We thank the reviewer for the positive assessment and helpful comments. We have reviewed the manuscript and revised the manuscript accordingly. 

Comment 1. l.102 depth / Fig. 1a: 

Response: Figure 1a shows the water depth of the Brazil Basin study region. The seafloor depth over the ridge region is approximately 3000 - 4500 m, while the adjacent off-ridge abyssal plain has a mean depth of approximately 5500 m. We will revise Fig 1 and show model bathymetry more explicitly.  

Comment 2. BTC/BCC at 4000 m and other depths 

Response: BTC (energy transfer for KE between the eddy field and mean flow) is consistently about one order of magnitude larger than BCC (energy transfer for APE between eddy and mean flow). We extended the BTC and BCC calculations to additional depths (3000 m and 5000 m) in the deep water column. We find that the relative magnitudes of the two conversion terms remain similar to those at 4000 m.  

Comment 3. Missing units 

Response: We thank the reviewer for pointing this out. We have carefully checked all figures, colorbars, and captions. Missing units will be added where appropriate, and the figure captions will be revised for consistency. 

We thank the reviewer for these thoughtful comments, which helped us clarify the central message and strengthen the physical framework of the manuscript. We appreciate the concern regarding the broader significance of the comparison and we have carefully revised the manuscript accordingly. 

Comment 1. Choice of resolution 

Response: We thank the reviewer for this comment. Based on the scale and effective resolution of the represented bathymetry, we view the 1 km and 3 km configurations as distinct dynamical regimes in the Brazil Basin. The 3 km simulations could only marginally permit some smaller-scale motions with an effect resolution of approximately 7.5 km, while the GIGATL1 runs are indeed submesoscale-permitting in the Brazil Basin.  

Nevertheless, we understand the reviewer’s broader concern. To address this point, we have extended the analysis to include a 6 km horizontal resolution case, GIGATL6T, including tides. Also, for this case we performed the corresponding Lagrangian particle experiment. The 6 km case provides a coarser comparison and shows the expected progression with resolution: coherent deep eddies are largely absent, and bathymetric smoothing prevents the formation of a bottom boundary layer where mixing increases because of energy conversion when the currents interact with the topography. As a result, the diagnosed diapycnal diffusivity is lower than in the 3 km and 1 km simulations and nearly constant throughout the water column. We will add and discuss this case as integral to the revised manuscript (Supplement Fig.1).

Performing an even higher resolution run (which would require a nonhydrostatic configuration) is beyond the scope of the current work, which aims at quantifying with a very specific example why simply increasing model resolution to km scale and/or developing AI-based parameterizations from submesoscale permitting simulations may not reduce (global) ocean models structural uncertainty.

Comment 2. Bathymetric smoothing  

Response: 

We thank the reviewer for this comment. We agree that a direct numerical sensitivity test would provide a stronger assessment of the role of bathymetric smoothing. We are currently running a 1 km test simulation with downscaled 3km bathymetry which will be added when ready. Unfortunately, through the summer semester our supercomputer was under maintenance for a prolonged time, and the simulation is not yet completed.  

This comment, though, prompted us also to revisit the physical framework underlying this part of the manuscript. Energy stored in the mean flow can be transferred to the eddy field through barotropic and baroclinic instabilities, while eddy–internal-wave interactions can provide a sink of eddy energy and a source of internal-wave energy.  Polzin (2010) showed that internal waves propagating through a three-dimensional mesoscale strain field can exchange energy with the eddies through correlations between wave stresses and the mesoscale rate of strain. Both oscillatory tidal flows and slowly varying geostrophic currents interacting with rough bathymetry can radiate internal waves and exert topographic form drag (Bell, 1975; Nikurashin and Ferrari, 2010). Then nonlinear transfers toward smaller vertical scales enhance shear and ultimately support wave breaking and turbulent dissipation. 

The geostrophic mean flow in the simulations is broadly consistent across the runs (very similar between GIGATL1, 3 and 6, and from a very preliminary analysis in agreement with ARGO-derived estimated), suggesting that strong eddy fields in the simulations are not primarily associated with a bias in the mean-flow reservoir. In contrast, GIGATL1T exhibits substantially larger EKE and shear than GIGATL3T. The diagnosed horizontal energy conversion from the eddy field to the internal-wave field and the internal-wave energy also increase in GIGATL1T, consistent with the enhanced interaction between the eddy field and the higher-resolution topography. However, the enhancement in internal wave energy in GIGATL1T is much smaller than the increase in EKE. This supports our interpretation that, although the higher topography resolution in GIGATL1 allows more energy transfer from the eddy field to the internal-wave field relative to GIGATL3, internal-wave generation remains strongly limited even at 1 km resolution. As a result, the internal-wave pathway cannot provide a sufficiently efficient sink, especially for the enhanced eddy energy in GIGATL1.