Cloud Shadow Detection via Ray Casting with Probability Analysis Refinement Using Sentinel-2 Satellite Data
Round 1
Reviewer 1 Report
“Cloud Shadow Detection via Ray-casting with Probability Analysis Refinement using Sentinel-2 Satellite Data”
General comments
The article proposes a novel cloud shadow detection algorithm that utilizes probability analysis of geometric relations outputs in improving the shadow mask results as elaborated in sections 2.2.1 cloud detection, 2.2.2 candidate shadows, 2.2.3 satellite viewpoint and sun position, 2.2.4 ray casting scene, and 2.2.5 statistical improvements.
Specific comments and recommendations
1. It is not clear on how the experiments are setup/designed to obtain the results that are presented in section 3. Results.
2. In table 7, the in the first column, last row, the authors should elaborate on the %’s 226%, 679%, and 2018%, as it is not clear neither explained in the discussion section.
3. In page 6, section 2.2.1, the authors discuss two types of probability, i.e., cloud probability (CLP), and Cloud probability (CLD), I suggest that the use similar abbreviations e.g. CLP1 and CLP2 to signify dealing with the same concept under different circumstances.
4. In page 7, section 2.2.1, line 238-239 “We apply a flood-fill algorithm using the 8-neighbours of each valid cloud pixel to identify all the clouds in an image”, the authors should provide a reference/citation to the algorithm which they refer to.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
In this manuscript, the authors have proposed a method for determining cloud shadows via geometric refinement using a fast inverse texture mapping ray-casting process. The paper is well-written, but I have the following comments:
1. The ‘Introduction’ and ‘The Related work section’ is quite repetitive. The authors should either get rid of the repetitive parts from any one of the section or combine them to a single section.
2. What is the motivation behind choosing that particular study area?
3. For the cloud detection, it is not exactly clear whether the authors have introduced any processes on the own or they have completely used the algorithm used by Sentinel-2 for cloud detection. The authors need to clarify this.
4. How are the false positives for the cloud shadows removed?
5. For the ray-casting, the authors have introduced few simplifications to decrease the computational cost. The authors need to analyse how these simplifications might negatively affect the results.
6. The author should provide a more detailed discussion why their method gave better results compared to the other methods already available in the literature.
7. It seemed that the authors have considered the dataset where the area was quite green. How will the results be affected if the area was covered with snow?
The English is quite okay
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have greatly improved their manuscript. The Authors addressed my concerns.
