General comments

This paper presents a novel way to predict rip current and shore break wave hazards, providing a new means to forecast these coastal hazards ahead of time. It uses a relatively novel physics-based approach to predicting these processes. The methods have been calibrated and validated over a single summer season at a beach in France using lifeguard perceived hazard estimates, indicating that the system performs very well against those test data. This contribution has the potential for wide reaching impacts in coastal hazard prevention through forecasting rip and shore break hazards (which is alluded to in the introduction), as long as the authors can address some of the issues I raise below regarding transferability of the approach.

The paper addresses a relevant scientific question that is within the scope of the journal. The approach is novel for the purposes of hazard prediction and the methods are clearly outlined. The results are sufficient to support the conclusions reached.

In the introduction you argue that previous approaches have been 'validated on a limited number of beaches' and that 'more generic surf-zone hazard models' are required 'to be applied to a wide range of sandy beaches'. Given this justification for developing a new method of rip forecasting, I think more discussion is warranted on how feasible the methods would be for a large number of beaches (e.g. a national-scale rip forecast) and how transferable the methods are for non-lifeguarded beaches. I.e., is this a method that's useful and accurate but is only really feasible for a handful of high-risk sites where it's worth undertaking a lengthy calibration effort to gather lifeguard perceptions, or is this a method that can realistically be generalised and tested on a regional or national-scale? If so, briefly indicate how you would propose for this to be done?

Specific comments

Line 53: consider citing Stokes et al. (2024) 'New insights into combined surfzone, embayment, and estuarine bathing hazards' here for a current reference. They predominantly forecast estuary hazards, but they also forecast rips with a process-based model

Line 57: You should also cite Scott et al. (2014), 'Controls on macrotidal rip current circulation and hazard' here, which was their earlier paper describing the data and analysis, while this reference describes the application to lifeguarding. I would also mention that this final approach relied on lifeguard incident data for calibration/validation of thresholds

Lines 70-72: This could do with rewording - it's not been mentioned yet what the free parameters are, which in itself is not a problem, but it makes it confusing when you mention setting up the parameters with either beach morphology (presumably an input parameter) or lifeguard-perceived hazard data (presumably the target variable). Consider briefly summarising what the free parameters are, or explain how they relate to beach morphology and lifeguard perceived hazard.

Lines 72-73: This may be dealt with later in the discussion, but you mention earlier that a shortcoming of previous systems is that they were calibrated on only a few beaches, and that more generic models are needed to apply to a wide range of sandy beaches. However, in this study you only use a single beach. I would suggest in this final paragraph of the introduction you should manage the readers expectations - is this new system demonstrated on a single beach, or is it shown to be applicable to a wide range of sandy beaches?

Lines 106-107: While the specifics of this Wavewatch model are not fundamental to your conclusions, it does influence your results to some extent, and I therefore think slightly more detail about the wave model is warranted. It would be good to know the extents of this model - is it a local area model or is it simulating waves along the entire French coast? How far offshore does it extend? Was it developed for this study? If this information is in another paper, then you can cite that instead.

Line 106: I suggest re-wording the sentence slightly - Wavewatch can use an unstructured grid, but it doesn't have to use one. I would briefly mention the min and max grid resolutions used and explain that the unstructured grid allows computational efficiency. You could after all resolve the French coast at 200 m with a regular square grid (although you wouldn’t necessarily want to!).

Line 110: As you've mentioned that the model can simulate coastal processes such as wave breaking, I think it's worth mentioning here  explicitly that the model is not being used to simulate surfzone conditions - it is being used to estimate wave conditions just outside the surfzone.

Line 112: Please state over what time period the results were gathered.

Line 134: I’m not sure I agree with how this is worded – rather than saying ‘alongshore variability in depth of the sandbar’ isn’t it more accurate to say something like ‘alongshore variability in depth between the sandbars and intervening drainage channels’

Line 147: The Larson et al. (2010) equation is absolutely fine for the purpose it's being used, but you should give some mention to what the equation includes and excludes. As a minimum, you should mention that this is a simple linear shoaling equation with a refraction law included, and assumes a simple linear seabed slope. It doesn't consider complex (barred) surfzone slopes or bed friction. To be more complete, you could include the full formulation (which is arguably appropriate, given that it is a key equation in your predictive system).

Equation 2: it is not completely clear from this formula which Hs is being used here - i.e. is this Hs at breaking, or does this Hs need to be locally defined? Please clarify this in the preceeding or proceeding text.

Equation 4: Intuitively, I would assume the velocity in the rip channel will depend on the gradient in wave setup between the bar and the channel, not the absolute difference between the two. However, I see from your diagram below how this can be neglected in an idealised case. As these gradients are present in equation 3, please explain in conceptual terms (and referencing the figure below) why the x and y dimensions can be safely neglected and are no longer important in order to estimate the flow velocity.

Line 171: Wave power at breaking can be derived from the wave breaker energy (Eb = 1/8 rho g Hb^2) and shallow water group velocity (Cg = sqrt(g hb)) as Pb = EbCg. This would seem a more obvious way to compute breaker wave power. I assume your motivation to derive a different formula here is because the local beach slope is not considered in the Larson (2010) calculation of Hb, and that this therefore provides an inferior estimation of wave power if you happen to know (or can estimate) the local beach slope. Please explain and justify in the text why you choose this approach over a more common airy wave theory approach.

Line 175: This equation is commonly presented as h = Ay^(2/3). Please cite where this version comes from and define how you use this equation. For example, a is usually a sediment dependent scale parameter and b is usually taken as 2/3; How have you defined them in this application?

Line 184: ‘in between, offshore wave breaking occurs’ - I suggest using a different term here as ‘offshore’ sounds seaward of the surfzone/ sandbars. However, I assume you are referring here to wave breaking on the sandbar?

Line 191: The quantile-quantile approach you used to transform your values into a 5-level scale warrants explanation in the text. How exactly was this done?

Line 195: Please explain in more detail how thresholds were determined to distinguish the five levels.

Line 247: to be completely transparent, the top 4 lifeguard perceived hazard values are understimated by the model. However, the correlation and performance is generally very impressive.

Discussion: Another point that may be interesting to investigate (although entirely optional) now that you have well calibrated models, is what proportion of time this beach exists at each hazard level. This would simply require running the models over a longer time frame (a few years of wave and tide data, for example) and plotting the distribution of different hazard levels for rips and shore break waves. As I say, this is an entirely optional suggestion.

Line 266: Another approach that is probably worthy of discussion and that follows from previous papers (scott et al (2014), for example) would be to test the developed models against lifeguard recorded incident data. This would have benefits and limitations (e.g. mixing risk and hazard), but it would be interesting and valuable to see how well the models pick out periods of incident occurrence. This may be one of the only feasible ways to test the model’s applicability on a large number of beaches, where gathering lifeguard perceptions may not be so feasible.

Lines 266-267: ‘The validation approach proposed here can be applied anywhere pending lifeguard hazard assessment can be performed’ - This is not a trivial undertaking! Can you comment on how many lifeguard observations would be required at each site to robustly tune the models?

Lines 288-289: For completeness, can you comment on how Wf performs when applied hourly? Presumably, the model you present here performs better at hourly resolution as it captures tidal variation?

Line 291: daily mean hazard would also help lifeguard managers roster lifeguards, some days ahead, to the beaches where they will be most needed

Line 304: This is an important point, as it suggests that the predictive method is not sensitive to the sort of changes in the bar and channel that might be expected within a single season. Can you comment on what a typical range of bar elevations and channel depths are expected to be (at least at this beach)?

Line 318: Please comment on the calibrated gamma value you found in this study - it is significantly lower than would be typically expected. How sensitive is the model to this value? What correlation would be obtained if you used a more typical value for gamma?

Line 348: This sounds like it's a limitation of your study, but that's only true if predicting overall risk is of interest (for lifeguard resourcing, for example). I suggest re-iterating here that even without predicting exposure the present system provides useful prediction of the underlying level of hazard, which is the primary factor of interest to both the public and lifeguard services.

Conclusions: As per my general comment regarding the discussion of transferability to other sites, I think the conclusions need to at least briefly address how feasible it is to apply the developed models at a large number of sites. i.e. how can these models feasibly be calibrated/validated at other locations? You need to be more realistic about how feasible it is to collect the required parameters at other sites. The rip model morphological parameters (bar crest depth and channel depth) are 'simple', but they are not trivial to measure. As the authors know, the surfzone is notoriously difficult to survey and is not routinely surveyed by monitoring programmes, and only at select locations globally is measured occasionally for research purposes. Therefore, the conclusions need to briefly address how these parameters are expected to be gathered for future application of these models, especially if they are to be useful for a large number of sites. Are the calibrated values used here expected to be applied elsewhere (along with some form of validation)? Are the parameters expected to be re-calibrated at each new location against lifeguard observations? or perhaps through direct surveying of the surfzone morphology?

Technical corrections

Line 2: Change ‘expose’ to ‘be exposed’

Line 2: I suggest changing ‘The most severe and widespread natural hazards’ to ' The most severe and widespread natural bathing hazards'

Line 8: Change ‘from July 1 to August, 2022’ to ‘during July and August of 2022'

Line 9: Change ‘into’ to ‘into a’

Line 12: Change 'where wave forecast is available’ to ‘where a wave and tide forecast are available’

Line 39-40: Rather than ‘due to alongshore-variable sandbar depths’ I think it would be more accurate to say something like 'alongshore variability in depth between the sandbars and intervening channels'

Line 52: Change ‘increased understanding in rip current dynamics’ to ‘increased understanding of rip current dynamics’

Line 69: Change ‘quantitative estimate’ to ‘quantitative estimates’

Line 81: Change ‘rip current are ubiquitous’ to ‘rip currents are ubiquitous’

Figure 1: The text within panel (b) doesn't show up well unless you zoom in on it. I would suggest changing the text to another colour. In the caption below the figure the abbreviation ‘SMGBL’ should be spelt in full on it's first use. This would also make it more consistent with the previous photo credit

Line 103: Change ‘operate surf-zone hazard forecast’ to ‘operate a surf-zone hazard forecast’

Lines 103-104: Change ‘we used numerical wave hindcast’ to ‘we used a numerical wave hindcast’

Line 133: Change ‘the deeper channel’ to ‘the deeper channels’

Line 148: Change ‘consider simple’ to consider a simple’

Lines 155-156: U, V, and w should all be defined here (currently only V is defined). Also, please specify where h should be defined as you have a depth over the bar and a depth in the channel. Which is this h supposed to represent?

Line 159: change ‘proceeds as follow :’ to ‘proceeds as follows:’

Line 160: you should define h_b and h_c here

Figures 3 and 4: the x and y axes of (d) are not clearly defined. Also, gamma appears here as Y which on first reading seems like a new parameter.

Line 167: Change ‘for break type’ to ‘for breaker type’

Line 171 and equation 6: Parameter names change from Essb and Hsb here to Esb and Hssb below. Please check all parameter names are consistent.

Line 180: it is not clear where this wave height is being defined. I assume this is breaker height at the sandbar? If so, it would be clearer and more consistent to refer to this as Hsb (as per earlier definition for rips)

Line 181: Change ‘model proceeds as follow :’ to ‘model proceeds as follows:’

Line 190: Change ‘Hsl’ to ‘SHl’

Figure 6 caption: Change ‘pdeth’ to ‘depth’

Line 228: Change ‘RHl’ to ‘SHl’

Line 264: Change ‘perception can influenced’ to ‘perception can be influenced’

Figure 11 caption: the terminology of ‘shore-break wave intensity I’ is inconsistent with the parameter naming used up to this point (Esb)

Line 305: Change ‘are modified based a the quantile-quantile’ to ‘are modified based on the quantile-quantile’

Figure 12 caption: Change ‘with the blue (dotted) blue lines’ to ‘with the solid (dashed) blue lines’

Line 323: Change ‘rip current tends’ to ‘rip currents tend’

Lines 330-331: Change ‘it provides a direct Information on’ to ‘it provides direct Information on’

Line 351: ‘allow to compute the time evolution’ -  I think it would be fairer to say ' allow to estimate the time evolution'