Review of “New insights into combined surfzone and estuarine bathing hazards” by Stokes et al.

This manuscript examines surfzone and estuarine flows at Crantock Beach with XBeach simulation. Modeled simulations were calibrated with and tested against observations at this field site. The flows were quantified in terms of bather hazard metrics (maximum offshore flows and particle exits) and used to inform a forecast of bather hazard. This manuscript is mostly well written and clearly articulates the gaps it seeks to fill. However, I have a few concerns, which are discussed below.

Hydrodynamics: The paper describes a range of simulations with varying tidal levels and wave conditions, primarily representative of relatively quiescent conditions (Hs < 1.4 m). However, only results for varying wave power and tide are presented here. I have two concerns regarding this: First, the time series shows several larger wave events and more oblique waves than represented in the selection of simulations. Why did the authors select the set of hydrodynamics to simulate? Are they representative of times when swimmers are at the beach? Considering there are many users in HS3 conditions, should larger wave events be considered? Second, wave power is the only wave property presented here (Hs * Tp). Wave direction is only mentioned as ‘unimportant’ during mid-tide but is not otherwise described in the paper. Presumably, wave direction is important if waves are adequately oblique to reduce the energy entering the embayment. Wave direction is also shown to be important for channel rip hazards (e.g., Dusek and Seim, 2013). Is there little dependence on wave direction here due to the embayment geometry? Additionally, previous work on channel rip currents (e.g., Moulton et al., 2017) suggests that wave height is important for rip speeds but does not incorporate wave period. Assuming wave breaking is triggered at a wave height to depth ratio, wave height defines surfzone width. By presenting these values solely as wave power (which combines both period and height), this paper potentially glosses over some of the dynamics relevant for ‘blocking exits’, etc.

Hazard quantities: The paper primarily relies on two metrics to define hazard: maximum offshore flows and percent surfzone exits. Both are helpful metrics to assess bather hazards. Percent surfzone exits represent a free-floating bather's likelihood of being ejected offshore (but does not represent speed). Maximum offshore flows target how fast a bather may be advected offshore and, therefore, the feasibility for a bather to react (swim) from an offshore flow (but does not represent the offshore flow distribution). These metrics do not represent the number of locations with sufficiently strong offshore flows to eject a swimmer offshore. Providing an additional metric could help. For example, this could be represented as the percent of the locations alongshore with Uoff exceeding a threshold value or, possibly, the Uavg and the rms alongshore of Uoff. This similarly ties into the section on how these flows change with morphology. While Uoff,max does not vary strongly, the position and possibly this distribution of these flows change. This could be explored with an additional hazard metric. 

Hazard forecasts: While most methods and results are thorough, the forecasting bathing hazard section needs to be better described. I found the definitions of different terms and how the hazards were predicted and allocated difficult to follow, especially since some definitions are different in the figure caption. For example, the use of seemingly redundant terms (risk, hazard) and the overly brief explanation of the hazard scoring.

Limitations: The paper should describe the limitations of these hazard predictions. For example, the modeled velocities are often underestimated, resulting in non-conservative hazard estimates. The surf-beat depth-averaged model cannot represent all rip current types (e.g., flash rips). The findings are highly tuned to this specific estuary and may not represent other combined surfzone and estuary flows.

Figure quality: Figures 9 and similar layouts are very challenging to read. The quivers are barely visible, and the figures are grainy. Consider using plots similar to Figure 8c,f.

Line-by-line:

L126: I expect the SfM DEMS to perform well at GCP locations because those locations were input into the algorithm to resolve camera geometry. Thus, this may not represent the accuracy of the DEM well. Can the DEM accuracy be checked by comparing the regions overlapping with both surveys (intertidal zone)?

L133: How was this RMSE computed? By comparing with?

L143: List drifter dimensions. Is it truly a surface drifter or representative of a depth-avg current?

Figure 5: Define acronyms in Figure.

L264: How do these tuned values compare with previous studies?

L336: Report bias since the sign is important here (i.e., if the flows are over or underpredicted).

Figure 8: Why are there large drifter passes at single points in the inner surf zone? Do drifters stagnate there?

Section 4.4: Is this supposed to be a new section?

Figure 10: Error in the x-label -> HsTp/overbar{HsTp}.

Section 4.6: The introduction claimed that the increase in swimmer rescues in recent years may be due to changes in the river channel. Here, the authors could add a simulation with synthetic bathymetry representing the previous, potentially less hazardous river channel to see how it compares with these morphologies.

L555: The spatial variability described here should be incorporated more often throughout the paper.

L600: NOAA’s rip current hazard forecasts consider hydrodynamic conditions (https://oceanservice.noaa.gov/news/apr21/rip-current-forecast.html). Perhaps specify that this is the only forecast model considering estuarine flow.

Reference:

Dusek, G., and H. Seim. "Rip current intensity estimates from lifeguard observations." Journal of Coastal Research 29.3 (2013): 505-518.

Moulton, Melissa, et al. "Comparison of rip current hazard likelihood forecasts with observed rip current speeds." Weather and Forecasting 32.4 (2017): 1659-1666.

General comments:

This observational and modeling study explored swimmer hazards in an understudied setting, where estuary mouth flows encounter surfzone currents including bathymetric rip currents and boundary rip currents with large tidal variations. The authors found that river channel morphology can facilitate not only strong estuary flows, but also strong rip current flows when the river channel modifies wave breaking patterns, similar to what occurs in a surfzone bar-channel system. In addition, prior studies of surfzone hazards have typically found hazard to depend on the water level, which modifies wave breaking, with no dependence on the tidal phase (ebbing vs flooding); in contrast, this study found bathing hazard was different during rising versus falling tides when ebbing or flooding estuarine flows were interacting with the surf zone. I found these conclusions to be very interesting, novel, and supported by the analysis. I do have several concerns about (1) the framing of the paper, (2) the forecasting hazards analysis, and (3) the clarity of the text and figures.

(1) The paper emphasizes estuary mouth flows and bathymetric rip currents. Some discussion of headland/boundary rip currents is included but should be expanded given the clear importance of the boundary flows in this system. In addition, the paper lacks discussion of other rip-current types like flash rip currents, which I would expect to be present, as well as embayment rip currents, which would form in the center of an embayment rather than at the boundaries. Given the importance of the flows resulting from the embayment geometry, I wonder if the title and framing of the paper should be adjusted to “Combined surfzone, embayment, and estuarine bathing hazards.”

(2) The forecasting bathing hazard section isn’t clearly described and is lacking important details. In particular, the analysis relies on a look-up table of hazard statistics from a prior study, but the authors don’t provide a summary of this study or it’s applicability to their study. The authors emphasize how this study site and combination of processes is understudied, so it warrants some explanation why a hazard model developed for a different setting would be the right choice here. The relationship between risk, hazard, and exposure isn’t explicitly stated, and there is some redundancy in presenting all of these separately. Proportions of hazard scores are presented, but a forecast skill assessment should be performed.

(3) Prior to publication I think some improvements to the clarity of the text and figures are needed, particularly to emphasize limitations of the study and make sure that key results are discernable in the figures (see Specific comments).

Specific comments:

L38-43: This text suggests that rip-current patterns/behaviors, rather than speeds, classify the hazard level. This is then inconsistent with the next statement that places importance on speeds. I think this section would be clearer if the authors started with a statement that it is expected that a combination of factors, including pattern and speed, influence the hazard.

L46: Here it is concluded that flows in estuary channels may pose “an equal or potential even higher bathing hazard than rip currents” because the speeds are equal or greater than the speed of rip currents. I don’t think this is known.

L75: “embaymentisation ratio (length/depth)” Maybe “headland amplitude” and “embayment width” would be clearer terms than depth/length? Labeling these scales in Figure 2 could be helpful too.

L75-77: When do boundary rip currents occur, in which fast flows are along the headlands, versus a headland circulation, in which the fast flows are in the middle of the embayment? If a headland circulation is occurring some of the time, could this enhance the flows out of the estuary channel?

L82: It may be helpful to label features such as “ebb tide delta” on the figure.

L98: Did you introduce available data on water users?

L133: Give some information on how the echosounder and UAV datasets are merged. Is this the same as the process described later for the model bathymetry?

L136: Flow measurements were at 0.1 m above the seafloor. It would be worth describing why this is expected to be a good representation of swimmer hazard, and if there are times when it might not be.

L144: Are there concerns about when drifters may be poorly tracking the currents, e.g., if they are scraping the seafloor or surfing waves?

L175: can delete “respectively”

L200: “tidal variation was imposed uniformly on all four corners of the model domain” – what does this mean?

L206: Wave directions are mentioned here, but not wave directional spread, which seems to vary tidally in the observational record. How does this influence the results of this study? I don’t think spread was varied in the model runs? In addition, it would be helpful to know if the observed wave spectra are well described by a single peak period and direction, or if a wave systems approach would be more accurate. How does this affect the results? I suggest referencing the observations here to say how the model spans the observational conditions (shown for this year, is this similar for other years?).

L213-223: Does this method of forcing the estuary flows miss any river-estuary interactions, or does it include them because it’s based on a measurement near the estuary mouth? A small discharge is added to “conservatively account for fluvial flow” – can you elaborate and say if the results are sensitive to these choices?

L232: “Each virtual drifter was advected for 20 minutes, or until they had returned to a safe water depth (<0.7 m).” Could you elaborate on these choices and how they affect the results? Why 20 minutes? Why is a safe water depth 0.7 m? Why not keep running the drifters to see what happens next even if they enter safe water?

L246: should this be lowercase u_off?

L328: Mention what may be the cause of these fluctuations (e.g., related to infragravity pulsations, instabilities, or flash rip currents?).

L355: Given the spatial complexity of the flows, a point to point agreement may not be expected. You could plot the modeled maximum flows and flow range within a spatial region around the observational point for a more fair comparison?

L395-404: There is little or no discussion of rip currents here. How does the variation in rip current strength and characteristics with wave power or other factors influence the hazard metrics?

L427-429: The ebb shoal delta acting as a bar rip system is interesting. Is this mentioned elsewhere in the literature, e.g., papers on flows near a small river mouth encountering a surf zone (Kastner et al., Rodriguez et al., 2018).

L510: It seems the authors are following a prior method (Austin et al., 2013), but more explanation here of how these thresholds and scores were developed is needed to understand this section.

L512: Assuming I’m correct that Risk = Hazard x Exposure (which isn’t spelled out here), and Exposure is a measured quantity, it seems redundant to me to present accuracy results for both Risk and Hazard.

Figure comments:

Figure 1: x and y axes are not labeled in the top panel.

Figure 5: The panel b transect where the Gannel estuary enters the beach shows the topography over a long distance, which does not seem necessary, and makes it harder to see the channel. It may be more useful to see a transect across the shoals to show the scales of the estuary channel and other channels connected to the swimmer hazards.

Figure 6: Were other days similar? This could be interesting to look at as a composite, though maybe that would be difficult given the variation in the behavior with tidal range.

Figure 7: Do other time periods look similar? It could be interesting to show a scatter plot comparing measured and modeled flows.

Figure 8: This figure compares observed drift tracks with gridded results of model drift tracks. It may be helpful to also show example model drift tracks (not expected to reproduce the observed tracks, but presumably similar patterns), and gridded observational data, so that there’s a more one to one comparison. Does this figure only show two of the three regimes described in the text (L374-379)? I suggest using colormaps that are colorblind friendly and perceptually uniform (e.g., Thyng et al., 2016).

Figures 9&11&13: Vectors are small and very difficult to see. Colormap for the vectors includes dark blue, which is the same color as the background water color. This figure shows the variation in the wave breaking, water levels, and inundated bathymetry, but is not readable for information about flow patterns.

Figure 10: This is an interesting figure, though a bit difficult to interpret. It’s clear there is a difference between rising tide and falling tide. Showing line plots of Uoff vs the wave factor for three example tidal elevations for rising and falling tides may be helpful as a summary of this figure.

Figure 12: I’m not sure I find the change plots very helpful. It may be more useful to show how the flows differ for the different cases, to show different spatial patterns, but similar magnitudes?

Figure 14: Is proportion the most useful comparison? This doesn’t show anything about the timing of events. Why isn’t a forecast skill assessment shown? I’m not quite sure what to take from this figure.

References:

Kastner, S. E., A. R. Horner-Devine, and J. M. Thomson (2019), A Conceptual Model of a River Plume in the Surf Zone, J. Geophys. Res. Ocean., 124(11), 8060–8078, doi:10.1029/2019JC015510.

Rodriguez, A. R., S. N. Giddings, and N. Kumar (2018), Impacts of Nearshore Wave-Current Interaction on Transport and Mixing of Small-Scale Buoyant Plumes, Geophys. Res. Lett., 45(16), 8379–8389, doi:10.1029/2018GL078328.

Thyng, K. M., Greene, C. A., Hetland, R. D., Zimmerle, H. M., & DiMarco, S. F. (2016). True colors of oceanography. Oceanography, 29(3), 10. link: http://tos.org/oceanography/assets/docs/29-3_thyng.pdf, https://matplotlib.org/cmocean/