I read with interest the manuscript "Chaotic oceanic excitation of low-frequency polar motion variability" submitted by Börger et al. for possible publication in Earth System Dynamics. The paper utilizes the OCCIPUT large ensemble with 50 realizations of time-variable eddy-permitting ocean mass and flow fields to calculate effective ocean angular momentum functions characterizing the excitation of changes in the solid Earth's orientation with respect to inertial space. The paper is very well written und certainly fits into the scope of the journal. I recommend this work for publication as soon as a number of comments have been reasonably well addressed.

(1) The analysis presented in this paper is based on the SPACE2018 series of Earth Orientation Parameters as processed at the JPL. Authors should explain in more detail why SPACE2018 is used here instead of the associated COMB2018 series, or a more recent reprocessing of the same data (i.e., COMB2019). Authors should also consider to use the newly published EOP series from the ITRF2020 computation that are operationally updated as EOP 20 C04 by scientists from the Paris Observatory. In any case, it needs to be discussed in the article how a particular choice of the EOP series might affect the interpretations of the results presented here.

(2) Time-variable gravity field representations from GRACE that are additionally augmentend by SLR and DORIS observations to extent the time-series have been used only rarely in Earth Orientation Parameter research. In view of the cautious comments provided by the authors in line 240, I propopse to explicitly show the hydrological angular momentum functions derived from GRACE+SLR+DORIS for the whole time-period 1995-2015, and compare it with GRACE-based excitation functions -- ideally derived from publicly available Level-3 products, like the Cost-G combination solution available via gravis.gfz-potsdam.de to make results traceable -- and an independent model-based hydrologic excitation function published elsewhere. Please note that a detailed discussion of the contributions from Greenland and Antarctica is not necessary at this point.

(3) It is quite suprising to see that the largest interannual surface mass variations outside Greenland and Antarctica are found on the Malakka peninsula in South-East Asia. This is not really intuitive from a hydrometeorological perspective and calls for further investigations. In particular, it should be thoroughly checked if poorly treated tectonic signals associated with the 2004 Sumatra-Andaman earthquake (and later events in neigboring areas) are responsible for this feature. Please report in detail about any modifications made to the GRACE+SLR+DORIS processing, which is not yet really well covered in the scientific literature.

(4) Authors speculate in both abstract and conclusions about possible implications of this work for EOP prediction, but fail to elaborate it further in the article. I suggest to remove this comment from the abstract in order to avoid raising unrealistic expectations with the reader. In any case, rigorously assessing the potential consequences for EOP predictions should be left for future study.

Börger et al. used the outputs of 50 ensemble OGCM simulations, driven by the same atmospheric data from the DRAKKAR forcing set (DFS 5.2) with slightly perturbed initial conditions. Then, the authors computed monthly OAM, separating the three output variables, p_b,  and into the common signals to all 50 members and those uncommon signals (eq. 3). Based on those common and uncommon signals, the authors computed a series of OAM data and showed some interesting results. In particular, Figure 2c shows an intriguing peak in the “forced” signal (L169-170), which is clearly not annual but rather broad and significant; the same peak can also be found in Figure 3c; I have never seen such a peak that is unexpected and deserves to be analyzed in more detail.

The goal of this paper is to “separate the ensemble OAM estimates into forced and intrinsic components and assess their contribution to the observed wobble excitation” (L44-45). However, I wonder if the goal can be accomplished from the presented approach. While the authors consider the common signals as “forced” and the uncommon signals as “intrinsic” (or “chaotic” in places), I do not agree with the authors’ understanding (or their terminology). I understand that the truly “intrinsic” signals should also be included in the “forced” ones and that the uncommon signals after the ensemble simulations are simply due to different initial and boundary conditions, indicating simply the uncertainties of the OGCM simulations, whereas it is important to quantify the uncertainties. Thus, instead of agreeing with a statement in Line 279 “variability in χ^o,i is a considerable fraction (43–50%, see Sect. 3.1) of the total oceanic excitation of polar motion even on interannual time scales”, I have rather understood that the simulation outputs have considerable uncertainties. I would suggest changing the scope of this work and focusing more on the analysis of the derived interesting “forced” OAM signals.

The problem might be because the authors used “eddy-permitting” model instead of “eddy-resolving” model. If the latter “eddy-resolving” model was used, it would have much finer spatial resolution and allows to more accurately compute the fine-scale ocean dynamics; the “atmospheric-driven” component should also more accurately include the intrinsic chaotic ocean variability. The authors might be recognizing this point in view of the sentence in Lines 279-280, “A caveat to be acknowledged…”.

Minor comments:

Line 56: There are still large uncertainties in Qc, and the value 179 is rather high.

Line 248: Is there any evidence for the effect of the core on interannual wobble excitation? While the present work assumes pure elastic deformation, 1.10 and 1.608 in A1, anelastic deformation will rather need to be considered in longer timescales. 

The main interest of that paper is to show that the "intrinsic" circulation in the ocean excites significantly the polar motion at both inter-seasonal and interannual time scales. Stemming from meso-scales eddies, this intrinsic circulation is chaotic, and till now is not fully captured by oceanographic observations. However, it can be simulated, as done by the authors.

In many respects, the authors are reshuffling the deck when it comes to modelling polar motion.

They show that the inter-annual polar motion could result from the oceanic circulation, without the need of the core-mantle interaction as advocated by many recent papers.

I am not sure sure whether the authors realized that their study could also modify the current understanding of the Chandler wobble. Indeed, according to Fig. 2b, the mass term of the intrinsic excitation at the level of 1 mas could contribute to Chandler wobble in a very significant way. In the forced part of the ocean angular momentum, I wonder whether the authors considered the pole tide as a source of forcing. Could the authors address that important question in the revised version?

The paper is well written, the approach is well presented. Only legends in Fig. 5 and 6 and captions of tables 1/2 deserve some light improvements: the legends for excitation functions have to be well split, in tables write "Percentage of Explained Variation (PVE)".