An Enhanced Residual Feature Fusion Network Integrated with a Terrain Weight Module for Digital Elevation Model Super-Resolution

Round 1

Reviewer 1 Report

Dear authors, 

I found your manuscript a well developed idea, comparing its performance against other well-known approaches, showing your model better metric scores under the two scales of reconstruction.

Also, I enjoy how you take every module (attention, losses) and evaluate the models performances with and without the improvements, clearly showing their vital role to get better results.

It is interesting how the errors are greater near to the valley line (Figure 14). What do you consider would cause this behavior? weighting? D8 algorithm?, other?

The images and tables were pretty illustrative and allows a clear interpretation.

Best regards, 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this paper, an enhanced Residual Feature Fusion network (ERFFN) with fused terrain features is proposed to meet the needs of data reconstruction of missing parts of high-resolution DEM data or only obtaining a small range of similar terrain. There are some advice I give to you.

1.Unclear interpretation of model structure. 

The structure of Figure 3 is wrong. What parts do the four residual blocks refer to? The local features obtained are not identified. Finally, the input of the next RFM is obtained. 

2. The ESRAM module is used in the model is not identified in Figure 2, whether the RFM module in Figure 2 is RFM or ESRAM.  You should directly write the module after the combination of two modules.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript "An enhanced residual feature fusion network integrated with a terrain weight module for digital elevation model super-resolution" is clear and well-written. The topic is very interesting and the methods are adequately described, results and discussion are clear as well. However, I have some remarks concerning the editing of the paper:

1) Line 168 - rec onstruction - remove space between letters c and o

2) Figures: 1, 6, 11, 12 - Thousands should be separated by comma. It is missing in legend of elevations. 

3) Figure 7 - scale bar under China - values should be 'round', e.g. 500 m, 1000 m, 2500 m

4) lines 225 - 233 - justification

5) Line 228 - citation format is inconsistent 

6) Sections 3.3, 4.6 - justification

7) Line 360 - 'In Figure 8 shows' - should be 'Figure 8 shows'

8) Table 1 - why all ERFFN values are bold, despite not being the lowest in some cases? It may confuse the reader. 

9) Line 452 - 'the MAE of the terrain lines evaluation all show worse performance than the overall MAE of the DEM image'  - check the correctness of this sentence 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

The theme of this paper is of great significance. In addition, the paper has a clear description in terms of structure, principle description, map presentation, experimental verification, conclusion and discussion, and is an excellent paper. However, several aspects of the paper need to be done with minor revision before published.

1) The key words of this article do not reflect the main features and innovations of this article. It is suggested to replace them.

2) In the experimental verification, only the classical statistical MAE, RMSE and so on are used as the evaluation indicators. Can the machine vision evaluation indicators be added. From the experimental setup, there is still no ablation experiment.

3) In the conclusion of the paper, the limitation analysis of the method lies not only in the data, but also in the technology itself, which needs some analysis.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf