the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The Mediterranean forecasting system. Part I: evolution and performance
Giovanni Coppini
Emanuela Clementi
Gianpiero Cossarini
Stefano Salon
Gerasimos Korres
Michalis Ravdas
Rita Lecci
Jenny Pistoia
Anna Chiara Goglio
Massimiliano Drudi
Alessandro Grandi
Ali Aydogdu
Romain Escudier
Andrea Cipollone
Vladyslav Lyubartsev
Antonio Mariani
Sergio Cretì
Francesco Palermo
Matteo Scuro
Simona Masina
Nadia Pinardi
Antonio Navarra
Damiano Delrosso
Anna Teruzzi
Valeria Di Biagio
Giorgio Bolzon
Laura Feudale
Gianluca Coidessa
Carolina Amadio
Alberto Brosich
Arnau Miró
Eva Alvarez
Paolo Lazzari
Cosimo Solidoro
Charikleia Oikonomou
Anna Zacharioudaki
Abstract. The Mediterranean Forecasting Systems produces operational analyses, reanalyses and 10-day forecasts for many Essential Ocean Variables (EOVs), from currents, temperature to wind waves and pelagic biogeochemistry. The products are available at a horizontal resolution of 1/24 degrees (approximately 4 km) and 141 unevenly spaced vertical levels.
The core of the Mediterranean Forecasting System is constituted by the physical (PHY), the biogeochemical (BIO) and the wave (WAV) components coupled offline, consisting of both numerical models and data assimilation modules. The 3 components together constitute the so-called Mediterranean Monitoring and Forecasting Center (Med-MFC) of the Copernicus Marine Service.
Daily 10-day forecasts are produced by the PHY, BIO and WAV components as well as analyses, while reanalyses are produced for the past 30 years about every ~3 years and extended (yearly). The modelling systems, their coupling strategy and evolution is illustrated in detail. For the first time, the quality of the products is documented in terms of skill metrics evaluated on a common three-year period (2018–2020), giving the first complete assessment of uncertainties for all the Mediterranean environmental variable analyses.
Giovanni Coppini et al.
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2022-1337', George Zodiatis, 24 Jan 2023
The article is very interesting and useful for the MedSea downscaling modeling systems.
Please clarify that the "Sea ice coverage..." at lines 191-192 concern the North Atlantic Ocean domain used to produce the boundaries for the Mediterranean WAV model.
Citation: https://doi.org/10.5194/egusphere-2022-1337-CC1 -
AC1: 'Reply on CC1', Giovanni Coppini, 31 Jan 2023
Dear George Zodiatis,
thanks for the comment and your appreciation for our work,
We will take into account your comment on the "Sea ice coverage..." and clarify better how is is used to produce the boundaries for the Mediterranean WAV model.
Citation: https://doi.org/10.5194/egusphere-2022-1337-AC1 -
AC5: 'Reply on CC1', Giovanni Coppini, 05 Apr 2023
Sea ice coverage fields (obtained from ECMWF IFS) are used for the North Atlantic wave model (wave energy in the WAM model is dissipated due to the presence of ice) that provides lateral boundary conditions to MEd-waves. Following this comment, the text in Line 191-192 will be changed as follows: “ Sea ice coverage fields used by the North Atlantic wave model are also obtained from ECMWF”
Citation: https://doi.org/10.5194/egusphere-2022-1337-AC5
-
AC1: 'Reply on CC1', Giovanni Coppini, 31 Jan 2023
-
RC1: 'Comment on egusphere-2022-1337', Anonymous Referee #1, 07 Feb 2023
Review general comments
The paper is very interesting and provides very accurate, robust and useful information for those developing operational ocean forecasting systems and for users of operational forecasting products. The plan of the paper is clear : to provide accurate information about main features of state of the art operational system and on scientific quality of the products which are delivered in the framework of the Copernicus Marine Service. Most of the information provided in this paper is also available in Copernicus Marine documentation and it is a very good initiative to publish this information in peer review journal. I recommend that the authors provide in the introduction more precise information about the Copernicus Marine Service framework and about product quality strategy and how this paper is focusing this strategy.
The scientifical analysis of main uncertainties, analysis and/or forecast errors is poorly described in the paper, I suggest to provide more information where possible on the source of uncertainty, on the missing processes in the forecast system, on the errors and uncertainties in the forcing fields … This should at least be mentioned in the section 3 introduction part, to indicate that the paper doesn’t provide a strong and detailed analysis of main drivers/stressors of forecast uncertainties based on sensitivity study (or other experimental framework) but only statistics and accuracy numbers based on a reference simulation produced to calibrate an operational forecasting system.
I recommend the publication of this paper, if the authors provide answers to the following questions and a revised version of the paper taking into account the main suggestions.
Questions and suggestions:
1 Introduction :
I strongly suggest adding in the introduction part information about product quality activities in Copernicus Marine, QUID documentation and why it is important to publish this information in peer review journal.
It could be useful to add two references i) to the ETOOFS guide « Implementing Operational ocean Monitoring and Forecasting Systems Alvarez, Ciliberti and Bahurel 2022 and especially a citation to the chapter 4 containing a section dedicated to ocean forecasting system validation and to ii) Sotillo et al 2021 who describe validation and product quality sqtrategy in copernicus marine. (Sotillo, M. G., Garcia-Hermosa, I., Drévillon, M., Régnier, C., Szczypta, C., Hernandez, F., Melet, A., Le Traon, P.Y. (2021). Communicating CMEMS Product Quality: evolution & achievements along Copernicus-1 (2015- 2021). Mercator Ocean Journal #57. Available at https://marine.copernicus.eu/it/node/19306)
This paper is the part I, it could be useful for reader to have information about the partII, which topic and how it will be related to part I.
In the introduction you provide quantified useful information for the transport at Gibraltar strait and for the mean wave period and significant height, it would be good to add also information on uncertainties and on variability instead of just only a mean value. It’s also related to the main objective of the paper to provide quality information on ocean model simulation which should include uncertainties.
It will be good to provide a clear definition of offline coupling (line 66) where are the feedback between the model components, which variables … There is often confusion between forcing, two way forcing, offline coupling or full coupling.
Line 71 : could you explain wheather this standard is also applied for the other MFC in copernicus marine and whether it is good practice for operational oceanography.
2 Description of the Med-MFC core components
2.1.1 Numerical model description
Could you explain in more detail how the exchanges with the Atlantic ocean are implemented and how these exchanges between med and atlantic are better resolved. are there any references on this development?
Can you describe the changes in the bathymetry that have been made in the different critical areas (adriatic, straits atlantic border). What are the reasons for these modifications ?
The barotropic time step is 100 times smaller than the baroclinic time step, is this justified by code stability or other concerns? This seems large compared to other model configurations already published.
2.1.2
Could you explain if atmospheric forcing is a mixed of analysis and forecast, or only atmospheric analysis during the ocean analysis phase and only atmospheric forecast during the ocean forecast phase ? And what’s the higher temporal resolution from year 2020 ? is it 1h ?
You are using closed boundaries in Atlantic for WW3 model, this is strange and not consistent with the justification to have a Atlantic model for the boundary condition of the WAV system. Could you comment on this choice and justify why there is different implementation for these two models.
You provide the salinity of the river discharge in table A4, there are differences depending of the river, I did not find justification of these differences in Delrosso 2020. How do you explain and justify these differences?
2.1.3 The data assimilation component
You describe a method for rejecting an observation based on a quality check. Could you explain how this square departure is computed in the methodology ? For each individual Temperature or salinity profile along the vertical? For each sla track? In a spatial box, temporal window?
Correction to the background is applied once a day, does it mean that it is applied during the last time step of the day, there is no IAU method used to apply analysis increment?
2.2.2. Model initialization, external forcing and boundary conditions
Could you explain what is the impact of the lateral forcing with full wave spectrum in the atlantic and why it is not applied in the WW3 configuration?
2.3.2. Model initialization, external forcing and boundary conditions
Could you provide more information on the initialisation procedure for the BGC model. Does the BGC model initialisation use a constant profile for each area and for each variable for the initialisation? How long the model is integrated to smooth these discontinuities between the areas?
Could you provide more information about the newtonian damping. Where exactly? Only for the Atlantic boundary? The Atlantic part in the bio system is smaller than for the physics if I am right?
Could you provide more information on the atmospheric pCO2 forcing, it is not clear how this is computed. Is it a constant mean value applied for the Med sea? what is extrapolated?
You didn’t provide very precise information on the computation of error covariance for the biogeochemistry. Is it the same characteristic than for the physical assimilation system? Which resolution, length of the simulation to compute EOF? Also 3-year simulation or there are other constraints related to bgc processes and differences in term of observations
- Quality assessment
3.1. PHY component skill
You explain that salinity is characterised by a negative bias, this is not what is shown in tab2. Negative bias is only in the first layer, below the bias seems to be positive.
Regarding the temperature bias, the bias seems to be negative in the upper layer and positive below 60m depth. Could you explain your assumption and how overestimation of shortwave flux will produce a warm bias only below the mixed layer?
Concerning the negative salinity bias due to mixing at Gilbraltar (fig 4) is it something verified with statistic in appropriate boxes for example in the Alboran sea or in a western part of the med sea ?
About the spatial variability of the SLA error (line 406), you suggest it could be impacted by the distribution of observations but sea level variability and the eddy kinetic energy should be much larger. Could you comment on this, is it link to your comment on model inaccuracies? Could you identify which components of the model are affected (forcing, assimilation method, numerical scheme, missing processes ...?)
You have large differences of sla error between satellites. How do you explain these differences between the satellites? Do you use the same measurement errors for all the satellites? Is it due to the satellite coverage?
3.2. WAV component skill
You explain at the beginning of the paper that forecasts are not assessed in this paper, is it different for the waves?
It is difficult to see in the figure what is explained in the text, for example we can’t see the underestimation for very small wave heights (<0.6m), neither the underestimation for MWP<7s. If I am right, in fig 6, there is an underestimation for the period <5s and overestimation for period > 7s. Do you think the overestimation is significant for 2m waves? It's difficult to trust this information with figure 2. Could you provide more information and argument to consolidate these conclusions.
Could you explain better the interpretation or the figures.
You haven’t shown the seasonal results, could you say whether the best results in winter are for the height or for the period or both?
Line 432, could you explain what is the CORR deviation in the figure and the correlation coefficient commented in the text.
Line 443 : Is it underestimation instead of overestimation ?
3.3 BIO component skill
L475 : there is no illustration of spatial gradient in fig 10, the figure only show the seasonal cycle
L483 : In table 5, the RMSD error is the order of 10 to 40m and not a meter, depending of the domain and the variable. Could you explain how the uncertainty is estimated
L490 : Consistency between the observation and the model seems to be good in all the areas, but could you comment on some of the differences, are they significant ? For example, at the surface in the alboran sea, there is not a good agreement for alcalinity or DIC. It's also the case for alcalinity in the Aegean basin. Is it not possible to add the mean profile computed with the observation?
- Conclusions and Future Perspectives
The analysis of the source of uncertainty is missing. There is some information for the physical part and for the waves but nothing for the biogeochemistry. That should be very useful to synthetise main source of uncertainty and error in the conclusion for the different systems.
L563 : could you explain what is expected assimilating sea level in coastal stations? Complementarity to altimetry sea level? No plan to assimilate altimetry close to the coast and on the shelf?
L580 : could you explain the link between these new model development/improvement and the expected impact on the metric and score computed in the previous section.
Corrections
L 51 : von schuckmann instead of von Schckumann
L 78 ref is missing
L195 : global physical model instead of global wave model
L417 : right panel instead of lower panel
L442 : ECMWF forcing instead of ECMWF is forcing
Figure 7 : MYKKON buoy is missing?
Figure 8 : left panel, the yellow dash line should be in the foreground as on the right panel. There is colorbar and no information about the grey cross and the yellow dashline in the legend.
Figure 10 : is it model forecast or model analysis in black ?
Table 3 : westward transport should be 0.87 instead of 0.087
Table 5 : mean values are missing for the Chl and Nitrate for the upper layer.
Citation: https://doi.org/10.5194/egusphere-2022-1337-RC1 - AC2: 'Reply on RC1', Giovanni Coppini, 05 Apr 2023
-
RC2: 'Comment on egusphere-2022-1337', Anonymous Referee #2, 12 Feb 2023
The manuscript content is mostly technical. It presents the main steps taken to construct a complex operational ocean forecasting for the Mediterranean Sea, which contains physical, wave and biogeochemical components. It should be highlighted that each component has its own data assimilation system, so that important effort was made to extract the most relevant information from observations to benefit the system forecasting skills. The main goal of the paper is to present the current quality of the operational system components by comparing the analysis and - for specific variables, such as significant wave height - the background (forecast) with observations, in situ and/or by satellites. In the text (L350) it is made clear that only the analyses will be evaluated and that the system short-range predictability will be assessed in a future work. However, the WAV and BIO components were also verified by using the background, i.e., the short-range prediction. Please, provide the adequate information with respect to this emphasis.
The text is very well written and contains a broad range of references to works that led to the forecasting system construction. However, it would be useful to add a new reference, by Napolitano et al. 2022, (https://doi.org/10.3389/fenrg.2022.941606) about a physical and wave forecasting system for the Mediterranean Sea that uses the MED system as initial condition and lateral boundary condition. It is another relevant use to MED system.
Please, it would be useful if more information is offered about the 2 way coupling between NEMO and WW3 and the wind forcing (L110). Does the speed of the ocean currents are considered to calculate the vertical momentum flux? Clementi et al (2017) paper is referred to for more information, but if you could give here this information it would be useful.
Also, despite using the monthly climatology for the river runoff inputs, the salinity at the river mouths are kept constant along time. Are there measurements that corroborate to this condition? At least at the mouth of the rivers with the largest fluxes, do you know about salinity variability from intraseasonal to interannual scales. Please, include a phrase commenting this condition.
With respect to the data assimilation systems employed in the PHYS, WAV and BIO components, is superob utilized? Does the system has this capability? It is very common the use of superob for the high resolution SST or longwave radiation data and SLA data. Please, mention in a short phrase if it is employed or not and why.
You mention that in WAV forecasting cycle, the model is initialized 24 h in the past. Do you use atmospheric analysis forcing during this past period?
I did not understand very clearly the forecasting cycle of the BIO component. Could you please clarify how the nutrients, DIC and oxygen are initialized. You mention (L255) that climatological profiles are used in the model initial condition in each subregion of Fig 3. Does the assimilation of chlorophyl and Argo BCG data change these vertical profiles of nutrient, DIC and oxygen in each forecasting cycle?
The figures are adequately prepared, but I miss a colorbar in Figs. 6 an 8. The work deserves publication, since it will be an important reference for the continuation of the evolution of the system.
Minor comments
L46-47. Please, use MED-MFC or Med-MFC throughout the text.
L65. The period 2017-2020 should be corrected to 2018-2020.
L78. Include a reference for the OceanVar.
L145-146 “SLA along track observations shallower than this depth are not assimilated”. I understand what you mean, but it would be better to rephrase as “SLA along track observations over waters shallower than this depth are not assimilated”.
L170-171. Please, fix the parenthesis used in the references.
L337-340. Please, you may use “three major improvements of the BFM model included: (i) the addition of ...; (ii) the revision of ... and so on.
L357-359. Please, clarify what you mean by “daily mean analysis products”. I understand you produce only one analysis per day with a single analysis increment at a specific time. Therefore, I do not understand how you can take daily means. You can take, for instance, annual means from daily outputs, but not daily means. In line 363, also refer to daily mean analysis.
L414. The skill of the WAV component is assessed both with the analysis and the background, but in L350-351 you have mentioned that the forecast skill would be assessed in a future work. Please, clarify the components that will be here evaluated only with analyses and with analyses and forecasts.
L431. Substitute “forcing wind model” by model wind forcing
L435. The unit is missing after 0.13
L442. Remove “is” from the phrase “that ECMWF is forcing underestimates”
Table 6. Please, correct the entry Phosphate RMSD x 0-10 m and superscripts of the variables Phosphate and Ammonia. The unit of the layers (m) is also missing.
L595. Replace WAB by WAV
Citation: https://doi.org/10.5194/egusphere-2022-1337-RC2 - AC3: 'Reply on RC2', Giovanni Coppini, 05 Apr 2023
-
RC3: 'Comment on egusphere-2022-1337', Anonymous Referee #3, 20 Feb 2023
The paper is of great interest to the oceanography community in the Mediterranean and not only, providing the up today updates of the three components of the MFS, i.e., the hydrodynamical, wave and biogeochemistry modelling systems, as well as of the DA. These three modules consist the Copernicus Mediterranean monitoring and forecasting system (CMEMS Med MFC). The quality assessment of the reanalysis’s products of the CMEMS Med MFC between 2018-2020 was evaluated based on in-situ and satellite remote sensing data using well accepted statistical indexes.
The paper is very useful also to the Mediterranean teams operating the national operational coastal forecasting system, downscaling from the CMEMS Med MFC.
I recommend the publication of the ms, after minor modifications, proving the clarifications and additions mentioned here below, as well as to consider the comments from the two anonymous referees.
- Why is needed 141 vertical levels in the hydrodynamical model? What is the benefit comparing to a 100 vertical levels? What are the criteria to use 141 and not less vertical levels?
- Clarify, why the need to use two different waves models? WW3 and WAM? If indeed is necessary, then, are there any inter-comparison between the results of the 2 wave models used?
- Before mentioning that " Sea ice coverage fields are also obtained from ECMWF” clarify that this parameter (sea ice) concerning the North Atlantic domain used for the lateral boundaries of the biogeochemical model. The reader is confusing as it is appeared in the current text without any prior explanation.
- Provide a paragraph or sub-section describing the cal/val of the surface forcing used in the CMEMS Med MFC, provide a relevant plot if available.
- There is no information what will be included in the Part II of the ms. Due to the fact that the CMEMS Med MFC products are used for operational downscaling and down-streaming in the Med-Sea, provide a paragraph mentioning the most known Mediterranean national operational downscaled coastal forecasting systems using the CMEMS Med MFC, as well as few successful down-streaming applications where the CMEMS Med MFC and the downscaled coastal systems were used (2 sound examples).
Citation: https://doi.org/10.5194/egusphere-2022-1337-RC3 - AC4: 'Reply on RC3', Giovanni Coppini, 05 Apr 2023
Giovanni Coppini et al.
Giovanni Coppini et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 443 | 193 | 25 | 661 | 4 | 5 |
- HTML: 443
- PDF: 193
- XML: 25
- Total: 661
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1