Dear authors,

This was an interesting paper that I enjoyed reviewing – it was a great opportunity to learn about belugas and multievent mark-recapture! Overall, the manuscript was well-written, and the code and supplements helped with understanding the model. I am not a mark-recapture expert, but it seems like an interesting development in that arena that might be potentially valuable to researchers working in other systems. Along those lines, my biggest concern is that the paper seems to occupy the “no man’s land” between a description of a novel model and an empirical application. To put it another way, if I were a graduate student interested in applying this model to a different system, I think I would have trouble figuring out how to modify the model and code to suit my needs. I think it would be very helpful to center the paper around a simpler synthetic example. Such an example might have a smaller number of true states (say, 4 or 5, rather than 14) and observed states (say, 10, rather than 72 in the age assignment matrix). Such a change would require some refocus and restructuring, but I think a lot of the information is already present. Below, I list some line-by-line comments and reactions. I hope this helps!

-L21-23: This sentence feels redundant with “hard-to-study” and “obtaining…data…can be difficult”. Suggest simplifying.
-L23: Suggest specifying mark-recapture models.
-L46: Apologies for an annoying grammatical comment, but I it should be adult–offspring (with an en dash, not a hyphen), since two nouns are being connected.
-L59 and throughout: Suggest omitting acronyms if at all possible – there were quite a few (YOY, MR, CIBW, etc.), which made the paper a bit more difficult to interpret.
-L56-109: Suggest including a bit more of a backdrop on mark-recapture and multievent mark-recapture techniques for readers who may not have extensive experience with these methods. For example, it is not clear from the introduction if multievent mark-recapture models have been used exclusively for offspring states, or whether they are more broadly applicable to cases in which ecological states (e.g., disease status, whether a site is occupied by a breeding pair or just an individual, etc.) are imperfectly observed.
-L68-69: Suggest updating this topic sentence to include the idea that uncertainty also arises when young are observed but are of unknown age (and can possibly be from previous years/reproductive events).
-L93: Suggest omitting “typically”.
-L87-109: This section is rather long (e.g., lines 87-90 and 98-102 felt redundant). Also, if the objective of the paper is to present a novel model, I wonder if this level of detail about the focal organism should appear here.
-L150-151: Looking at the data, I only see 23 observation states (out of the 72 listed). I’m assuming this means that the other 49 states were never appeared in the data? (E.g., state 12, seen with 4-yo calf).
-L153-154: It would be great to report the agreement rate between classifiers if that information is available.
-L158-165: While reading this, I wondered about the influence of false positives (i.e., calves being assigned to the “wrong” adult or age class) – is this something that you considered, e.g., in the simulation study?
-L166-207: This section is well-written and generally makes sense, but I think the model would be more accessible to readers if a simpler, generalized example with fewer states were presented such that the matrices could be presented in the main text.
-L225-288: I know very little about belugas – does this imply that, if a mother dies, the calf also dies? Is this in line with your understanding of the species (e.g., are “orphans” ever “adopted” by other adults)?
-L243-254: I tried out the code, and everything ran smoothly, so kudos for making that available. One thing I was curious about: did you explore including temporal variation in or effect of covariates on the variables of interest (survival, reproduction)? In lines 418-420, you note that the model can be extended in such a fashion. It would be interesting, for example, to allow survival to vary by year (as a random effect) so that average survival across years, as well as each year’s survival, could be estimated. Assuming such an extension doesn’t prevent model convergence, this would provide arguably richer information for stakeholders.
-L260-261: Minor, but when I read these equations, I originally interpreted the e as the mathematical constant (~2.71); might be more clear just to write estimate0.5,i – truth or something like that?
-L269-270: This result surprised me – can you think of a mechanism driving lower survival of non-breeders? I would assume that breeding / having a calf might incur significant energetic costs or render breeding adults more vulnerable to predation.
-L282: Believe there is a typo here – should be “sample”?
-L305-307: Suggest softening this statement a little. I am also a little skeptical that static demographic estimates from 12 years of data can reveal the influence of exogenous stressors (I’m assuming these include things like climate fluctuations, etc.), which presumably vary through time.
-L320-327: This is interesting – any ideas regarding the mechanisms behind this? Abiotic conditions? Predation risk? Low resource density?
-L342ff: I like this section, as I think it is good to explicitly discuss assumptions of the model, but I wonder if it would be better-suited for the methods?
-L374: Suggest adding a component to the simulation to address the issue of offspring being associated with adults other than the mother. What you’ve written here largely makes sense, but ultimately it is impossible to evaluate the consequences of violating this assumption with empirical data when truth is not known.
-L377ff: This section felt like a bit of an afterthought. For the current venue of MEE, I think the paper would be stronger if a more generalized case of the model were presented, with the beluga analysis included as example, rather than the focus.
-L595: I found these codes difficult to interpret, even after looking up the definitions multiple times. Suggest spelling out the codes in full so they can be more quickly interpreted.
-L610: Great figure. Very helpful!

Dear authors,

The revised manuscript feels like a stronger and more broadly applicable piece than the original submission that I reviewed. I particularly appreciate the addition of a simpler example. Overall, the revision is well-written and seems like a valuable addition to the mark-resight literature. I have only minor comments on this version.

L40, and throughout: Verbiage like “reasonably precise” and “relatively precise” strikes me as rather vague; is there any way the precision of these estimates could be described quantitatively, or compared to less-precise estimates from simpler models?

L47: Suggest omitting “for the first time”

L110-112: I wonder about the similarity / parallels between the model here and “categorically marked” methods in which individual identity is partially known (e.g., individuals can be classified as male/female, juvenile/adult, but cannot be uniquely distinguished; Augustine et al. 2019). Might be worth mentioning if you think there are parallels.

L137: Suggest omitting “wholly”

L158: Suggest rephrasing as, “In this example, …”

L219-223: I like this summary – succinct and clear.

L347-348: Here, “sample size” refers to the number of individuals, rather than the number of resightings, right? If so, the description here is fine.

L361-362: Suggest briefly mentioning what these are – I imagine parameters like survival and breeding probability?

Literature cited
Augustine et al. 2019. Spatial capture–recapture for categorically marked populations with an application to genetic capture–recapture. Ecosphere 10(4):e02627.