In the Heat of the Night: Comparative Assessment of Drone Thermography at the Archaeological Sites of Acquarossa, Italy, and Siegerswoude, The Netherlands

Round 1

Reviewer 1 Report

The paper presents an interesting application of drones and overall is well written and presented. I would find it helpful to be guided as I am reading to the types of anomalies in the images are being looked at.

Some additional background in the thermal image processing and the methodology behind that could strengthen the paper as well. 

Why is there a greater field of view in the the 30 m image compared to the 100 m image in Figure 8? Are the two switched? 

There are minor typos or bad grammar throughout the paper that need attention. Some examples:

Unfinished sentence Line 550. 

Line 570 is missing units

Line 592, do should be changed to does

Among others. 

Author Response

Dear scientific editor and reviewers,

Thank you for the time invested in reviewing our paper. We are very happy to receive such detailed and valuable feedback, and we have worked hard to improve the paper accordingly. Below you will find the responses to your queries, in which we did our best to answer the questions, follow up on the suggestions (we indicated the actions we took, and the paper has been edited using track changes), and provide an explanation where we did not follow suggestions. The only revisions we did without track changes are those concerning references, that is now in proper MDPI style.

Best wishes,

Jitte Waagen, also on behalf of the co-authors

Reviews

Reviewer 1

I would find it helpful to be guided as I am reading to the types of anomalies in the images are being looked at.

>> Ok, we added more guides to anomalies in the manuscript.

Some additional background in the thermal image processing and the methodology behind that could strengthen the paper as well. 

>> Agreed, adapted section 2.2 (Analyzing heat) to include more detailed information on thermal file format and specifics of photogrammetric processing using Pix4D Mapper, as well as a link to an online guide into processing thermal images.

Why is there a greater field of view in the 30 m image compared to the 100 m image in Figure 8? Are the two switched? 

 >> Sharp observation! However, it is actually not a different FoV, it is just that the 30m images didn’t align properly in photogrammetric postprocessing, resulting in a distortion making the mapping too narrow in the east-west orientation. As this is inconsequential to both the interpretation and the purpose of this figure (showcasing a similar spectral mark recorded at two different altitudes) and technically uninteresting (because it was a result of using too few GCPs), We redid and cropped the image to avoid a strong visual difference and focus the reader on the content.

There are minor typos or bad grammar throughout the paper that need attention.

>> We did a thorough general check on language and sentences, using track changes, corrected the examples given by reviewer 1 and a many more.

Reviewer 2 Report

This work shows valuable data captured by a thermal camera using drones. However, the authors failed to present these data in a well-managed way. For example, there is even no section and sub-section number all over the manuscript. It also lacks and comparison or quantitative data to support the conclusion. I suggest the authors re-organize all the data and try to use the data to support scientific funding, rather than put all the figures there. The detailed procedure and setting of the drone for capturing these data should also be presented. Otherwise this paper is more suitable to be submitted to Remote Sensing rather than Drones. 

Author Response

Dear scientific editor and reviewers,

Thank you for the time invested in reviewing our paper. We are very happy to receive such detailed and valuable feedback, and we have worked hard to improve the paper accordingly. Below you will find the responses to your queries, in which we did our best to answer the questions, follow up on the suggestions (we indicated the actions we took, and the paper has been edited using track changes), and provide an explanation where we did not follow suggestions. The only revisions we did without track changes are those concerning references, that is now in proper MDPI style.

Best wishes,

Jitte Waagen, also on behalf of the co-authors

Reviews

Reviewer 2

For example, there is even no section and sub-section number all over the manuscript.

>> That is indeed sorely lacking, and now fixed, we adapted to the required Introduction, Materials and Methods, Results, Discussion and Conclusion sections, with appropriate paragraph numbering to improve clarity of the structure of the paper.

It also lacks and comparison or quantitative data to support the conclusion. I suggest the authors re-organize all the data and try to use the data to support scientific funding, rather than put all the figures there.

>> Quantitative support for the conclusion is not essential in our opinion, as the interpretation is subjective, and based on relative contrasts in relation with complementary sources of evidence. So we did not follow up on this point, although we provided, in response to other queries as well, a lot more information on the figures, legends and units.

The detailed procedure and setting of the drone for capturing these data should also be presented. Otherwise this paper is more suitable to be submitted to Remote Sensing rather than Drones. 

>> Ok, we added information on the drone technology in the beginning of section 2.1 (Recording heat). As for more details on the procedure and setting, we am not entirely sure what is being requested here; in this 2.1 section the flight operation, including software (and settings), flying strategy, ground control, etc. are being presented at a level of detail that is required to understand the resulting data, at least in our opinion.

Reviewer 3 Report

The presented research is on a topic of relevance and general interest to the readers of the special issue and the journal. The manuscript offers a much-needed perspective on the application of low-altitude thermography-based archaeological prospection with drones, delving into the thermodynamic characteristics of buried remains and their surrounding landscape matrix. I found the manuscript to be very well written and felt confident that the authors performed careful aerial surveys, and analyzed thoroughly thermograms, thermal ortho-image mosaics and surface models about the areas of archaeological interest. However, some improvements could be made regarding presentation and interpretation of results. Furthermore, a few points may require clarification before the manuscript is ready for publication.

Major Comments

I recommend following the authors’ instructions for sectioning the manuscript and numbering the sections, as the introduction, methods, and application should be better separated.

The four thermodynamic properties that affect soil, and therefore prospection via thermography, are extensively introduced in the theory section, and mentioned in the results sufficiently enough, but are missing from the methodology. There should be, in my opinion, at least some description of how they have been considered during data acquisition and how they affect the recording workflow.

Most of the figures need revising. Those that have been manipulated should clearly state how they have been manipulated. Those that depict reflectance, temperature, height difference, or amplitudes should have a color scale with legend. I would additionally recommend not using colors within the specific pseudo-color scale of an image to indicate features; for example, in Fig. 8, white has been used to indicate the identified features, but white is also used for specific temperature values. Also, the captions should state which features/potential features have been overlayed on the original images, and how the figures are partitioned.

There should be, in my opinion, a comparison with figures for both sites showcasing how the archaeologist (or another specialist) selects between all the produced thermal ortho-images (different times, heights, post-processing scenarios) the optimal scenario to perform the interpretation of features.

Lines 185–187. Has any thermal sensor calibration been performed beforehand to model or anticipate such sensor noise?

Lines 201–202. Of these algorithms, probably the one related to texturing would be worth mentioning, as it is essential to transfer temperature values correctly (and their spatial variation) from captured images to thermal ortho-image mosaics.

Minor Comments

Lines 44–50. A few important publications the authors may have missed:

     Adamopoulos, E.; Rinaudo, F. UAS-Based Archaeological Remote Sensing: Review, Meta-Analysis and State-of-the-Art. Drones 2020, 4, 46, doi:10.3390/drones4030046.

     James, K.; Nichol, C.J.; Wade, T.; Cowley, D.; Gibson Poole, S.; Gray, A.; Gillespie, J. Thermal and Multispectral Remote Sensing for the Detection and Analysis of Archaeologically Induced Crop Stress at a UK Site. Drones 2020, 4, 61, doi:10.3390/drones4040061.

     Šedina, J.; Housarová, E.; Raeva, P. Using RPAS for the Detection of Archaeological Objects Using Multispectral and Thermal Imaging. European Journal of Remote Sensing 2019, 52, 182–191, doi:10.1080/22797254.2018.1562848.

     Thomas, H. Some like It Hot: The Impact of next Generation FLIR Systems Thermal Cameras on Archaeological Thermography. Archaeological Prospection 2018, 25, 81–87, doi:10.1002/arp.1588.

Lines 218–220. Perhaps the manuscript does not need to be too technical from a geomatics point of view, but at least some information regarding the obtained spatial resolution on the thermal orthoimage-mosaics would be essential to understand the size of features that can be identified with each sensor and height scenario.

Lines 241–242. It is unclear how the ‘thermal 3D model’ would allow better interpretation, since the thermal orthoimages and the surface model must be processed further separately to allow interpretation. It may be a good option for visualization purposes, but it is not distinguishable from the images. This section should be expanded or rephrased.

Lines 329–332. Has any post-processing helped with identifying feature B? it is almost indistinguishable.

Lines 352-359. This section is slightly confusing. Please rephrase in a clearer manner, considering what the scene sampling distances are during acquisition, the size of the investigated remains, and what parameters of the camera/optical system are relevant for images that are going to be used for photogrammetry and for images that are going to be used directly.

Lines 375–376. By ‘calibrated’, do the authors mean undistorted? How did the authors perform this process? Were these image derivatives extracted from the SfM software after the photogrammetric block processing?

Lines 664–665. Since SfM software has been used to produce ortho-image mosaics, is the sacrifice of area coverage over spatial resolution an issue? Please rephrase accordingly.

Author Response

Dear scientific editor and reviewers,

Thank you for the time invested in reviewing our paper. We are very happy to receive such detailed and valuable feedback, and we have worked hard to improve the paper accordingly. Below you will find the responses to your queries, in which we did our best to answer the questions, follow up on the suggestions (we indicated the actions we took, and the paper has been edited using track changes), and provide an explanation where we did not follow suggestions. The only revisions we did without track changes are those concerning references, that is now in proper MDPI style.

Best wishes,

Jitte Waagen, also on behalf of the co-authors

Reviews

Reviewer 3

I recommend following the authors’ instructions for sectioning the manuscript and numbering the sections, as the introduction, methods, and application should be better separated.

>> See above, we adapted to the required Introduction, Materials and Methods, Results, Discussion and Conclusion sections, with appropriate paragraph numbering to improve clarity of the structure of the paper.

The four thermodynamic properties that affect soil, and therefore prospection via thermography, are extensively introduced in the theory section, and mentioned in the results sufficiently enough, but are missing from the methodology. There should be, in my opinion, at least some description of how they have been considered during data acquisition and how they affect the recording workflow.

>> We discussed in 3.1.2 and 3.2.2 (Thermography capturing strategy and conditions) the considerations about the local thermodynamic properties of soil and expected archaeology, in 3.1.3 and 3.2.3 (Spectral marks) the effect of those conditions on the acquired data, and 3.1.5 and 3.2.5 (Interpretations) overall interpretation on the interplay of thermodynamics, acquired data and archaeology. About the recording workflow, since capturing at both sites is part of a systematic test using so they can be compared, workflows were intentionally applied in the same manner, and not changed because of local thermodynamic conditions. We am not really sure what we should add further in response to this query.

Most of the figures need revising. Those that have been manipulated should clearly state how they have been manipulated. Those that depict reflectance, temperature, height difference, or amplitudes should have a color scale with legend.

>> Yes, that’s very true, so we now adapted the figures adding colour legends and units. Of course the temperatures are only approximations, as the calculations for true temperature readings have never been made, because those are irrelevant for their purpose in this research.

We left figure 2 and 4 as they are, since 2 is just an example of features that are not historically/archaeologically interesting, and 4 is just an example of visualisation (we reduced the importance of the figure in response to another comment, so it is now just a demonstration).

The amplitudes of the GPR have been normalized, so adding a colour legend does not make anything clearer.

We adapted all captions to describe accurately the data visualised in the figures.

I would additionally recommend not using colors within the specific pseudo-color scale of an image to indicate features; for example, in Fig. 8, white has been used to indicate the identified features, but white is also used for specific temperature values.

>> Ok, true indeed, redid the figure and changed to red.

Also, the captions should state which features/potential features have been overlayed on the original images, and how the figures are partitioned.

>> See above, we rewrote all captions.

There should be, in my opinion, a comparison with figures for both sites showcasing how the archaeologist (or another specialist) selects between all the produced thermal ortho-images (different times, heights, post-processing scenarios) the optimal scenario to perform the interpretation of features.

>> The time and altitude comparison is there for Siegerswoude (figure 8 and 9), which was explicitly part of the test. To limit the use of similar figures showcasing the same affects, we did not show the altitude and time comparison for Acquarossa, which is very much the same. As for the postprocessing, the now cited technical workflow was used and no testing was performed between various postprocessing scenarios or visualisations.

Lines 185–187. Has any thermal sensor calibration been performed beforehand to model or anticipate such sensor noise?

>> No, and we are personally not aware of (but would be happy to learn about!) any procedure possible with the used equipment to deal with vignetting (due to the heating of the sensor body) or the pixel-to-pixel variation for which the Flat Field Correction was done.

Lines 201–202. Of these algorithms, probably the one related to texturing would be worth mentioning, as it is essential to transfer temperature values correctly (and their spatial variation) from captured images to thermal ortho-image mosaics.

>> Good point, but there is unfortunately no information on the exact texturing algorithm from Pix4D that we could find. However, Pix4D produces the images using the radiometric metadata part of the thermograms to produce the index maps, which are Float32 raster files, capable of very accurately displaying the recorded values. We added a section on this that to the text.

Minor Comments

Lines 44–50. A few important publications the authors may have missed:

     Adamopoulos, E.; Rinaudo, F. UAS-Based Archaeological Remote Sensing: Review, Meta-Analysis and State-of-the-Art. Drones 2020, 4, 46, doi:10.3390/drones4030046.

     James, K.; Nichol, C.J.; Wade, T.; Cowley, D.; Gibson Poole, S.; Gray, A.; Gillespie, J. Thermal and Multispectral Remote Sensing for the Detection and Analysis of Archaeologically Induced Crop Stress at a UK Site. Drones 2020, 4, 61, doi:10.3390/drones4040061.

     Šedina, J.; Housarová, E.; Raeva, P. Using RPAS for the Detection of Archaeological Objects Using Multispectral and Thermal Imaging. European Journal of Remote Sensing 2019, 52, 182–191, doi:10.1080/22797254.2018.1562848.

     Thomas, H. Some like It Hot: The Impact of next Generation FLIR Systems Thermal Cameras on Archaeological Thermography. Archaeological Prospection 2018, 25, 81–87, doi:10.1002/arp.1588.

>> We did not miss James et al and Thomas which are in the paper, but indeed the other two were yet unfamiliar. We are very happy with the suggestions for the other publications.

Lines 218–220. Perhaps the manuscript does not need to be too technical from a geomatics point of view, but at least some information regarding the obtained spatial resolution on the thermal orthoimage-mosaics would be essential to understand the size of features that can be identified with each sensor and height scenario.

>> We agree, we will add the appropriate statements on GSD.

Lines 241–242. It is unclear how the ‘thermal 3D model’ would allow better interpretation, since the thermal orthoimages and the surface model must be processed further separately to allow interpretation. It may be a good option for visualization purposes, but it is not distinguishable from the images. This section should be expanded or rephrased.

>> Agreed, although the basic photogrammetric thermal index is still useful to project on the 3D geometry, as it elucidates basic relations between the two datasets. We will rephrase accordingly.

Lines 329–332. Has any post-processing helped with identifying feature B? it is almost indistinguishable.

>> Just some contrast enhancement, but it is indeed much better visible on figure 8.

Lines 352-359. This section is slightly confusing. Please rephrase in a clearer manner, considering what the scene sampling distances are during acquisition, the size of the investigated remains, and what parameters of the camera/optical system are relevant for images that are going to be used for photogrammetry and for images that are going to be used directly.

>> We am not entirely sure what is being requested here.

Lines 375–376. By ‘calibrated’, do the authors mean undistorted? How did the authors perform this process? Were these image derivatives extracted from the SfM software after the photogrammetric block processing?

>> These are individual, single thermograms directly from FLIR Thermal Studio, no photogrammetric postprocessing. I added ‘individual’ to make that clearer. The calibration here means having FLIR Thermal Studio calibrate sensor readings for distance, atmospheric temperature etc. for better temperature readings, but this is actually not relevant for these images, so I removed the term ‘calibrated’ to avoid confusion.

Lines 664–665. Since SfM software has been used to produce ortho-image mosaics, is the sacrifice of area coverage over spatial resolution an issue? Please rephrase accordingly.

>> Yes due to time (battery) limitations this means you simply capture less area, notwithstanding the possibility of stitching images together into larger datasets. Rephrased to make that clearer.

Round 2

Reviewer 2 Report

The authors have modified the work according to the reviews' comments.