Thank you very much for your interesting contribution to the Special Issue. The manuscript discusses an important issue and should be of relevance for a broader audience. The detailed analysis of accuracy performance of different data sources for very small scale geomorphological features from a large variation of observation distances is significant. Especially, the focus on GCPs used by different data sources for referencing and its significance for multi-temporal assessments of (i.e. low-magnitude) change detection is interesting. During the inter-active discussion process the authors explained their approach and answered all the comments by the referees in detail and made changes to the manuscript, where necessary.
However, there are two important aspects, which need further consideration. On the one hand, the assessment of one small excerpt within a large area captured via high resolution topography (HiRT) methods raises questions about the transferability of the results to other applications and the statistical foundation of the accuracy performance. For instance, how would different topographies and surface properties (e.g. texture) affect the performance of each method? This would also be relevant considering multi-temporal data acquisition. Thus, the investigations will need extension to other locations, e.g. at the same large field that has been captured already. This should be highlighted in the manuscript. On the other hand, it has been mentioned that different settings of parameters and its influence at the final model were not assessed as they were beyond the scope of this study. However, this can result in significantly different surface models. How do you assure that your settings are the most suitable (also in direct comparison to the TLS data set)? This needs more consideration, experiments and explanation in the manuscript.
Furthermore, there are some minor issues, described below, which need to be discussed.


Minor changes:

Accuracies regarding the SfM adjustment performance after GCP implementation are interesting information, especially because the authors perform a multitude of different SfM variations, which should be included in the study. Please, add these values.

In the abstract the authors state identical method and equipment can result in incorrect estimates of multi-temporal data due to registration error. If the registration approach is similar and assumed error of the GCPs nearly negligible, this cannot be the case. However, there are deviations, if measurement of GCPs (either in RGB images or intensity data of Lidar) is not exact (e.g. due to target fitting). Please, rephrase to a less rigid manner.

In the introduction, please shorten line 12 to 21 at page 2 because the authors do not use airborne Lidar in their study and thus the reference to their study is not as obvious.

In chapter 2.2 please state the accuracy of the applied TLS device.

Please, state the accuracy of each GCP after 20 seconds collection interval (chapter 2.2.1) because as the work focuses on the impact of GCPs regarding (multi-temporal) registration aspects.

In chapter 2.2.2 at line 31 page 4, do the authors mean distance values (eventually leading to elevation values at smooth, plane surfaces) given the distance measurement accuracy (instead of vertical accuracy) because multiple reflections of the same surface originally result in this error due to the random measuring noise of the device? This can lead to height deviations but if the surface is more complex also positional errors are possible. Furthermore, please state how the 2 mm locational accuracy was measured?

In chapter 2.2.4 page 5 line 13 and line 28 please remove exterior because both put together - camera position and orientation - are referred to as exterior camera orientation

In chapter 2.3 please consider that the usage of points coinciding with GCPs is difficult in the case of SfM because in dependence of the GCP size image matching at the GCP location can be limited due to missing texture in the middle of these marked targets. Furthermore, in the approach of plane estimation the plane is �filled� with points for subsequent ICP. Please, add this information because else the mentioned point cloud could be misinterpreted with the actual ephermal gully representing data. Thus, furthermore scale becomes irrelevant during ICP because the points are created artificially as plane representation.

Page 7 line 3: Undersampling solely becomes relevant for positional accuracy of the interpolated model but not the points themselves. Please, rephrase.

The merit of the consideration of data scoring is questionable because there is no explanation to which extent different parameters are weighted or should be weighted. Please, be more specific to this regard.

In the results the authors refer to standard error but no explanation has been given. Please, add this information.

The evaluation of horizontal and vertical displacement reveals mainly alignment errors. Does this mean if this could be corrected or taken care of, deviations between the different methods become almost negligible? This would be an interesting result. However, because TLS accuracy has not been stated, yet, it is difficult to assess if the measured deviations are within the error domain of the scanning device, potentially neglecting any statements regarding the accuracy performance of all the tested methods. Furthermore do the results suggest that for ephermal gully observations solely close-range measurements are suitable?

In the discussion it is stated that GCP consideration cause great difficulties regarding multi-temporal assessment due to remaining alignment errors. Did the authors consider the GCP accuracy parameter, which is set during adjustment procedure? Was it close to zero, because else the models are not �lashed� to these positions during adjustment potentially resulting in different alignments for different data. Could the authors please be more specific?

In the conclusions, please moderate the statement regarding the necessity of GCPs for multi-temporal assessment. The authors refer to ephermal gullies but in other studies concentrating on change detection of larger magnitudes this might not apply.

Table 1 needs more specific explanation because in its current form the ground based SfM data acquisition scheme is not comprehensible. An additional figure might be helpful.

Table 4: What does the red line at the bottom (Lidar) refer to?

Please, add a legend to figure 3C because relief features are difficult to identify. Furthermore, the blue graphic does not have good quality. There can hardly anything be identified.

Figure 9 and 10 need larger font sizes.

Dear authors,

thank you for revising the manuscript. All stated concerns have been adressed thoughtfully and with great detail. The manuscript improved significantly. In its present form it will be a valuable resource to the scientific community studying ephermal gullies with high accuracy and resolution.
Thank you for considering this Special Issue for your submission.

Kind regards

Anette Eltner

Dear authors,

Considering the AE report and reviews, your paper has clearly received considerable scrutiny and constructive feedback that I can see you have taken on board during revision. I am satisfied with the outcome of this process and pleased to confirm acceptance for publication.

During the proof corrections, you might want to keep an eye out for the following. I suspect the copy editors may have additional corrections for consistency, for example in the use of commas and semicolons.

p. 1, line 10 -- Over time should be two words
p. 1, line 34 -- there is no Figure 13, and it also seems unusual for this reference to be the first figure referenced in the paper, unless you mean Figure 13 of the Vinci et al. reference, in which case I recommend clarifying this point
p. 2, line 12 -- choose between airborne and air-borne (both are used in this line)
p. 4, line 38 -- as a result?
p. 10, line 34 -- show, not shoe
Table 1 -- here the terrestrial surveys are abbreviated as "Grd" while in most of the manuscript they are referred to as "Ground"
Figures 6 & 7 -- same issue of using "Grd" here instead of "Ground" elsewhere

Thank you for submitting your work to Earth Surface Dynamics!