I thoroughly enjoyed reviewing this paper and commend the authors on the substantial amount of effect that has gone into this work. I think this will be an important addition to the literature on insect migration. Further, the comparisons to historical observations, although they cannot be done direct, provide further support for declines in insect populations worldwide.
I do, however, have several concerns about the methods and statistical analysis.
Throughout the paper, the authors rely on the distance across the pass (30 m) or the width of the video recordings as their measure of ‘effort’. But, I suspect insects do not migrate at a consistent elevation, and I think additional information on the area (distance x height) being monitored should be reported. Further this area is not considered, I’m concerned that this could affect the calculations of total insects traversing the Pyrenees.
In the analysis on migration timing, it’s not entirely clear how the ‘mass migration days’ were selected. It’s possible that a figure would help to clarify this approach or you could identify these days on Figure 3A. This analysis mentions 22 days of mass migration, but elsewhere the text says 38. More information is needed if only a subset of the 38 days were used. Second, because it’s not clear how different these days are from year to year, the timing of mass migration should be considered relative to day length and variable like sun elevation rather than hours and minutes through the day. The Results mentions ~1 hour before solar noon, but this should be reflected in the full analysis and Figure S1.
In the analysis of meteorological factors, the use of stepwise regression is risky as it many not necessarily identify the ‘best model’ (lowest AIC). I would encourage the authors to either develop a few model subsets based on hypothesis and predictions to compare or use an all subsets approach which can be done easily with the R package MuMIn. These results much be presented with delta AIC and AIC weights at minimum to support your ‘best model’. Please review Arnold 2010 (https://doi.org/10.1111/j.1937-2817.2010.tb01236.x) for informative model selection. Also, it is not clear from the text if there were multiple observers involved in counting insects in the field or on videos. If there were multiple observers or Malaise traps, then I would encourage the authors to add a random effect for observer and/or trap to their models. I would also encourage the authors to account for year, day and solar elevation or time of day relative to day length; year and day may be best incorporated also as random effects.
In the, the authors assume that, but it would be helpful to provide some metrics of elevation throughout this region to back up that this is the only viable migration corridor.
I am concerned about many of the assumptions made during the estimates of numbers of insects annually traversing the Pyrenees, including 1) insects solely travel through the Pass of Bujaruelo, 2) a 1.2 km section of this range is similar to the other 430 km, and 3) the area covered during the surveys adequately captures insect abundance. It might be helpful if the authors could provide a landscape model of elevations that might be suitable for insect migration to support these assumptions, but without further evidence I would strong recommend removing this analysis from the paper entirely. The other work is exceptional enough on its own to not include this analysis with so many unknown variables.
Lastly, throughout the entire methods section it was often challenging to follow the approach and how different sources of data were used. I recommend the authors thoroughly review this section to clarify for readers with no background information. Everything must be clear enough to be reproducible by a naive reader.
Minor comments (line numbers are based on those when the “All Markup” option is used):
L26: document or estimate?
L27: write out numbers < 10; 5 should be five. Check throughout manuscript.
L27: which 5 orders?
L50: use ‘common name’ for Lepidoptera or use scientific names for all three groups
L122: just a single Malaise trap?
L155-156: was it a the same 15 minute period every two hours each day? If not, how was the 15 minute period selected?
L155: what was the height of the area that was monitored?
L156: how does this time period line up with day length and the timing of sunrise or sunset?
L195: this time period contradicts that described above for the counts
L202: why average daily rather than a minimum daily?
L256: please add more details about this model, e.g., response, explanatory covariates, distribution, etc. See above comments on observer effects. This is a great example of many of the statements made throughout the Methods that provide insufficient information for a reader to understand and reproduce the work.
L403: 22 days reported in the Methods
L406: use
L570-572: are the any other studies documenting declines in insect migration that you could mention even if they’re not specific to this region?
L679-690: what about minimum temperature? Your analysis included the average, but minimum temperatures are often a better predictor of insect activity.
Fig 3A: please add tick marks on all axes.
Fig. 4: consider changing to 95% confidence intervals instead of SE. Also, in 4B, the SE has a very unusual shape; I would check that it has been correctly calculated.