the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Nov 2024
An assessment of the variability in temperature and salinity of the Baltic Sea from a simulation with data assimilation for the period 1990 to 2020
Abstract. A Baltic dataset covering 1990–2020 is reconstructed using a circulation model and data assimilation. Satellite observations of sea surface temperature and temperature and salinity (T/S) profiles are used to reduce model biases by a local Singular Evolutive Interpolated Kalman (SEIK) filter. The dataset is evaluated with assimilated T/S profiles and reprocessed grid observations, and the results demonstrate that the sea surface temperature, sea surface height, mixed layer depth, and vertical distribution of T/S are all reasonably reproduced. T/S trends at various depths in the Baltic sub-basins are analyzed from a reanalysis perspective, revealing a clear warming trend in recent decades, with a slight desalination trend in the northern Baltic Sea and a salination trend in the southern Baltic Sea. In particular, T/S trends of the Baltic Sea are larger in the south than in the north. In the Baltic Sea over the past 30 years, the temperature rises at a rate of 0.036 to 0.041 °C/year, with a larger warming trend below the thermocline than above it, while the salinity increases with a trend of -0.0036 to 0.049 PSU/year. In addition, seasonal variations are evident in the temperature at the surface, 60 m, and bottom, as well as in the surface salinity, whereas no clear seasonal variations are detected in the salinity below the surface and temperature at 100 m.
- Preprint
(4009 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2024-3283', Nima Zafarmomen, 07 Nov 2024
How does the choice of the Local Singular Evolutive Interpolated Kalman (LSEIK) filter impact the accuracy and stability of reanalysis, and what justifications were made for using a constant inflation factor rather than a spatially varying one?
Given that the reanalysis shows significant variability in simulation quality across different sub-basins, can the authors provide further details on how biases (especially for temperature and salinity) were corrected and what limitations persist in coastal and deepwater regions?
The study utilizes multiple validation datasets. How do differences between assimilated and validation datasets (e.g., OSISAF vs. CMEMS SST products) impact the interpretation of errors and the robustness of the results?
Given the reported warming and salinity trends, how do these findings compare with observed trends from other datasets covering different periods? Are the differences statistically significant and, if so, what could be contributing to these discrepancies?
The results indicate significant seasonal and depth-dependent variability in temperature and salinity. Can the authors elaborate on how these variations align with physical oceanographic processes and potential anthropogenic impacts on the Baltic Sea?
How do the authors address limitations of a 2-nautical mile horizontal resolution and the potential impacts of parameterization choices on model accuracy, particularly in areas with complex bathymetry?
Can more details be provided about the data quality control measures applied to observations before assimilation? How do these controls influence data coverage and assimilation accuracy, especially in areas with sparse data?
The study emphasizes the climate relevance of Baltic Sea warming and salinity trends. What specific recommendations can be made for policy or management actions based on the observed trends?
Were any sensitivity analyses performed on key model parameters or external forcing datasets to evaluate their influence on simulation accuracy and uncertainty in trend estimations?
How does this study's approach and findings compare to other similar long-term reanalysis efforts in the Baltic Sea? Are there specific methodological advantages that led to better accuracy or unique insights?
I appreciate the comprehensive approach taken in this study, particularly about the assimilation of temperature and salinity data. However, I believe that incorporating references to relevant works on streamflow assimilation could enrich the context and highlight parallel methodologies or complementary findings. For instance, streamflow assimilation studies often deal with similar challenges of data assimilation accuracy, bias correction, and spatiotemporal variability, which could offer valuable insights or comparisons. Including such references would provide a broader perspective on assimilation techniques and their applications, potentially benefiting both this study's conclusions and the wider research community. I strongly suggest citing below papers:
"Potential in improving monthly streamflow forecasting through variational assimilation of observed streamflow."
"Assimilation of sentinel‐based leaf area index for modeling surface‐ground water interactions in irrigation districts."
Citation: https://doi.org/10.5194/egusphere-2024-3283-CC1 - RC1: 'Comment on egusphere-2024-3283', Anonymous Referee #1, 19 Dec 2024
-
RC2: 'Comment on egusphere-2024-3283', Anonymous Referee #2, 21 Jan 2025
Review of "An assessment of the variability in temperature and salinity of the Baltic Sea from a simulation with data assimilation for the period 1990 to 2020" by Liu et al.
The manuscript provides a description of a model reanalysis simulation of the Baltic Sea with data assimilation for the study period 1990 to 2020. The generated data are used to study temperature and salinity trends at different depths and different sub-basins. The manuscript spends most of its content on the description, validation and discussion of the model simulation and less content on the temperature and salinity trends. From the current state of the manuscript, I do not know whether the manuscript is intended to be a model validation paper or a paper on the variability of temperature and salinity. Both parts need a lot of improvements in presentation and methods to be acceptable in the future. The trend analysis does not provide new information to the scientific community as trends have been studied a lot before. In addition, previous studies could also partially explain these trends due to changing dynamics and internal variability. This study overlooked this literature and does not add new understanding. The authors should consider splitting the manuscript into two papers in the future: one on the introduction of the model system and detailed validation of the simulation, and an improved version on T and S variability, trends and its dynamical origins, since otherwise the paper would be too long in my opinion. But the authors have to decide. Below are my main comments and suggestions on both topics.
Reanalysis and validation
-For the Baltic Sea, the saltwater inflow from the North Sea is the crucial process that renews most of the bottom water of the Baltic Sea. To see a model validation without a time series comparison of bottom salinities in the different basins is quite surprising. I would expect almost all bottom water trends to be governed by saltwater inflows. As the authors themselves state, the model does not capture this important driver of the Baltic Sea’s water masses. How can the reader trust the computed trends? In order to trust the trends, a plot of the time series compared to some central stations is necessary. In general, the model should be configured in such a way that this important physical process is represented by the model’s physics and not just included by data assimilation.
-In addition, trends may be biased in cases such as the following: a monitoring station was introduced in 2000 and is used for assimilation starting in 2000. Due to the introduction of the data, the model salinity hypothetically increases by 1g/kg in the following. This would artificially create a trend, although the model’s physics itself would not support a trend. Similarly, if the data assimilation were to suddenly stop during the model simulation. Do such cases occur in the simulation? If yes, the authors should be transparent about these potential biases in the trends.
-In general, the validation lacks a validation of time series. Most validation focuses on the mean state, but as this study focuses on trends, the validation should ensure that the model can be trusted in this regard, e.g. by showing time series.
Trends and variability
- Regarding this part of the manuscript, all results lack a lot of critical information for the reader to trust the results, as no errors, uncertainty ranges, and significance checks are presented.
-How do the trends compare to the trends of observations over the same time period? Comparing the results with trends from other studies may be biased by the different time periods considered.
-I do not think that the results add any new information to the Baltic Sea community. It does not add new understanding on the dynamical reasons for the well documented, observed trends which are only recreated.
-The whole presentation, from the introduction to the discussion to this part of the paper, is missing some critically important points, such as the large internal variability of the Baltic Sea (~30-year cycle). Due to the large natural variability in the Baltic Sea system, looking at trends over a period of similar length will most likely lead to aliasing artefacts. The authors touch on this part with their Figure 11, but do not discuss this critical property of the Baltic Sea.
-Many recent papers on this topic are not cited, including the Baltic Earth Assessment Report which reviewed the current state of knowledge of the Baltic Sea a few years ago.
General and specific comments
Introduction: Much of the recent literature is missing from the introduction. Many key points of the introduction are citing rather old papers, although newer literature with updated analyses is available.
L35-37: “Although several studies have focused on long-term changes in the Baltic Sea using observations over the last two decades (Fonselius and Valderrama, 2003; Winsor et al., 2001) [...]” It does not make sense to cite papers from two-decades ago to support this statement.
L37: “[...] spatiotemporal coverage limitations, especially in the deep Baltic Sea.” The Baltic Sea has one of the best and longest monitoring data coverage compared to many other marginal seas.
Section 2: Why are the parameters so different from the NEMO Nordic model of Kärnä et al.?
L71-73: What kind of vertical levels do you use? I assume z*? Can you resolve the halocline and thermocline with these z-levels?
L78: This equation of state is outdated. Why not TEOS-10? NEMO 4.0 has it.
L81: How is the conversion into in-situ temperature done? With the TEOS-10 equations?
L88: Why do you need isopycnal diffusion to close the Neva river inflow? This seems unusual.
L89: 3cm seems like a lot of friction. Why did you choose this high value? Other NEM Nordic setups have smaller values.
L94: Why is precipitation corrected? Any validation for this change?
L160: What is the climatology check?
L173/174: What is the source of the sea level data?
L179-183: Some typos. There are more occasions of wrong grammar and typos. I will not list them all.
The method section misses a description of trend analysis.
Section 3.3.: Spell out the acronyms somewhere.
L209/210: Can you be 100% sure that the different databases have no overlaps and that you do not compare the reanalysis with the data used for assimilation? I do not know the databases, therefore I ask.
L248-250: This shows that your model does not capture the essential physics of the salt water inflows, which are the central process that determines the bottom water properties. Therefore, I have little confidence in the trends of the model. Also, where does the salt come from in the DA simulation, if the model does not reproduce the physics of the inflows?
L258: I disagree because the model does not seem to capture one of the essential mechanisms, the saltwater inflows.
Section 4.1.3: SST products from satellites themselves are also imperfect.
Section 4.2.: Table 1 has no uncertainties, and it is not clear how the numbers were computed. Fig. 9 also needs uncertainty range and significance checks.
L338: The cold water range is not in Tab. 1
L340: The statement is obvious.
L348-350: Freshest water should not be found on the bottom, except in cells where river discharge is present.
L350-352: This result is not described in the previous text, but reads as a summary of the previous paragraphs. Again, this is not a new finding, see e.g. Kniebusch et al. (2019)
Section 4.2.1.: Are the trends significant? Without any significance checks, the whole section cannot be trusted.
Section 4.2.2.: The interdecadal variability of water mass properties is well described in the literature, which is overlooked by the authors, see one of my main comments.
L427: Incomplete sentence or paragraph
Figure 2: It would also be useful to see the temporal resolution of each station. Are there stations included which can provide data with very high temporal resolution, e.g. every 10 min? There exist such monitoring stations in the Baltic, e.g. the MARNET stations of the BSH in Germany
Fig. 5: The color of the colorbar does not align with the colors of the dots since the dots have alpha values.
Fig. 8: An additional panel with bar plots would be helpful for comparison, since the 60m range is quite large and could mask substantial MLD deviations
Fig. 9: This should include time series and trends from observations. This would help the reader to have confidence in the model results.
Fig. 10: Indicate where the trends are statistically significant.
Fig. 11: This could also be included in Fig. 9.
Fig. 12: Is this for basin averaged values? If yes, does the SD include both spatial and temporal variability? What would the same climatology look like for an observation?
Citation: https://doi.org/10.5194/egusphere-2024-3283-RC2 -
EC1: 'Comment on egusphere-2024-3283', Bernadette Sloyan, 28 Jan 2025
The reviewers provide critical comments and suggestions on the manuscript. They find that manuscript attempts to provide both a validation of the model assimilation method and ocean dynamical interpretation and assessment of trends are not successfully achieved. The authors should consider the reviewers comments carefully.
Citation: https://doi.org/10.5194/egusphere-2024-3283-EC1 - RC3: 'Comment on egusphere-2024-3283 (review 3)', Anonymous Referee #3, 30 Jan 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 197 | 53 | 13 | 263 | 8 | 6 |
- HTML: 197
- PDF: 53
- XML: 13
- Total: 263
- BibTeX: 8
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1