Review of: Improved winter conditions in SURFEX-TEB v9.0 with a multi-layer snow model and ice for road winter maintenance

This article addresses the implementation of a more advanced snow and ice model in the SURFEX TEB model. The intention is to improve TEB in urban environments, where it is used, for use in road weather modelling. As the title indicates, the modelling should improve the description of winter road conditions for road winter maintenance.

Unfortunately, the improved model falls far short of achieving this aim. This is mainly due to the lack of processes included in the model. These are listed by the authors at the very end of the paper.  ‘Traffic heating, salting, water splashing, and snow compaction impact the road surface conditions and the road surface temperature.’ I would add snow spraying and snow ploughing to this list as well as an improved description of water drainage. Since the authors include none of these processes then the model will not be able to reproduce real world winter road conditions.

Of the two experiments used for validation only the Col de Porte experiment is relevant for this model. However, this site is not a road and so has little relevance to the title of the paper. What the authors seem to have done is to have improved the TEB model so it can better model snow, but not made it relevant at all to roads. The reviewer presumes this to be one of the next steps.

Since the model does not include traffic or winter road activities the Hajala site was not suitable for model comparisons, where the aim of the model comparisons was to test the new snow model routine. The only reasonable comparison would be the surface temperature, which the model seems to reproduce quite well. This section can be significantly reduced if the authors use it more to demonstrate that the model is not suitable for road applications yet, rather than to try to find moments in the data which actually are reproduced in the model.

Even so, there is clearly a process involved in this model development and this paper would appear to be the first stage in a general application of the TEB model for applications concerning road weather. This reviewer then suggests that either the authors wait until they have implemented these relevant processes or that the paper be rewritten in a way that makes this article the first step in several model improvements.  This can be made clear in the title of the paper, for instance using a phrase like ‘Towards improved winter conditions … for road weather applications’. The authors would be wise to state what the process will be, including their next steps in making the model practically useable on real roads and use the Hajala site to demonstrate these needs.

Generally the article is well written, though there were a large number of small mistakes in the text. This should be proof read before the next submission.

Comments to the paper have been inserted in the original pdf as notes, attached.

This reviewer recommends major revisions.

Review of “Improved winter conditions in SURFEX-TEB v9.0 with a multi-layer snow model and ice for road winter maintenance” - Colas et al. 2024

General comments:

This paper details developments to the Town Energy Balance model, ‘TEB’, to improve the prediction of snow and ice on road surfaces. Specifically, a new representation of surface ice is added and the snow scheme in TEB is replaced with the multi-layer snow model ‘Explicit Snow’. The previous and updated versions of TEB are tested at two sites – an experimental road surface at Col de Porte in the French Alps, under controlled conditions, and a road with live traffic at Salo Hajala in Southern Finland.

The evaluation of results at Hajala clearly presented challenges as the impact of traffic and road salting are not represented in the model and the timings of snow clearance were not precisely known. However, the authors acknowledge this and conclude that it is necessary to incorporate these aspects into future modelling efforts. This seems like a valuable conclusion.

Overall, the improved version of TEB was found to perform better than its predecessor at predicting road surface temperature and snow depth. It also outperformed the current operational road weather prediction model in France, ‘ISBA-Route/CROCUS’, which was used as a benchmark at the controlled experimental site of Col de Porte. Given CROCUS provides a more detailed snow scheme, this finding does in principle support the authors’ premise that an urban model would be better suited to road weather prediction although it is unclear why the benchmarking was not also performed at Salo Hajala. If the finding is reproducible at the real world site then it would be a much stronger argument.

This work presents a useful advance to road weather modelling efforts as well as highlighting further areas for improvement. However, the findings could be strengthened with some further analysis as well as considerable improvements to the clarity of the manuscript. As such, I recommend major revisions.

Major comments:

1. Why were the two benchmarks not evaluated at Hajala? Given the apparent unsuitability of either TEB or TEB-ES to predict conditions at a real world site, it would be helpful to know if an operational road forecasting model, such as ISBA-Route, performs any better. Ideally results from the two benchmarks should be shown for the Hajala site but if that is not possible it should be briefly explained in the text why not.
2. More detail is needed to explain what data was used to drive the model simulations at Hajala. Did you use the observations directly from the synop station at Hajala or output from either the Kriging process or the HARMONIE-AROME model runs described by Karisisto and Loven 2019? Either way, is this data publicly available and where can it be found (it does not appear to be included at the link provided in your data availability statement)? If the input data is not available then this needs to be made clear in your data availability statement. You also mention that “The hourly atmospheric forcing for the model consisted of a mix of observation data and ERA5 reanalysis” – please give more detail as to how the ERA5 data was mixed with the observations. Was this for gap filling? How much ERA5 data was used? Was it for all or just some of the fields?
3. It is not clear in the text whether section 2.2 is a description of an existing treatment of ice in TEB or a new treatment for TEB-ES. Authors need to make it clear which aspects of the model are novel in this work.

Minor Comments:

1. Abstract: A major finding of this study was that physical models need to include representation of various effects due to traffic and road maintenance activities to be able to accurately predict real world road surface conditions. This finding should be included in your abstract.
2. Line 134 and 135: I’m not sure I understand, if ice content growth is limited by water availability then what other limitations are being referred to when it says there are none? Perhaps the first sentence means “Contrary to the liquid water content, ice content can grow on both snow covered and snow free fractions of the road surface”?
3. Figure 1: It might be useful to have a diagram of original TEB infrastructure for comparison.
4. Line 165: Is that the correct reference? Vionnet 2012 describes a multi-layer snow model but does not compare it to a single layer one. A need for multiple snow layers is a stated motivation of that paper but is not well explained. They reference Etchevers 2004 which found that more ‘complex models’ (assessed based partially on the number of snow layers) were better able to predict the net longwave radiation flux.
5. Line 170: At least 3 layers – why? Is that based on a reference?
6. Line 249: Why do you change the way that natural soil is initialised in TEB-ES? What impacts will this have?
7. Line 250: It’s not clear what is meant by the term ‘water wear-off’ and it does not appear in the referenced paper. Please can you clarify what this means – is it the same as runoff?
8. Line 261: You mention that six different surfaces are installed at the site but only show one set of results for Col de Porte – did you just use observations from one surface or was it a mix of all six? Was there any performance differences associated with the different pavement types?
9. Line 282: Is it possible to quantify the errors associated with these measurements?
10. Line 297: I don’t understand the second half of this sentence, please clarify.
11. Figure 4: The observed snow depth occasionally appears to show increases when there is no observed precipitation – why is this? (Particularly small blip in observed snow depth on November 12^th and large increase on November 17^th).
12. Line 324: On what basis do you say that the precipitation forcing was wrong? Is there another observation source for precipitation phase? Presumably the partitioning of precipitation phase on the basis of air temperature isn’t always accurate, are you able to comment either here or near line 269 as to how appropriate this assumption is and how often it fails?
13. Tables 3, 4, 6, 7, and 8: What’s the difference between ‘false detection rate’ and ‘false alarm rate’? Does one of these columns correspond to the ‘true negative’ scenario referenced in line 350? It may be worth renaming one of these columns for clarity and/or describing what the headings mean in the table caption.
14. Figures 6b and 7b: The scale is not sufficient to distinguish changes in the observed SWE. The trends described in the text on January 21^st and 22^nd cannot be seen in this plot because the values are too small to be distinguishable on this scale. Consider increasing the height of the panel or an alternative method to ensure that the trends described in the text are visible in these figures.
15. Line 388-399: TEB-ES also appears to melt snow considerably slower than the observations. I think this sentence needs to be re-phrased to avoid overstating the performance of TEB-ES in this instance.
16. Line 451: Figure 7 does not show observed air temperature, only road surface temperature.
17. Line 460: I agree with the first sentence of this paragraph but it needs more detail to support it and it also seems unrelated to the rest of the paragraph. I would expect a separate paragraph here discussing both the reduced reliability of observations at Hajala and also the impracticalities of using live road observations to validate a model that doesn’t represent those conditions.
18. In your zenodo dataset, the readme file states that raw observations and validation data can be found in the zenodo dataset associated with Karisto and Loven’s paper but I can only see observations for validation data there – the only files containing the atmospheric forcing fields you mention are model output from HARMONIE simulations. You also state that ‘raw observations for SW and LW ERA5 radiation are available’ but ERA5 is not raw observations – it is a reanalysis product that involves a lot of modelling. Please be very clear, both here and in your paper, what data you are using comes from observations and what comes from models/reanalysis.
19. The zipped data file in your zenodo dataset appears to be corrupted – when I try and unzip it I get error messages saying the file is ‘invalid’ or ‘empty’.

Technical corrections:

1. In general there are numerous instances of inaccurate grammar and sentences that could be written more clearly. I suggest that with further proofreading, this manuscript could be edited to become much easier to comprehend.
2. Line 118: remove first instance of ‘water’
3. Line 156: Need to distinguish that M is the melting rate of ice as opposed to the melting rate of snow (Rmelt) which is also used in that equation.
4. Line 162 (eqn. 9): Is d_r1 the same as d_R1 used in eqn. 2? If so, these need to be the same symbol, if not, need to specify what d_r1 refers to.
5. Line 342: Do you mean TEB-ES is more accurate than ISBA-Route/CROCUS rather than TEB? TEB-ES outperforms ISBA on most statistical measures in table 2 but TEB does not.
6. Figures 5 and 8: Increase font size
7. The figure caption for figure 9 is very difficult to comprehend, especially the last sentence.