The article highlights the work on active seismic tomography by building a 3D velocity model using two different approaches, e.g., thin- and fat-rays in a geothermal testbed characterized by a fault structure and associated seismicity. 8 boreholes available in the vicinity of the study area were used for this purpose. The results are correlated with the wireline logs and the geological observations. The article in its present form has several deficiencies in terms of technical details, presentation of the results, and the conclusion. Please find the attached .pdf file with specific comments highlighted over each row. Here, I am only mentioning the main points from my side. They are as follows: 

1. The title involves four keywords: fault-structure, geothermal, seismicity, and tomography. Currently, the "Introduction" section provides no information on the geological setting of the geothermal testbed, the main fault structure (MFZ), and the seismicity in the area. In my opinion, this information should be introduced briefly, clearly stating the problem statement. 

2. A major work is required on the figures. For example, in Fig. 1a, it is very hard to read anything in the inset figure - the same goes for the legends used to describe the geology.  Fig. 1b shows the fault zone, but is not described at all anywhere in the text, only referenced in section 3 dedicated to data description (Row - 103). A proper re-writing of the figure description is also required. 

3. Section 5 requires a major re-schuffling in my opinion. It was very odd to read the inversion setup, then the tomography comparison between thin- and fat-rays, and finally visualize the velocity model. In my opinion, section 5.3 should be first with information on inversion setup, then 5.2, and finally 5.1.  

4. More technical details are required on the tomography. Currently, it is only descriptive without full details! I suggest also showcasing the picks against offset information or a similar approach, a resolution test for the obtained 3D velocity model (checkerboard?), ray-path coverage, etc., to provide the readers more confidence. 

5. Section 5.7 should be rewritten. Currently, it lacks a proper flow of information. There are a lot of sudden jumps between sentences without proper explanation. I could not understand what the author(s) mainly want to convey here.  

6. Several claims were made in the conclusion without any discussion or proper description in the article. For example, it is being said in Row-345 that "ray tomography is less dependent on the model parametrization of the forward and inversion grids (compared with thin rays)" without a discussion in the article. 

7. At places, a few words had been very loosely used. A few examples are: 'a remarkable spatial correlation' (Row-9), 'a nice signal-to-noise ratio' (Fig. 2), 'we show in this paper' (Row-184), 'in the middle of the volume' (Fig. 9), etc. I suggest avoiding such a form of writing unless it is clearly defined what it means.

8. Lastly, throughout the article, several other studies done in the same site have been referenced. At times, the main outcomes of those studies are either not stated or the main information is only described. In my opinion, those results should be shown in this article rather than just referencing the readers to them, especially when we are using that information to cross-check our results. For details, please see the .pdf file.

I hope the provided comments are valuable to the authors and will help in rectifying this article. 

Schwarz et al. applied fat ray travel-time tomography to calculate 3D velocity models and compiled a large dataset of 42,843 manually picked first-arrival waves from eight boreholes. This manuscript demonstrates that the fat ray method has higher image quality than traditional methods. In addition, this manuscript validates the derived 3D model using ground truth information from logging and geological observations. In summary, the authors have done a lot of literature review, but there is a lack of information on the geological setting of the geothermal study area, fault structure, distribution of ground monitoring stations, and regional seismicity, which needs to be provided and why high-resolution imaging is needed. In addition, there are still some deficiencies in the verification process of the results. For example, whether the total inversion parameters of the entire model match the data. Although the authors use coverage to show this, it is still unclear about the scale that can be resolved. The checkerboard test is helpful in defining the resolution scale. The verification of well logging data can only show the resolution of certain points in depth, but it cannot guarantee that the same results are maintained at other locations in the study area. In addition, the selection of damping and smoothing parameters does not seem to be clearly explained in the text. After all, this is an important parameter for inversion. The following are my main comments on the manuscript:

Line 111: The authors explain that they can compile a relatively large data set, including 42,843 manual P-picks, with an average picking uncertainty of about 0.15 ms. What method or software is used to estimate such a result? Please explain the details.

Line 139: Regarding the argument that the discretization for the forward modeling and the ray segment lengths for the inverse problem can be different, please provide a reasonable explanation about this argument.

Line 170: Why was the homogeneous model (5300 m/s) chosen for the initial model? Are there no previous 1D velocity model results for this study area? After all, the 3D velocity model obtained in this manuscript appears to be layered (Lines 200~207).

Line 236: The authors state that the comparison of the 3D seismic model with the borehole logs and core shows good consistency. Is there any literature or other supporting data to verify the borehole data?

Line 242: This manuscript observed a general decrease in velocity along the MFZ, but also a large amount of heterogeneity. This may be partly due to the limited spatial resolution of the tomography images. So how to define the spatial resolution of this 3D model? After all, the authors have used a very fine grid distribution. Although part of the reason has been explained in Section 5.6, it is still unknown how large the anomalies can be observed for the overall model.

Line 285: What is the magnitude distribution of the selected earthquake events? What is the temporal distribution of the records? Please give a detailed explanation.

Since this manuscript images a main fault zone with a very complex structure and strong inhomogeneity, there are still many doubts about the above inversion or verification process, so I am reserved about the interpretation of the results. The authors need to restructure the entire article because it is written in a jumpy way and sometimes it is really difficult to understand the meaning of the text. This manuscript needs a lot of revision work to make it have enough scientific volume.