Review of Measurements of frazil ice flocs in rivers by Pei et al

This article is a valuable contribution to the literature on frazil ice formation in rivers. It is a well-written description of field observations made under difficult conditions using novel instrumentation developed by the authors. The article could very nearly be published as is. However, I do have very few specific comments. I also have some suggestions for the authors. I believe these suggestions would improve the paper, but it is not required that the authors make any of the suggested changes.

Line 14. The term “relative depth” confusing. Is this the distance from the bottom of the channel or the distance from the water surface?

Line 22. It is suggested that the authors consider not using the term “sintering.” There is a long history of using the term sintering with regard to ice. The very first uses were applied to the adhesion of ice particles in air when they were held together with some pressure. The reduction in surface energy of the system provides the main driving force for sintering. (Blackford, J. R. J. Phys. D: Appl. Phys. 40 (2007) R355–R385) In the case of frazil ice flocs in supercooled water, however, the frazil discs can simply freeze together due to the heat transfer from the boundaries of the frazil disks to the supercooled water. There is no need to look for a reduction in surface energy of the system to cause the disks to stick together. Also, it is well known that flocs form only in supercooled water. Ice crystals in slush, a mixture of ice and water all at the ice/water equilibrium temperature, do not stick together. Perhaps you are using the word “sintering” in a very general sense to describe solid particles sticking together without regard to the mechanism causing them to stick. That use is imprecise and confusing. The exact mechanism causing the frazil disks to fuse together should be described.

Line 22. It is suggested that the authors consider providing more background on the process of floc formation. The frazil disks are transported by the flow. If the frazil disks are all moving at identical velocities, they cannot collide. Disk collisions require spatially varying disk velocities. Spatially varying disk velocities can result from spatially varying flow velocities and disk varying buoyant rise velocities. There are several mechanisms providing spatially varying flow including turbulent eddies of appropriate size and the influence of the stationary boundary at the channel bottom.

Line 25. It is suggested that the authors consider the vagueness of the term “grow.” In the previous sentence you write: “Frazil flocs grow in size either by the thermal growth of the crystals and/or by further aggregation of individual frazil ice particles or flocs.” Then you state “Once frazil flocs grow…” It seems to be that the word “grow” should be applied only to thermal growth of the crystals. Increase in size through aggregation is something different. Perhaps there can be two distinct types of growth, but you should make this clear.

Line 41 (and other locations). It is suggested that the authors consider not using the terms “residual supercooling” and “principal supercooling” and replacing them with more accurate terms. According to the authors, frazil ice formation has two periods. The first is the “principal supercooling” period and the second, which follows the first, is the “residual supercooling” period. There is a long history of using the term “residual supercooling” going back to the very first experiments of Michel (Michel, Bernard. Properties and processes of river and lake ice. Université Laval, Laboratoire de mécanique des glaces, 1972.). However, the use of the term “residual” is very unsatisfactory. Residual describes what remains after most of something is gone. However, the supercooled temperature of the water is not a residual of the higher levels of supercooled water temperatures that were temporarily present during the earlier principal period of supercooling. The water temperature at all times represents a dynamic balance between the heat loss at the water surface and the latent heat released by the growing frazil ice is suspension and the anchor ice on the channel bed. The water temperature is more-or-less constant during the residual period because the heat loss at the water surface and the latent heat released by the growing ice are equal,  . In summary, residual supercooling is not left over, it represents a dynamic heat balance exactly as in the principal period. The authors should consider replacing “principal supercooling” with “transient supercooling period” and “residual supercooling” with “steady-state supercooling period.”

Line 106. Table 1. It is suggested that the authors consider adding an additional term to their “Summary of the study reach characteristics” table. It is suggested that they add the term e, the turbulent energy dissipation rate per kilogram of fluid. This term strongly influences the heat transfer from suspended particles and the secondary nucleation rate. This can be estimated for both channel flow and laboratory tests. This parameter would allow the reader to compare field sites with previous laboratory tests. The units are generally in Wkg^-1 with dimensions of m²s^-3.

Line 117. Change “capture” to “image.”

Line 265. Change "4.2 Heat flux analysis” to “4.2 Heat flux analysis at the water surface”

Line 265. Heat flux analysis. It is suggested that the authors verify the accuracy of their heat flux analysis at the water surface by modeling the water temperature decline early in the transient period prior to the formation of ice. This could be done for deployments NSR-L.1, NSR-L.3, and NSR-L.4. Two basic and reasonable assumptions would make the model simple and straightforward: that there are no significant gradients of temperature in the longitudinal direction (parallel to the flow velocity) and that the water temperature was well mixed in the vertical direction.

Line 394. Revise section starting with “Arakawa (1954) discovered …” and ending with "time to grow irregularly.” (Line 398) It has long been realized that the stability of the edge of the ice crystals is controlled by the formation of temperature gradients in the water at the ice/water interface when the surrounding water is supercooled (Mullins, W. W. and R.F. Serkerka (1964) Stability of a planar interface during solidification of a dilute binary alloy. Journal of Applied Physics, 35, No. 7, 444-451). The perfect disk shape of frazil ice results from the anisotropic crystalline kinetics combined with the turbulent suppression of temperature gradients surrounding the crystals. Given the ability of turbulence to suppress gradients through mixing, unstable disk growth is typically a special case. Irregular particles generally indicate that the frazil ice particle has been in quiescent regions with exceptionally low turbulence levels. In these regions temperature gradients can form in the water surrounding the ice particle. Small perturbations of the ice crystal boundary encounter colder water because of the temperature gradients and grow more rapidly.

Line 385. 6. Discussion. It is suggested that the authors address these two related questions in this section. 1. How do you explain the near constant supercool water temperatures during the steady-state supercooling period based on your observations of suspended frazil disks and flocs? 2. What fraction of the total ice created in the water column is being sampled by the apparatus? The total ice created can be estimated based on the surface heat flux and the water temperature.

General comments and suggestions:

The authors present a well written study which investigates the properties and concentrations of frazil flocs in reaches of the North Saskatchewan River (NSR), Peace River (PR), and Kananaskis Rivers (KR) using a submersible camera deployment called FrazilCam. The manuscript presents in-situ observations of floc shape, size, and concentration which is of use to the frazil and river ice communities for improved modelling efforts. The following general suggestions can be considered by the authors for potentially improving the quality of the manuscript. Specific comments follow these general suggestions.

1. The methodology used for classification of flocs was challenging to follow at times. Perhaps adding in a flowchart or process-flow diagram would aid in reader comprehension to better follow all the processing steps.
2. The heat flux analysis requires additional clarification and minor reanalysis. Namely, description of equations used and some minor addendums to methodology. Please see specific comments.

Detailed  comments and suggestions:

Study Reaches:

L108-111: Is the data obtained from Kellerhals et al. (1972) the most up to date site data available?

L114: The site map is well made and clear.

Instrumentation, Methodology, and Deployments:

L115 and L126: The study modifies the FrazilCam system developed by MacFarlane et al. (2017). Can the specific modifications from the original system be made more clear in this section? Was the FOV the only aspect modified (L126-127).

L147: Which frequency Aquadropp ADCP was used? My concern is for blanking distances, as the river sites have quite shallow depths (Table 3).

Image Processing:

L221-223:  Were the preliminary experiments conducted by yourself? If not, please provide some more context.

L234: ‘S’ is not defined prior to its use here.

L237: Perhaps it should be made clearer earlier in the study that only 1 deployment coincided with an entire ‘principal’ supercooling event.

L231-264: The addition of a process-flow diagram or flowchart would largely help the reader understand the methodology used. Additionally, it would be useful to consider provided quantitative measurements of how many images were taken (in total), followed by how many were removed at each processing step.

L255: One key missing piece of information was the specific sampling time used for each deployment. On L131, it is noted that 5 images at 1 Hz every 9,15, or 19s were acquired depending on field conditions. It would be more transparent to describe under what case/deployment each sampling time was used. If sampling times were mixed for a given deployment, this should be also stated and justified.

Heat Flux Analysis:

L273: It is uncommon in heat flux analysis to explicitly consider an albedo in the longwave spectrum as nearly all radiation in this spectrum is thought to be absorbed at/near the surface (shown by your use of a very small albedo of 0.03). It is recommended to remove this.

L277: Mean water temperatures are used here for conducting a surface energy balance. While vertical turbulence may be well developed and river depths relatively shallow, some degree of caution should be presented on this matter. Surface temperatures may deviate significantly depending on flow and meteorological conditions. The assumption that the river reach is fully vertically mixed should be stated explicitly in this case.

L278-280: Satterlund (1979)’s parameterization relies on data from Aise and Idso (1978) from continental Montanna and is extended with Stoll and Hardy (1955) for measurements in Alaska. Perhaps it may be more prudent to select a more well-used scheme for clear sky conditions shown effective in higher latitude regions of North America (e.g. Efimova, 1961). Key et al. 1996 provide a review on the matter using data from Alaska and the Northwest Territories on their review of parameterization schemes.   It is left to the authors’ discretions to keep the current scheme or adopt one of the above-mentioned after reviewing the noted references.

Efimova, N. A. (1961). On methods of calculating monthly values of net longwave radiation. Meteorol. Gidrol., 10, 28-33

Key, J.R., Silcox, R.A., Stone, R.S., 1996. Evaluation of surface radiative flux parameterizations for use in sea-ice model. J.Geophys. Res. 101 C2, 3839–3849

L281: Where was Bowen’s ratio obtained from? In addition to the above-mentioned, please describe equations used for the flux analysis within this section. It provides the reader with the information readily, rather than having to access several other sources to understand the approach taken.

Results:

Floc Shape, Size and Concentration:

L303: It was quite interesting to record such a large floc size (99.69mm). As I understood, the FrazilCam in this study had an increased FOV relative to its predecessor. Would recording a max floc size such as this in KR-E1 suggest perhaps the FOV may need to be further increased? Perhaps there may be potential for biasing floc sizes too low, as larger flocs that are unfavourably oriented interact with the polarizers and break.

Floc Size Distribution:

L319: You note that lognormal distributions are reasonable fits for the distributions. Would you be able to provide a quantitative measure of the fit for each of the histograms?

Time Series:

L332: It is understandable that not all sites are presented within the contents of the manuscript. I do however believe that the reader is left curious as what the other sites might have looked like. Perhaps it can be considered to add in the other deployments data (similar or simpler versions to Figure 8,9,10) as supplementary data for further transparency.

L357-358: The usage of hourly component flux data for correlating floc properties and concentrations is a bit questionable given the timescales of the deployments (~1-3hr). It is described between L155-167 where this data is obtained from. If these sites (minus the NSR reach) have sub-hourly data that can be used for the heat flux analysis, please consider updating.

L375, 380, 383: Figures 8,9, and 10 would benefit from the addition of air temperature data.

Discussion:

L483-484: A larger limitation would be the use of hourly data rather than neglecting surface conditions and sediment heat fluxes. This is likely too coarse for the intended goal with correlating net surface heat flux with floc properties. This can be considered going forward for future studies on the matter.