In the paper, ADM-LES simulations of a ducted balloon wind turbine are performed for different inflow conditions and positions of the balloon. Results are then compared in terms of loads and velocity and turbulence intensity profiles in the wake.  In this frame, a new automatic framework for the generation of the grid is also proposed. The reviewer believes that the activity is innovative and worth of investigation. The methodology seems to be adequate and consistent throughout the whole activity. Nonetheless, a greater effort must be spent to highlight the originality and the contribution of the publication. Some specific comments:

• Little bit of background is required in the abstract;
• The introduction is complete and extensive, but lacks some structure. In particular, two questions remain unanswered: 1) what are the advantages and disadvantages of balloon turbines with respect to other AWE systems?; 2) what is the common layout that is expected for wind farms using this technology and how does it justify a high-fidelity  study of the wake of these machines?;
• The sections 2, 3.1 and 3.2 are very general and probably redundant for a technical publication like this. Please reduce them, reporting only the details specific to this work;
• Section 3.3: the use of XFoil in the preliminary design phase is acceptable (please refer to the approach, not the airfoilTools database). However, its use for final design and simulation is questionable, especially for airfoils like the S809 with pronounced stall characteristics. At least a validation of the adopted polars with the experiments available in the literature is needed. It would also useful to specify here what conditions are gonna be simulated and how the turbine is controlled;
• First part of Section 4 should be moved to Section 3, in order to complete the overview of the selected test case. To what real life operating conditions do the selected tilt angles correspond?;
• Line 234: what criterion was used for the computation of the timestep?;
• In the majority of the results there is a typo: “separation” instead of “separation”.

The reviewer recommends the publication of this paper after the comments above have been properly addressed

The paper conducts large eddy simulation (LES) of a ballon wind turbine, employing an actuator disk (AD) method to parameterize the effect of the turbine. The authors attempt to ensure consistent spatial resolution of the LES via an a priori estimation of the resolved turbulence kinetic energy. The results focus on the wake recovery in terms of the mean velocity and turbulence intensity, as well as the aerodynamic forces acting on the ballon. The study considers the effect of incoming flow direction and magnitude.

Overall, I found the paper to be rather lengthy in its description of the AD method and associated momentum theory, which is not novel. I expected to see sensitivity of the computational domain size, which was not provided and my intuition suggests is small, especially the upstream distance between the inlet and the balloon turbine. Secondly, the analysis to ensure consistent spatial resolution relies on a RANS simulation for estimating the turbulence kinetic energy and dissipation rate for calculating the turbulence integral scale. The details of the RANS simulation are not provided. Moreover, why the RANS solution can be considered accurate is not justified. Thirdly, the grid independence assesses the pressure coefficient distribution along the balloon periphery, which is not convincing for assessing grid independence of the results. More convincing would be to demonstrate the grid independence of the wake recovery, separation zone size and strength, and other parameters that would be expected to be more sensitive to the grid. Fourthly, the validation against experiment is not well documented--the authors do not comment on the spatial resolution of the validation study and whether it is consistent with the main study. The result is that the validation--which does show good agreement--does not convincingly demonstrate the accuracy of the main study results.

The present studies are intended at investigating the wake characteristics of a balloon wind turbine using an actuator disk theory solved through an LES solver. Although the topic itself is of interest, I do feel the authors need to improve the paper for publication. The following aspects are some comments from my side:

• The authors provide too detailed formulations on the LES and actuator disk theories while they do not attempt to improve them. These can be omitted.
• Numerical description is too weak. The authors shall provide more information about the approaches used to solve the sets of flow equations, e.g., time integration? discretization? convergence acceleration?
• Temporal discretization studies are not performed. How can we sure the solutions are accurate with respect to time size?
• Upstream domain looks a bit too small.
• Validation studies use conventional wind turbines, while the case being studied is much more complex.
• If the authors claim to use LES, is the energy in the proximity of the balloon wall resolved well?
• The authors focus the studies on the wake, but the mesh behind the balloon does not seem to be well refined such that it can well resolve small eddies. Perhaps plot the Q or Lambda2 criterion?
• Last minor aspect: remove the 6th point in Conclusion as nothing is written there.