Random Forest Classifier for Cloud Clearing of the Operational TROPOMI XCH4 Product
, Mari C. Martinez-Velarte 1, Maarten Sneep 2, Mark ter Linden 2 and Jochen Landgraf 1Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsPlease see the attachment.
Comments for author File: 
Comments.pdf
As marked above.
Author Response
Please find attached our response to the comments of the reviewer.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript presents an alternative approach that can be used to clean clouds from TROPOMI XCH4 data. Overall, I found the manuscript well written. The authors claim that the performance of this approach is comparable to that used with VIIRS filtering and even increases data availability. The validation of the technique is shown for data from 12 stations. It would be interesting to discuss whether this increase in data is not related to the presence of some type of clouds. It is well known that thin clouds could be interfering with the data. Is there any chance to ensure that this new data is not related to thin cloud? Are the stations analyzed close to any maritime coast? This could also be commented on since the presence of clouds in those areas can have a strong impact on data recovery. The seasonal regime must also be considered in the analysis and discussion, since the study focuses only on summer. In some areas, winter and spring are the seasons with the most clouds and therefore there is more noise in the observations. With this in mind I believe the manuscript will be suitable for publication
Author Response
Please find attached our response to the comments of the reviewer.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsComment
Manuscript “Random Forest Classifier for Cloud Clearing of the Operational TROPOMI XCH4 Product” submitted to Remote Sens by Tobias Borsdorff, Mari C. Martinez-Velarte, Maarten Sneep, Mark ter Linden, and Jochen Landgraf is devoted to the problem of cloud clearing approach based on a random forest classifier (RFC). The authors have done a lot of work to develop a method for clearing clouds to obtain accurate data.
In this study, authors introduced a new machine learning approach based on the Random Forest Classifier technique, which replicates VIIRS cloud clearing using only TROPOMI data. The classifier is trained on a subset of 5 years of collocated TROPOMI and SUOMI- NPP VIIRS data (about 20000 orbits). To this end, authors used parameters derived from the TROPOMI CO retrieval, which is inherently sensitive to the presence of clouds and is processed prior to the retrieval of TROPOMI XCH4 in the mission operational pipeline. This strategic choice simplifies the integration of approach into the existing processing framework.
In addition, in this work presented an efficient and robust method for despriping TROPOMI data, relying on median smoothing techniques. This approach was validated with ground-based TCCON measurements to show that it is not changing the bias, but slightly improves the standard deviation of the bias. This is a positive result since removing the stripe noise from the data should only improve the scatter and not introduce a bias change. The destriping will be suggested as an update for the operational TROPOMI XCH4 and CO retrieval in future. The results highlighted the effectiveness of RFC cloud clearing approach, showcasing its ability to match the performance of VIIRS filtering with a very similar bias and about 7 % more data. It will also apply RFC for improved quality filtering of TROPOMI XCH4 data in future. When the dependence on SUOMI-NPP data is eliminated, faster processing becomes possible.
1. Is there any deviation of XCH4 data value between SUOMI-NPP VIIRS and TROPOMI. If yes, which ones?
2. In the absence of data from SUOMI-NPP VIIRS, is it possible to determine XCH4 from TROPOMI data?
3. Is it possible to use the RFC cloud cleaning approach proposed to you to determine other greenhouse gases, for example, CO2, water vapor, NOx, etc.
4. In Fig. 3a, data over Siberia is given, but in Table 1 Siberia is not in the list. Please explain the criteria for using data over Siberia only in 2022?
5. Are XCH4 values for August 2022 compared with TROPOMI data for 2017-2021? What deviations are there at August, please?
6. RFC cloud scrubbing provides a slightly larger amount of data (2182 vs. 2035 daily averages) and may have fewer outliers, how to explain that the Pearson correlation coefficient remained the same in both cases at 0.9?
It should be noted that the article is written in an understandable language, not overloaded with unnecessary terminology. The conclusions of the authors are well founded. This work must be continuing for aim described in section of discussion: expanding its application to glint geometries observed over the oceans; apply RFC for improved quality filtering of TROPOMI XCH4 data in future.
In case when the dependence on SUOMI-NPP data is eliminated, faster processing becomes possible. This is in particular interesting for the scientific data application monitoring CH4 point sources or assimilating TROPOMI data in near-real-time as done by CAMS-IFS for TROPOMI CO.
The article is of scientific interest and is recommended for publication after minor revision.
Author Response
Please find attached our response to the comments of the reviewer.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors have replied to my suggestions and comments and modified the manuscript when necessary. I think the manuscript is in good shape now and ready for publication.
Comments on the Quality of English LanguageAs marked above.
Reviewer 2 Report
Comments and Suggestions for AuthorsI consider the manuscript suitable for publication in the journal, as the authors have made at least some comments on the points made in the previous review.
