1. In figure 2d, what is the gray color in Th3? 

2. For the supervised neural network, what is the network structure used in the study?

3.  In the confusion matrix, people have to check every element in the matrix to analyze the prediction accuracy. It is better to provide an overall/single index for an easy analysis.

4. In the analysis of the seismic cube for internal structure of the GRAME area, different subareas have been chosen in different technologies (Figs. 4-11, such as a,b,c,d,e). Could you focus on the same area and apply the different seismic data analytic methods for a consistent evaluation?

5. For the lithology prediction, what are the inputs for the classification?

6. Please rewrite the discussion section, as some new figures and equations are introduced. And also the discussion should be more focused on the main findings and limitation of the current research, as well as some suggestions for future researches.

7. It is also for the conclusion which is too long. Please shorten it.

The authors describe a case study for an advanced geothermal carbonate reservoir characterization

based on 3D seismic attribute analysis, lithological classification and structural imaging. 

The approach is demonstrated for a region with considerable geothermal potential and undergoing

development for geothermal utilization. The study site is located within the Molasse basin

in southern Germany, covering the city of Munich and surrounding area.

The paper is well written and well-structured. The work is based on data of high quality. 

Data analysis is carried out with existing state-of-the-art methods. Figures are of excellent quality. 

The results are described and discussed regarding both methodological aspects and geological interpretation.

The motivation for the study is reasonably explained by the very complex situation

and specific challenges of carbonate reservoirs for geothermal exploration.

The underlying data are described at the beginning of the Methods.

The authors could think about a separate chapter to describe the Data.

But the presented version is also clear enough with given references

for the data.

The advanced data analysis is nicely categorized into four seperate approaches:

+ single attribute analysis

+ multi-attribute analysis

+ neural network-based lithology classification

+ fracture orientation analyis

I suggest to extend a bit the explanation of the neural network-based clasification.

My assumption at this point is, that a few readers would like to know more technical

details such as 

+ type of neural network

+ architecture

+ specifications of network (e.g. learning rule, internal functions)

+ software used / implementation

Commendable for me in this manuscript is the critical discussion of seismic attributes

as given in the discussion chapter.

Maybe it would be worth to mention that seismic attributes are signal properties, and

not inherent rock physical properties.

As an example, frequency based attributes could be influenced by different factors 

including inherent seismic attenuation, complex geological structures

such as thin layers or gradient structures with potential shifts of signal frequencies, 

data processing or acquisition footprints.

At least from my perspective the formulation "reservoir control factors

that may affect the physical properties of the seismic signal" (line 550)

is suggesting that the signal properties of the seismic reflection waveform directly

represent subsurface rock physical properties. As descibed above, the causality is

more complicated in my opinion.

But this might be an overcritical comment.

The important support for the interpretation of results in this paper is given

by the presented empirical correlations between attributes and borehole data.

The compehensive attribute analysis and their interpretation is a very nice

contribution to improve the characterization of geothermal carbonate reservoirs.