production through desalination of sea water is needed to complement the municipal water supply. The storm Hans triggered extreme water and organic material transport to the intake stations of the desalination plants clogging the filters, and coinciding with cold sea temperature spells, that collectively disturbed the water treatment process. We analyze the ocean-dynamical drivers of these coastal hazards and their impacts, and present a pathway toward a tailored forecast system to the desalination plants combining available observations, operational ocean model outputs, and statistical models. The linear coastal response to southwesterly winds of storm Hans was found to be the primary driver, either directly or indirectly, of the coastal hazards impacts. The linear regression models building upon this finding show a potential for future development of forecast framework to inform the water management on Gotland, if continued reporting and observational efforts can be secured.
- Preprint
(10487 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-1344', Giovanni Scardino, 29 Apr 2026
-
RC2: 'Comment on egusphere-2026-1344', Anonymous Referee #2, 05 May 2026
Massini et al have analyzed the data associated with two storm events in the Gotland region that impacted water desalination plants on the island. They have effectively utilized available oceanographic and atmospheric data to present an observational analysis of the hazards from those events to the desalination plants, and proposed a methodology to estimate future hazard forecasts.
While the observational and analysis part of the framework is sound and well documented, the forecast and broader applicability beyond the 2 historical events and identified locations is questionable based on the results presented in the manuscript. I have further described this concern and other minor issues below. I hope that the authors and the editor find these comments helpful for further improving this manuscript for future readers.
- With the linear regression based forecast model as a primary contribution of the manuscript, the fit is generally quite poor (max R2=0.62 at KVA and 0.46 at HER; and negative R2 for 2024 cold spells at KEV). Additionally due to lack of many events in the prediction period (F2024), the validation of the model is difficult. In fact the model failed to predict any of the 5 extreme transport events at HER in 2024. This is a major limitation of the study, that reduces the support of its utility as a broadly applicable hazard forecasting tool for the region.
- Tables 1 and 2: Can the authors include another row in both training and validation boxes for the number of predicted events, in order to compare against the observed events?
- Section 3: Can the authors include the fitted coefficients in the Results section, and comment on the predictive power of their chosen variables based on the coefficient sign and values.
- Sections 4 and 5: The authors’ conclusions regarding the linear dependence as evidenced by the results of linear regression model are not well supported given the weak regression fits. I recommend the authors revise the language to better reflect the fit, and/or indicate that a linear dependence model is not sufficiently well suited for this hazard forecasting.
- Line 475: What is the false positive rate of predictions from the linear regression fits?
- Line 78: The parenthesis around (cold water) are unnecessary.
- Line 102: Sentence is unclear.
- Line 106: personal -> personnel
- Figure 1: In *(b)* and (d), red sticks indicate…
- Figure 2: Superimposed on panels (a,*d*)
- Line 135: Please clarify whether the symbol should be strictly less than (<) or less than equal to (≤).
- Line 142: The last sentence by itself is unclear. Perhaps the authors are contrasting their linear regression approach with more complex regression models? It will be helpful to expand on this sentence, and provide examples of higher complexity models and alternative approaches. Inclusion of additional literature review summarizing other common methods used for similar hazard assessment will also be helpful, along with their current limitations to serve as additional motivations of this study.
- Line 167: …might lead *to* the accumulation of…
- Line 193: accumulation of -> its accumulation
- Line 218: Can the authors add information about how “quality-marked observations” are marked/determined?
- Line 224: Can the authors add why the data cannot be used directly?
- Line 241: Sentence is unclear.
- Eqs 1, 2: Can the authors include complete descriptions of notations, e.g., Tr, Cov, Var, etc.?
- Line 247: Does the equation assume normal distribution?
- Line 254: Why is the vector correlation represented by a single scalar value = 2?
- Line 254: Typically correlation ranges from 0 to 1, so I am unfamiliar with how a value of 2 is derived and its interpretation. Can the authors add more information about the basis of the equations and interpretation of results?
- Eq 3, 4: Can the authors include complete descriptions of notations?
- Line 326: We chose *the* winds…
- Line 329: os -> is
- Line 432: missing closing parenthesis
- Line 483: The data in parenthesis - R2=0.4 and 50% does not match Figure 7h or any of the other HER panels.
- Line 489: The authors mention couple of events at HER but list 5 events in Table 1. Can they clarify?
Citation: https://doi.org/10.5194/egusphere-2026-1344-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 441 | 156 | 34 | 631 | 29 | 38 |
- HTML: 441
- PDF: 156
- XML: 34
- Total: 631
- BibTeX: 29
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This work is very interesting and well-written, considering the topic of desalination associated with the storm Hans impact. In particular, the findings of this work are useful for coastal management, mostly along the island areas that need desalinization plants.
I have few comments reported line by line, that can be freely considered:
Line 10-14:In the abstract, when you mention the "linear coastal response" to the winds of storm Hans, it is unclear what this linear response specifically refers to. Furthermore, you report outcomes from linear regression models, but a brief sentence specifying the type of model implemented (e.g., which predictor variables were used) or which aspect of hazard forecasting it supports should be added in the abstract.
Line 38-40: The aspects related to low water resources are only briefly mentioned here. I suggest adding some details about the hydrogeological features of Gotland (e.g., thin soil layers, impermeable bedrock, efficient drainage) to better explain the high surface runoff and the consequent limited groundwater recharge that characterize this area and lead to water stress.
Line 55-60: The description of the desalination process is clear, but it would be helpful to specify at which stage(s) the clogging problems typically occur (e.g., intake pipes, drum filters, membrane nanofilters). This would strengthen the link between the hazard (material transport) and the reported operational impact.
Lines 73-85: It could be useful for readers to summarize the three coastal hazards and their impacts in a small table, which would facilitate comparison and provide a quick reference for the hazard-impact pairs described in the text.
Lines 97-100: You mention that storm Poly caused no reports of significant disturbance at the desalination plants, despite comparable wind speeds to Hans. This is a key observation. I suggest adding a brief hypothesis here (e.g., differences in storm duration, wind direction steadiness, or antecedent conditions) to guide the reader before the detailed analysis in Section 3.
Line 120: The flowchart summarizing your approach (Reported impact → Analysis → Identify drivers → Develop forecast) could be presented as a figure with a proper caption. Alternatively, if you prefer to keep it as text, I suggest integrating it into a continuous sentence rather than leaving it as a bullet list, to improve readability.
Line 130: A more accurate approach to represent the material transport proxy would be to consider the magnitude of the horizontal velocity vector (i.e., the composition of u and v components) rather than using the zonal (u) component alone, especially given the complex coastal circulation patterns around Gotland.
Line 347: You state that "surface chlorophyll decreases in areas of upwelled cold water". Does this refer to a decrease in chlorophyll concentration (i.e., lower biomass) or a decrease in the spatial extent of the bloom? Clarifying this would help readers understand whether the effect is due to physiological inhibition of cyanobacteria growth or physical redistribution of the bloom.
Line 575-580 (Sec. 4.2, future outlook): You mention that continued reporting from desalination plants is needed for future forecast development. I suggest adding a short recommendation on what specific parameters should be routinely logged by plant operators (e.g., filter pressure drop, cleaning frequency, water temperature at intake) to best support model validation. This would provide actionable guidance for stakeholders.
Table 1 (Page 23): The table shows that validation R² values for KVA in F2024 are lower than in-sample, but the number of extreme events is zero. It would be helpful to add a note in the caption or footer explaining that the validation metrics for KVA in 2024 are based on a period with no extreme events, which affects the interpretation of predictive skill.