This study attempts to understand the causes of decadal variability of dissolved oxygen in the Gulf of Alaska, observed at the Ocean Station P (OSP) at 145°W, 50°N. The oxygen timeseries at the OSP is the longest record of dissolved oxygen. The decadal variability of this data has been studied intensely over the last 20+ years and there are many hypotheses proposed to explain this variability. This study put forward the idea that the subduction of thermocline waters generates oxygen variability in the western Pacific, which would then propagate eastward following the circulation pathway of the North Pacific Current. Approximately 10 years later, the signal reaches the OSP and is observed there.

This hypothesis itself is not new, but what is new in this study is that a significant correlation was found between dissolved oxygen at OSP and the isopycnal outcrop area in the western Pacific which was reconstructed from the historical observations (EN4). Furthermore, the outcrop area is controlled by the density of the sea water at the surface, which further depends on the variability of sea surface salinity, rather than temperature. However, this variability does not exhibit significant correlation with the dominant mode of climate variability in this region. The maximum lag-correlation values are on the order of 0.5 to 0.6, which would explain about 25-36% of the total variance. This seems a significant contribution to the total oxygen variability, while it also leaves significant room for other mechanisms as well. My overall impression is that the manuscript contains significant progress on this problem and merits publication. Having said this, I would like to raise two points that the authors can consider for a revision. 

At L75, the authors choose to analyze data after 1982 only. I ask the authors to reconsider this choice because dissolved oxygen data from OSP exists from 1956. Since the focus of this study is the decadal variability with approximately 20-year timescale, the statistical significance of this analysis is critically limited by the effective sample size. The additional 26 years of data can capture additional full cycle of the signal potentially. Figure 2 indeed supports that the EN4-derived outcrop areas show the same pattern of maxima and minima as EN4-OISST. Then, it is possible to include the additional, extended timeseries before 1982.

At L135, it does make qualitative sense that a larger outcrop area indicates more ventilation, and a smaller outcrop area indicates less ventilation. However, not enough reasons were provided to justify why the spatial extent of the surface outcrop area is used as the only proxy for the amount of ventilation. There are atmospheric reanalysis data products available for estimating buoyancy and wind stress forcing. The latter can provide an estimate of Ekman flow. Mixed layer depths and geostrophic circulation can also be estimated from EN4 products. It makes me wonder if there were any reason that direct subduction estimates weren’t used in this study. It is possible that these calculations were already performed by the authors or prior studies by others. If this is the case, it would be important to include this information.

I really enjoyed reading this manuscript. It is well-written, explores a simple idea very effectively and puts the results in context in a way that makes them potentially very useful.

I think this manuscript could be published as is (aside from a single typo) -- the figures are very readable and the text succinct and clear. However, I nevertheless will provide a few suggestions that I think would improve the impact of the manuscript.

Line 39: Could also mention that similar cycles have been seen in DIC (e.g. Figure 7 in Franco et al 2021) -- this would allow you to pick up this point again at the end and suggest what you saw might also apply to CO2/DIC.

Line 235: I found the lagged correlations (Figure 5) really interesting as a way to assess how long it takes water to get from the NW Pacific  to OSP (and the NE Pacific). My immediate question was: what does it look like for other isopycnals? It’d be powerful to see an increase with density and also really useful to have an idea of how fast signals are transmitted at different depth/isopycnals. I think you could make a nice comparison with Ueno & Yasuda 2003’s Figure 7 (based on simple model).

Line 342: Excellent point! It'd be really useful if you were able to suggest some ideas for what they might look like.

Line 347: Here's where I think the typo is. Is/are word/s missing? I find this phrasing confusing.