High-Resolution COSMO-CLM Modeling and an Assessment of Mesoscale Features Caused by Coastal Parameters at Near-Shore Arctic Zones (Kara Sea)
Round 1
Reviewer 1 Report
Overall, well written paper but needs some clarifications. I will accept it with some major changes and listed below.
- why not compare the mean values in the table and figs? like to see figs or tables for mean values, not only biases
- why not show vertical air velocities? is there any diff between cloudy and cloud free conditions?
- what caused the strong winds?
- how can you explain point measurements versus model grid area diffs? what are the scale issues that affect biases?
- see Gultepe et al (JGR) for explaining ice surface effect on clouds/fluxes. impact on turbulence fluxes are critical for dynamical stability. Also see J of Atmosphere& Ocean for turbulence fluxes.
- how large scale conditions affected weather conditions e.g. fronts or low pressure etc.
- finally you need a discussion section for uncertainties/issues.
Author Response
Response to Reviewer 1 Comments
Point 1. why not compare the mean values in the table and figs? like to see figs or tables for mean values, not only biases
Response 1. Thanks for the Reviewer for this issue, but from our side, this needs clarification. Tables are presenting average statistics over all 15 stations based on comparison of observed and modelled values in the corresponding grids for two months. Figures are presenting an example of probability distribution functions at several stations, and wind speed patterns in case studies for specific times. In our opinion, it doesn’t make sense to present any long-time mean values of wind speed as well for statistics in Tables, as in any Figures. Probably, authors didn’t understand this remark thoroughly, and pleased to clarify this issue. Thanks!
Point 2. why not show vertical air velocities? is there any diff between cloudy and cloud free conditions?
Response 2. The focus of this paper is on the surface wind speed patterns. Of course, authors keep in mind that wind speed has a complicated 3-dimensional structure, vertical velocities are very important in the many mesoscale circulations’ developments including downslope winds, mountains overflow, interactions between flow and land-sea surfaces, etc. However, vertical velocities are not the main features in the cases of flat islands overflows (e.g., Belyi and Dikson islands) considered in the paper on the given resolutions (~ 3 km grid). Many features need to engage not only vertical velocities and surface winds, but other boundary layer levels also, and some hydrodynamic parameters of flow and so on. This is not the main objective of this paper. The paper is devoted mostly to investigation of model capability to reproduce surface flow characteristics and its transformation by interaction of flow with any obstacles, coastlines, islands, land-sea transition, many mesoscale features developing in this interaction. Firstly, we would like to demonstrate the capability and success of the model in reproduction of many these features in surface wind speed patterns. One of the next steps of our research will be 3-dimensional analysis including the Reviewer’s suggestions. On our opinion, this goes beyond this paper at the moment.
Point 3. what caused the strong winds?
Response 3. Authors have given a short description of general synoptic situation in each case and provided some maps and the corresponding vertical sounding profiles in the Supplementary Materials. Authors described synoptic conditions more detailed in text, according to explanation of the strong winds causes. The corresponding text was added in lines 428 – 431, 538 – 539, 591 – 592.
Point 4. how can you explain point measurements versus model grid area diffs? what are the scale issues that affect biases?
Response 4. Differences between observations and modelling data in the nearest grid point could have various sources. There are inconsistencies of surface (land, sea, soil type), altitude, long distance between considered points. This issue is very important in atmospheric modelling and these errors are inevitable, which refers to an additional research and paper, in our opinion. Generally, these differences are among the main issues discussed in paper. The Discussion section demonstratese some cases of inconsistencies between modelled and observed wind speeds, including comparison examples of 12 km and 3 km grid patterns for the Malye Karmakuly case. Scale of islands, obstacles and coasts curvatures are demonstrated almost in all cases. Generally, it was shown quantitatively using statistics of verification results that 3 km grid simulations capture more mesoscale features of 3 – 10 km size quite better than 12 km ones.
Point 5. see Gultepe et al (JGR) for explaining ice surface effect on clouds/fluxes. impact on turbulence fluxes are critical for dynamical stability. Also see J of Atmosphere& Ocean for turbulence fluxes.
Response 5. Thanks to the Reviewer for appropriate references. It is concerned to the turbulence fluxes, its interactions with clouds and ice surface. However, all considered cases in paper are not concerned with sea ice conditions, which was checked previously. One of the reasons in experiments periods choice was sea-ice free periods to investigate not only wind speed patterns, but waves also. The latter was not included in this paper at the moment. Therefore, in our opinion, it is not necessary to analyze the issue of sea ice effects on clouds and fluxes. The Reviewer is right that turbulent fluxes are critical to dynamical stability, cloud formation, vertical mixing and some 3-dimensional flow properties. As authors have mentioned in Response 2, the 3-dimensional analysis was not included in the scope of tasks of this paper. Since the focus variable is the 10 m wind pattern, the complete thermodynamical case studies did not carry. In our opinion, including this analysis in the paper, as well analysis of Point 2, will shift the subject of paper to the complete case studies physical investigations, and move away from surface wind speed analysis and model capability to reproduce it.
Point 6. how large scale conditions affected weather conditions e.g. fronts or low pressure etc.
Response 6. This issue is linked to the Issue 3. Authors will give detailed synoptic overview for each considered case including frontal analysis, cyclones movements, etc., and its transformation on mesoscale. As for general large-scale conditions of these cases (i.e., synoptic climatology), authors have referred to the corresponding papers (e.g. according to the Novaya Zemlya bora). Moreover, authors would like to note the role of spectral nudging technique in the large-scale conditions’ reproduction. Spectral nudging is the best choice to assimilate large-scale meteorological fields additionally, inside the model domain. There are many papers described a notable positive effect on model success in large-scale variables reproduction (Von Storch et al, 2000; Feser, F., Barcikowska M., 2012; Miguez‐Macho et al, 2004, etc.). Tables 1 and 2 have shown the significant improvement of simulation results, Figure 4 gives a nice example of this improvement.
Point 7. finally you need a discussion section for uncertainties/issues.
Response 7. In most cases we have completed the Discussion section placing the abovementioned corrections and issues mentioned by the Reviewer 1 including the issues from the second Reviewer concerned to this section. The text was added and changed in lines 103 – 107, 440 – 442, 468 – 471.
Reviewer 2 Report
The recent rapid climate change has caused serious regional environmental changes in the coastal Arctic region. However, it is still a very challenging task to accurately simulate physical/dynamical processes over this region. This study attempted to simulate the mesoscale circulation under various surface conditions in the Arctic coastal region using the COSMO-CLM model. It seems that authors have tried to find the optimal condition by comparing model experiments of different options. Also examined are various mesoscale phenomena through various case studies. I think the purpose of the study seems to be very timely and appropriate for the scope of Atmosphere. However, in order to recommend the publication of this paper, I think a revision is necessary considering the following points.
The researches on the Arctic regional climate system mentioned in the introduction mainly occurred in winter, but the simulations were run mainly for summer-autumn. It is also not the season of major storms occurred in the Arctic region, which has been of interest in many recent studies. More explanation is needed for the experiment period.
More objective validation of the model performance is required. I wonder authors may say really “realistic reproduction” with the presented values in Table 1 & 2. Even though ERA-interim is used as an initial & boundary condition and spectral nudging is used, the presented model performance is not satisfactory in my view. I would like to see a more objective value or, if possible, a comparison with other models.
Many typos, non-conventional terms, and inappropriate use of references are found. Elaborate editing is required. Figures and table captions are also insufficiently explained, and units are missing. For example, in Table 1, I understand that the values are from 10m wind, but should be noted again here in Table caption. The correlation coefficient for what? Please change, to. unit should be given. In Figures 5, 6, 7, and 8, is the right panel just enlarged?
This paper has almost the nature of a report. In particular, the Discussion section is too verbose and difficult to grasp the point. Overall, the discussion part should be more organized and concise.
Author Response
Response to Reviewer 2 Comments
Authors thank the Reviewer for some conceptual and useful issues and high estimate of timeliness and suitability of this paper for the scope of Atmosphere.
Point 1. The researches on the Arctic regional climate system mentioned in the introduction mainly occurred in winter, but the simulations were run mainly for summer-autumn. It is also not the season of major storms occurred in the Arctic region, which has been of interest in many recent studies. More explanation is needed for the experiment period.
Response 1. The reviewer is right, of course, the main stormy season in Arctic is late autumn – early spring. However, there were really some reasons to select the late summer – autumn period. The first one was sea-ice free conditions, and therefore we could analyze the sea-land transitions and its influence on the surface wind speed patterns clearly, without any other side impacts. The second one was concerned to the next application of these simulations. We have done some waves simulations with WaveWatch III model according to COSMO-CLM modelled wind fields, but did not include it in this paper, because we would like to prepare another article in the near future. Ice-free conditions are the best to test wave propagation simulation, therefore we have chosen period of August-October as well known for being the most ice-free of the year. At the same time, we have chosen this period taking into account observed high wind speed events (including the Novaya Zemlya bora) caused storms. The text was added and changed in lines 281 – 285.
Point 2. More objective validation of the model performance is required. I wonder authors may say really “realistic reproduction” with the presented values in Table 1 & 2. Even though ERA-interim is used as an initial & boundary condition and spectral nudging is used, the presented model performance is not satisfactory in my view. I would like to see a more objective value or, if possible, a comparison with other models.
Response 2. The Reviewer has pointed that verification results are not satisfactory including comparison with reanalysis. Authors have mentioned that in our simulations we did not use any additional assimilation of station, satellite, radar and other data, since it is assimilated yet in reanalysis. Considering this fact, we can conclude the verification results are satisfactory, in our opinion. However, there are usually no overall advantages of regional long-term mesoscale modelling over modern global reanalyses in average values, but its main advantage is in extreme values simulations (e.g., Shestakova et al., 2020a, [92] in paper). This latter fact was demonstrated in our paper using detailed case studies. We could not to compare with other model simulations, excluding references on some WRF papers over this region (e.g., Shestakova et al., 2020b, [76] in paper), which confirm the abovementioned issues generally. It was shown also in many articles [Serreze et al., 2012; Tilinina et al., 2012; Akperov et al., 2018] that reanalyses and regional mesoscale long-term modelling estimates have differences of the same order as in our paper. We have used the standard statistical metrics and verification methods to estimate model performance, perhaps some other estimations would fit better. It is also necessary to mark out that Kara region is covered by surface observations quite poor, we have used as much data as possible. The text was added and changed in lines 456 – 460.
Point 3. Many typos, non-conventional terms, and inappropriate use of references are found. Elaborate editing is required. Figures and table captions are also insufficiently explained, and units are missing. For example, in Table 1, I understand that the values are from 10m wind, but should be noted again here in Table caption. The correlation coefficient for what? Please change, to. unit should be given. In Figures 5, 6, 7, and 8, is the right panel just enlarged?
Response 3. Authors did their best to fix and correct all found typos, inappropriate references, etc. and elaborated text, including Figures and Tables captions. Authors would be extremely grateful for the specific recommendations on some references and indications of appropriate terms utilization. List of references was corrected. Tables contents average statistics values over 15 stations. Average correlation coefficients in Tables 1 and 2 was calculated for time series of wind speed (m/s) by stations observations and by the corresponding nearest grids. Biases, RMSE, STD have m/s units. In all necessary cases, units were given. Mentioned Figures has the right panel just enlarged to show some details clearly. The text was added and/or changed in lines 27, 35, 38, 59, 66, 94 – 95, 194, 209, 368 – 369, Tables 1 and 2, 371 – 372, 388 – 389, 549 – 550, 557 – 558, 638 – 639, 640, 767, 949 – 956, line 14 in Supplementary Materials.
Point 4. This paper has almost the nature of a report. In particular, the Discussion section is too verbose and difficult to grasp the point. Overall, the discussion part should be more organized and concise.
Response 4. Thanks to the Reviewer for these recommendations. To clarify, the Discussion section is devoted to consideration different strong wind case studies, investigated issues of model simulation errors indicated in the Results section. Thus, there are discussed and estimated different detailed wind speed patterns caused by flow and surface interactions on diverse scales and resulted in some mesoscale features. Exactly these features are the main source of simulation errors. Maybe, the Discussion section is a little crowded by analysis and Figures, however most of this information is necessary to show and discuss model performance using scale and dynamical analysis, to reveal model discrepancies, ways to overcome them. Perhaps, the Discussion section should be a little more concise, but in our opinion, it needs to be saved in the same form, in general.
Round 2
Reviewer 1 Report
Manuscript seems is improved but i have no responses given to me. For this reason, I cant accept this paper for publication.
Please also add following 2 papers for introduction that emphasize, not only dynamics but microphysics is also critical in the Arctic weather:
Khvorostyanov, Curry et al 2003, 108, D9; A springtime cloud cover over the Beafort Sea...
Also importance of wind/physics in Arctic is studied for severe weather research:
Gultepe et al 2019: A review of high impact weather for aviation meteorology. pure and appl. geophy., 176, 1869-1921.
please modify and submit your responses.
Author Response
Responses to Reviewer 1 Comments
Point 1. Manuscript seems is improved but i have no responses given to me. For this reason, I cant accept this paper for publication.
Response 1. Authors have given 7 Responses to Reviewer’s Points on the previous stage. In many cases there were some objections and explanations, why we can not or did not consider it possible to follow all points marked out by Reviewer (e.g., Points 2, 5, 6). In case of Point 1 Authors have explained that we did not understand the comment thoroughly. However, Responses to Points 3 and 7 included text corrections suggested by Reviewer, which were listed in Response using the lines list (lines 290 – 293, 387 – 388, 435 – 436 in the current text version in “Fixes” mode). Response to Point 4 contained an explanation and discussion of issue.
Point 2. Please also add following 2 papers for introduction that emphasize, not only dynamics but microphysics is also critical in the Arctic weather:
Khvorostyanov, Curry et al 2003, 108, D9; A springtime cloud cover over the Beafort Sea...
Also importance of wind/physics in Arctic is studied for severe weather research:
Gultepe et al 2019: A review of high impact weather for aviation meteorology. pure and appl. geophy., 176, 1869-1921.
Response 2. Thanks to Reviewer for pointing out these references, they were added in Introduction (lines 54, 56 in “Fixes” mode) and in References sections. Microphysics is very important issue in atmospheric motions undoubtedly, including the severe weather features development. However, we would like to emphasize that cloud microphysics is not included in the scope of this work, the focus here is the surface dynamics and wind speed patterns analysis.
Point 3. please modify and submit your responses.
Response 3. According to the abovementioned explanations and discussion, Authors have done some minor changes in text listed above.
Reviewer 2 Report
The authors seem to have addressed my comments (based upon the answer), but I can't find the changes in the places (line numbers) mentioned in the authors' answers. I think it went wrong during the editing process. Please highlight the modified part.
Points: A description of the selection of the experiment period must be included along with references to existing studies.
Response 2: In this model, the reanalysis with observation data assimilated was used as boundary forcing, and spectral nudging was additionally performed. Nevertheless, the skill values (correlation coefficients) are less than reanalysis. It sounds strange to me.
Author Response
Responses to Reviewer 2 Comments
Point 1. The authors seem to have addressed my comments (based upon the answer), but I can't find the changes in the places (line numbers) mentioned in the authors' answers. I think it went wrong during the editing process. Please highlight the modified part.
Response 1. Authors apologizes for some possible editing issues in the “Track Changes” option. All changes and additions of the previous stages were highlighted in text in yellow compared to the first version of paper. In the current text version (in “Fixes” mode) there are lines 35, 38, 45 – 46, 53, 82 – 83, 91 – 95, 107, 181 – 185, 238 – 242, Tables 1 and 2, 253 – 254, 299, 302 – 304, 318 – 322, 330 – 333, 397 – 399, 406 – 407, 474 – 477.
Point 2. Points: A description of the selection of the experiment period must be included along with references to existing studies.
Response 2. The part of paper devoted to the experiments’ period selection was supplemented by the following references describing the wave climate features and storm events: [Stopa et al., 2016; Screen et al., 2011; Parkinson and Comiso, 2013]. The corresponding text with references inserted in lines 181 – 185 in “Fixes” mode.
Point 3. Response 2: In this model, the reanalysis with observation data assimilated was used as boundary forcing, and spectral nudging was additionally performed. Nevertheless, the skill values (correlation coefficients) are less than reanalysis. It sounds strange to me.
Response 3. Authors agree that there are no significant improvements in the best simulation results compared to reanalysis, moreover in most cases reanalysis is better in some skill values (e.g., in correlation coefficients and RMSE, but not in biases). However, there are different subject of discussion. The first point is that there was no additional assimilation of observations in model simulations. Moreover, the spectral nudging technique has captured the large-scale circulation features. We have used zonal, meridional wind speed components and temperature reanalysis fields above 850 hPa level with more than 500 km spatial scales for the spectral nudging technique. In this way, the spectral nudging controls the large-scale circulation within the middle troposphere. The second point, Authors have also noted that the main advantage of regional simulation is the extreme features reproduction, which was discussed in text with examples (in the Discussion section). Monthly averaged statistics could be generally worser reproduced by model because the mesoscale model with finer resolution could reproduce some features more detailed but shifted in space on a few model grids. At the same time, reanalysis data could reproduce some averaged values better, because it represents more broad area over the grid. However, there are usually no overall advantages of regional long-term mesoscale modelling over modern global reanalyses in average values. On the contrary, its main advantage is an extreme values simulations (e.g., Shestakova et al., 2020a, [97] in our paper), some WRF modeling papers devoted to the Arctic region (e.g., Shestakova et al., 2020b, [81] in our paper) confirm the abovementioned issues generally. Considering all mentioned issues, we would like to mark out that our modelling results (i.e., its statistical skill values) are quite typical for long-term mesoscale modeling of given resolution using spectral nudging without any additional observational data assimilation in comparison with reanalysis data. Therefore, Authors would like to conclude that these verification results are quite satisfactory taking into account all discussed circumstances, the modern state of mesoscale modeling and verification approaches.
