The authors present a new dataset of Moss Kill dates on Baffin Island, and compare them with an earlier dataset (from 2005), as well as nearby paleoclimate datasets and model simulations. The differences in the two datasets allow the authors to explore how the recent retreat has exposed older mosses, and the implications for the Common Era glaciation history of the region. The authors argue that the observed pattern of glacial advances is representative of the North Atlantic Arctic due to the similarities with other data throughout the region. Finally, the authors compare their results to model simulations, allowing them to explore the whether or not the peak of Common Era glaciation in the region, 1780-1880 would have led to glacial inception in the absence of anthropogenic addition of greenhouse gases. 

The manuscript presents important new data that provide key insights in to the recent glaciation history of the region. It is also well written and appropriate for publication in Climate of the Past. I recommend publication following a handful of mostly minor corrections. 

Calibration bias testing approach

Investigating how the calibration curve may impose biases on the results is a good idea, however I'm somewhat concerned about the methodology. I worry that a single realization of a random 160 samples could get you spurious results and peaks that are either artificial, or artificially high or low.

I can imagine two alternate paths to improve this. The first would be to expand on what you did, but rather than randomly sampling ages once, repeat that sampling 1000 or so times. Looking across this ensemble of random results, and specifically extracting so high end quantiles (90, 95, 99) would robustly show how the calibration curve impacts the likelihood of finding calibrated ages in certain time periods, and also tell you how likely it is for a random collection of dates to give you the observed results. One note on this - I wasn't certain from the description, but rather than randomly sampling radiocarbon ages, it might make more sense to randomly, with a uniform distribution, sample calendar, and then use a reverse calibration to create radiocarbon ages. Then add realistic analytical uncertainty and calibrate again.

The alternative strategy would be similar, but where you'd be looking to quantify the cumulative probability bias of each decade in due to the calibration curve, and then adjust, or bias-correct your results, effectively lowering peaks around 1280 and 1450 CE, while raising peaks in the 1300s. I think the first strategy is a more direct test. 

Minor points

I'm not sure I'm convinced by the argument on lines 347-349. Earlier you argue that the dead moss are removed within a few years except in rare circumstances. Having a multidecadal retreat where the moss is preserved, and then a new advance that entombs both old and new moss seems possible, but I would like to see other potential explanations for this discrepancy explored (and argued against, if appropriate). 

446: typo - I think "in" should be "is" here

457: Be explicit about what the data are saying that the model is getting wrong here. 

Figures 5, 6 and 8. I think it'd be very instructive to add a panel showing the difference between 2018 and 2005 data, highlighting what moss kill ages were exposed during recent retreat

The caption for figure 7 looks like it might be swapped with that of figure 8. 

844: I couldn't readily find the Dewar lake data at that website. Could you provide a more specific link?

Please add where the Hvitavatn dataset can be found. 

The article is generally well cited, and includes a good review of early work with the method and in the region, however I found several references cited in the text that weren't included in the references section, including Falconer 1966.

This is an excellent study and a well written and well illustrated manuscript. In fact one of the best I have seen in years and certainly suitable for Climate of the Past.The authors have dated a large material of Polytrichum mosses from ice caps on Baffin Island, and use the ages as a proxy for summer temperatures. Based on a comparison with data from Iceland, they develop a palaeoclimate record for the Late Holocene for the North Atlantic. The authors conclude that a new ice age may have been on its way during the cold period 1780 to 1880, had it not been for global warming caused by human-induced emission of greenhouse gasses.

I did not see any reference to figure 13 in the text

Also I could not find figure 16