Rapid-DEM: Rapid Topographic Updates through Satellite Change Detection and UAS Data Fusion

Round 1

Reviewer 1 Report

The article presents the technology of rapid updating of landscape information by combining different types of remote sensing data.  Detection of changes with mapping is performed by satellite images. DEM updating is performed by fusion of DEMs obtained by UAS and open-source DEMs. Using a priority queue allows mapping and updates to be directed to the areas with the most change. The proposed rapid update approach will undoubtedly be useful in analyzing changing territories. The article is written in good technical language, and the results of the development application are unmistakable.

The reviewer has only two comments:

1. It is necessary to define Spatial Resolution USGS NED DEM. They are different in the text (line 285) and in the table.
2. In Figure 10 (C) DSM looks like corrected (B) DTM. If this is not a mistake, there are questions about the photogrammetric processing algorithms settings.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This manuscript presents a framework to update DEM. Generally, the contribution is good and the manuscript is well structured. Several concerns should be considered to further improve the manuscript.

1. I do not agree this framework/process is a semi-automatic one. I understand that machine learning methods and human interference are both required in this process. When it comes to the term “semi-automatic”, however, we expect to see a more intelligent and efficient method. The classification and change detection are not new but based on the supervised approach and post-classification comparison, therefore the so called semi-automated framework should be revised.
2. The introduction part is a bit short. The synthetic utilization of multi-source data is a good point in this manuscript, thus the research status of the related part can be reviewed.
3. The barren land cover decided by BSI and NDVI may not be suited when using a different dataset. Why not just classifying the barren land class in the RF method?
4. It is confusing to use objects/polygons since these terms have an explicit but different meaning (usually refer to segments) in an object-based analysis, while your classification method is based on pixels. Thus a proper explanation is necessary.
5. Figure 2c is unnecessary, but a link between Figure 2a and 2b should be added to show the exact part of the study area in the Wake County.
6. You have listed classification features in Table 3. How can the samples include 12363 features? It doesn’t make sense.
7. Figure 4 is not appropriate. It is suggested to present the spatial distribution of samples in the Walnut Creek image. The box plot is normally used to show the spectral distribution of sample data.
8. I can barely identify the curve of fused DEM in Figure 13. Please make the plot better.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

All issues have been addressed.