This integration of two already integrated models is impressive, clever, and an important step forward for migratory connectivity and population demography research. I have several comments, mostly on wording or presentation of the results.

Major Concerns:
None, although the comments for the breeding population data below almost rise to this level.

Minor Comments:
Note: My line numbers are for the version with tracked changes (which is the one I saw first).
Lines 64-66: The first part of this sentence reads like the questions of migratory connectivity are largely already answered for birds, which is clearly not the intent. I would switch “mainly” and “been.”
Line 117: Gain in bias is a bad thing, unlike gain in precision. Maybe change to “change in precision and bias”?
Line 174: Should “recovery model” here be something like “tracking model”?
Lines 242-243: I understand what the authors mean, but “informs on” is usually what one criminal does to another. I’d switch to just “informs.”
Lines 252-305: I’m a bit confused about this section – random values were chosen for each demographic and migratory connectivity parameter. Was this done once, once for each simulated dataset, or what?
Lines 292-294: This one-sentence paragraph probably isn’t needed in this section. Then the subheading could become just “Model assessment.”
Lines 312-321: Should winter and wintering be switched to nonbreeding? They migrate to some places south of the equator (where it’s not winter).
Lines 362-364: What do the minimums and maximums represent? Are those years?
Lines 362-371: I’m dubious about turning what appears to be two population estimates per country into population numbers for each year through the assumption that population growth is the same every year. In addition to the likely violation of that assumption at the country level (or any level), this probably provides false precision to the ICPM/IPM by acting as if there were more count data than is actually available. It looks like the issue being addressed is different countries within a region having different count years; is that right? If so, I’m not sure what approach is available to handle these assumption violations, but an approximation could be made by using only two of the regional level “count” distributions that the current approach generates (say 2000 and 2012) in the ICPM or IPM.
A response to this comment could be that this is only an example species to demonstrate the new model, and therefore it doesn’t matter if the results have false precision. This is reasonable, but I would counter that because this is a new model, other scientists will be looking here for guidance on how to implement the model for their species.
I would like the authors to try the approximation I suggested. If it doesn’t work, at a minimum this issue should be mentioned in the Discussion.
Lines 371-378: The apparently required(?) assumption for Curlews that there is no observation errors in counts should also be addressed in the Discussion.
Figure 3: What do the 95% CI on bias and CV between simulations represent? It seems that this is similar to showing a box plot (but a bit less informative). But I don’t think “does the CI overlap 0?” informs the questions of whether the models are biased and precise. Since that’s what we’re trained to do with CI ranges, maybe box plots or violin plots would be better.
Lines 416-420: Is something missing here? It states the bias for the ICM & ICPM are less than the IPM, but that has a mean relative bias of 0%. Oh, I think I see, since the range of the CI for the IPM is larger, suggests that biases are often greater in magnitude for it. Absolute value of the relative bias may be a useful statistic here. Perhaps the difference in absolute value of relative bias between ICPM and IPM for each simulated dataset, summarized, would be even more useful, as it would indicate how often one model provided more biased results than the other.
It might be helpful to put RMSE back in, at least in a supplement. A measure of accuracy such as RMSE can help one make bias/precision trade-offs.
Line 487: See comment on lines 242-243.
Lines 544-549: This section of inserted text is well stated.
Line 258 says three nonbreeding regions were used in simulations, but Supplement 1 (Table S1.1) says two.