To understand possible adaptations to prolonged breath-hold diving capacities, the authors have measured parameters pertaining to body oxygen storage capacity in three tribes and 16 species of North American diving ducks, including blood haemoglobin concentration, volume percentage of red blood cells in blood, and myoglobin concentrations in breast and leg muscles.

These are valuable measurements that will be important to contribute to our understanding of the evolution of diving ability in birds, beyond and contrasting with the famous cases of penguins and alcids. However, I feel that the background, analyses and interpretations of the data could be improved. I list my major points below and include an annotated manuscript with several minor points for consideration, which will hopefully be useful for improving the manuscript.

1-Body mass, dive times and phylogeny
Body mass is generally one of the strongest single predictors of dive duration in mammals and birds. Its prior exclusion as a predictor just because of no correlation with Mb concentration seems missing an important determinant of dive duration. Its influence is both via its direct correlation with the size of the oxygen storage compartments of blood and respiratory gas space volume and muscle mass, and the allometric scaling of oxygen consumption rates. Did body mass correlate with dive time? Even if not, I feel that the average mass of species should be reported as a basic parameter in this study. Size differences within species could indicate differences in (seasonal) body condition or age/development, which both have been associated with differences in myoglobin levels in some species.

Furthermore, in the spirit of open and transparent data, all dive times used in this study, their references and justification for use in case of differing reports should be tabulated.

I feel also that for the same reasons, the underlying tree should be shown to good advantage in the main manuscript, indicating time calibration nodes and branch support values. The topology should be discussed in comparison with phylogenies from the literature and accession numbers or other relevant details of sequence data should be given.

At some point the authors also need to explain that the total amount of oxygen available from a given type of store also depends on the size of the storage compartment, e.g. for Mb the mass of gastrocnemius or pectoralis muscle, and for Hb the total blood volume and fraction and saturation level of arterial and venous blood. This may well all differ between diving and non-diving ducks.

2-Statistical analyses
It is not clear why PGLS wasn’t used in all analyses and why there was a switch PIC. Please explain this or analyse all data in a consistent statistical framework. You may well find that single predictors (such as Mb gastrocnemius, or indeed body mass) in PGLS analyses have AIC values not substantially worse (e.g. within 2 units) than more complex 2 or more predictor models. You can also compare nested models using likelihood ratio tests for significant differences.

There are no p-values and error ranges given for any predictors in the models listed in the supplementary table. The table would also benefit from including the fixed (or intercept) value in those models. Furthermore, the results of single predictor models (using PGLS, not PIC) could be given, including body mass in those and more complex models including body mass as well. Several models are nested and therefore likelihood ratio tests could be performed to check whether more complex models are significantly better than simpler models. There is also the problem of multi-colinearity and over fitting of models unless such checks are performed.

The rationale for grouping species into tribes for statistical analyses in Figs 1 and 3 but using species data in correlational analyses needs to be explained. I suspect there were no significant differences in a phylogenetic ANOVA when using species data and hence species were grouped into tribes. I may have missed this, but was there ever a significant difference between Mb in the 2 muscle types within species? Using a paired comparison may improve the power of such tests. One could expect leg propelled avian divers to have significantly higher Mb levels in leg muscles compared to wing-propelling breast muscles, the latter of which may exercise maximally only under aerial (aerobic) flight.

Regarding the outlier tests, were these done on the combined data for each tribe? This may have removed data in a non-random fashion, if some species had data concentrated at the upper or lower margins of data for their tribe. What did you do with outliers? Just indicate them in box plots? Or remove them from further analysis?

3-Ethics
The method of 'harvesting' needs much more detail to address any ethical concerns and to allow comparison with other studies. How were animals killed? Were relevant ethical guidelines followed?

This is a greatly improved manuscript that addresses all my previous points and is much clearer. There are just a few points that require attention:

1. Dive time of dabbling ducks
   It is great to see all sources for dive times being listed, but table 1 speaks of depths and doesn't give units. Furthermore, the values for dabbling ducks are all set to zero. I have seen at least some of these species submerging totally in the wild, so setting all these dive times to exactly zero needs an explanation or more detailed definition of what you mean by diving.

Dabbling ducks, especially in northern regions where shallow ponds may freeze over, are known to dive for food, see overview and literature in the following reference that includes some of the species that also appear in your table:
Furilla, R.A. and Jones, D.R., 1987. Cardiac responses to dabbling and diving in the mallard, Anas platyrhynchos. Physiological zoology, 60(4), pp.406-412.
In the absence of mean dive data for all species of these dabblers, I suggest to use the general average and maximal values indicated in the above reference of 5 (mean) and 10 s (max duration). This will also allow log transformation of dive duration, should that lead to a more normal distribution of data.

1. The negative intercepts in Table 3 and the figures (which I have seen only later)suggest that data were not log-transformed before analysis. Have you checked for normality with the response variable and each predictor? Often log transformation (base 10 or natural) of data is necessary to achieve this. This will change parameter estimates. The equations suggested by the single predictor models do not seem to match the scatter plots of the data in Figure 5, indicating something has gone wrong. Please check whether this is just a result of swapping x and y axis (see comment Nr 3 below) or is a wider issue than that.

2. Figure 5: It is customary to show the predicted (dependent) parameter on the y-axis and the predictor variables on the x-axis. This also seems to be the way the model test has been set up and how model parameters have been obtained. This would also allow direct comparison of the single predictor model parameters for [Mb] (g) from Table 2 with panel d of this figure. The straight lines need explaining. Are these the resulting linear curve fits? If so, their position seems odd and not matching the trend in the data. Please check.

Minor comments:
Line (L) 11: insert 'in blood' to make absolutely clear that this is not Hb trapped in the gastrocnemius

L 531: Please give the units of measurement and clarify whether this is dive depth as indicated in the 2nd and 3rd column heading, or dive duration, as in the table heading.

L553: It seems from the figure that these are posterior probability values (from 0 to 1), which is different from bootstrap support values. Please clarify.

L558: I didn't understand this immediately. It may be clearer to state 'the length of the scale represents 10 million years'. Or to state the units directly underneath the scale bar.

L567: This is an expected correlation but deserves a comment on why PGLS hasn't been used (I assume because of multiple individual data points per species). This may be a little distractive and you could shift this figure to the supplementary material.