the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Apr 2023
Optimal variables for retrieval products
Abstract. The increase of satellite instruments sounding the atmosphere will make it very likely that several instruments simultaneously measure either the same vertical profile or vertical profiles related to nearby geo-locations, and users will consult fused products rather than individual measurements. Therefore, the retrieval products should be optimized for the use in data fusion operations, rather than for the representation of the profile. This change of paradigm raises the question if a more functional representation of the retrieval products exists. New variables for the retrieval products are proposed that have several advantages with respect to the standard retrieval products. These variables, in the linear approximation of the forward model, are independent of the a priori information used in the retrieval, allow to represent the profile with any a priori information and can directly be used to perform the data fusion of a set of measurements. Furthermore, the use of these variables allows to reduce to about one third the stored data volume with respect to the use of standard retrieval products.
- Preprint
(404 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on amt-2023-42', Anonymous Referee #2, 23 May 2023
General comments
The proposed concept aims to significantly decrease the volume of the processed data during a fusion of several profiles obtained from atmospheric composition measurements using remote sensing. Indeed, in future we may effectively have several profiles collocated or being close in term of location. The formalism for the concept is well described. However, even if it is not quite the objective of this piece of work, the results of this concept need to be confronted to those of the classic retrieval method and more importantly to be validated by independent observations.
Specific comments
See in the attached pdf file
-
AC1: 'Reply on RC1', Simone Ceccherini, 04 Jul 2023
I thank the reviewer for the useful comments. In the following, I answer the general comments (included in “boldface” for clarity). In the attached pdf file, I answer the specific comments and, whenever required, I describe the related changes implemented in the revised manuscript.
General comments
The proposed concept aims to significantly decrease the volume of the processed data during a fusion of several profiles obtained from atmospheric composition measurements using remote sensing. Indeed, in future we may effectively have several profiles collocated or being close in term of location. The formalism for the concept is well described. However, even if it is not quite the objective of this piece of work, the results of this concept need to be confronted to those of the classic retrieval method and more importantly to be validated by independent observations.
New variables for the retrieval products are proposed and the advantages of these variables with respect to those generally used so far are analyzed on a theoretical basis. Validation of these variables by means of application to real cases requires the involvement of data provider and data user communities and should be a follow up of this paper. I agree with the reviewer on the importance of the validation of the method and indeed the conclusions of the paper invite the communities of data providers and data users to test and validate the efficiency of this new interface.
Specific comments
See in the attached pdf file
Answers in the attached pdf file.
-
AC1: 'Reply on RC1', Simone Ceccherini, 04 Jul 2023
-
RC2: 'Comment on amt-2023-42', Anonymous Referee #1, 12 Jun 2023
It is unclear to me what this paper adds. It appears that the paper is a continuation of a discussion in the paper by Ridolfi et al (2022) which is more complete than the current document.
The terminology is confusing: the author is using a term data fusion. maybe i do not understand this but if i look at how the Ridolfi et al is defining data fusion it looks more like a data combining - averaging than fusion. I would understand data fusion more as a combining two different data sets from totally different sources (e.g. model and observations) than combining two different observational datasets. When combining observations with model simulations, reference is made to data assimilation, which is a different topic then discussed here.
Here the author discusses a method to combine two different observational datasets.
The abstracts suggests something new. However, already for decades there are instruments mounted on the same satellite platform, making observations in different spectral regions of the same atmospheric properties: vertical profiles of T, q, O3, etc. For iinstance IASI-MHS or HIRS and AMSU etc. So there is already a wealth of real data where new ideas can be tested. We do not have to wait for future missions like FORUM or IASI-NG, before the value of a new approach can be demonstrated.
The use of averaging kernels to mitigate the difference in vertical resolution and to mitigate issues with the mean state of the background has been reported. Rodgers has published several papers on this. And this is thus an accepted procedure to compare retrievals from different instruments. And for sure can used to derive new products which enables an more comprehensive way to combine products from different sources as discussed by the author.
But i believe there are some more fundamental problems then vertical resolution when combining similar products derived from semi simulateous observations by different instruments. Each instrument has different characteristics, viewing geometry, spatial sampling, Point Spread Functions etc. How these problems affect the final product is not addressed in the paper.
The author should demonstrate the approach he is advocating on real data.
My read from the paper is that the data producers should not only distribute the products, but also averaging kernell information. This is only possible if the product is generated using the methods based on OE.
If observations are not nearly coincidence in space and time, some sort of assumptions about the time/space evolution of the derived products needs to be made. It can be expected that the more complete the assumption the better the combined product will be. In that case Data Assimilation is around the corner and we would not need the new products proposed by the author as level 2 data assimilation introduces different problems, in particular correlated errors. Level 2 data assimilation is possible, but requires a different set of products.
In summary: to bring the proposed methodology to the attention of the target user community, the author should spend time to apply the method to some practical examples and then re-submit.
Citation: https://doi.org/10.5194/amt-2023-42-RC2 - AC2: 'Reply on RC2', Simone Ceccherini, 04 Jul 2023
Status: closed
-
RC1: 'Comment on amt-2023-42', Anonymous Referee #2, 23 May 2023
General comments
The proposed concept aims to significantly decrease the volume of the processed data during a fusion of several profiles obtained from atmospheric composition measurements using remote sensing. Indeed, in future we may effectively have several profiles collocated or being close in term of location. The formalism for the concept is well described. However, even if it is not quite the objective of this piece of work, the results of this concept need to be confronted to those of the classic retrieval method and more importantly to be validated by independent observations.
Specific comments
See in the attached pdf file
-
AC1: 'Reply on RC1', Simone Ceccherini, 04 Jul 2023
I thank the reviewer for the useful comments. In the following, I answer the general comments (included in “boldface” for clarity). In the attached pdf file, I answer the specific comments and, whenever required, I describe the related changes implemented in the revised manuscript.
General comments
The proposed concept aims to significantly decrease the volume of the processed data during a fusion of several profiles obtained from atmospheric composition measurements using remote sensing. Indeed, in future we may effectively have several profiles collocated or being close in term of location. The formalism for the concept is well described. However, even if it is not quite the objective of this piece of work, the results of this concept need to be confronted to those of the classic retrieval method and more importantly to be validated by independent observations.
New variables for the retrieval products are proposed and the advantages of these variables with respect to those generally used so far are analyzed on a theoretical basis. Validation of these variables by means of application to real cases requires the involvement of data provider and data user communities and should be a follow up of this paper. I agree with the reviewer on the importance of the validation of the method and indeed the conclusions of the paper invite the communities of data providers and data users to test and validate the efficiency of this new interface.
Specific comments
See in the attached pdf file
Answers in the attached pdf file.
-
AC1: 'Reply on RC1', Simone Ceccherini, 04 Jul 2023
-
RC2: 'Comment on amt-2023-42', Anonymous Referee #1, 12 Jun 2023
It is unclear to me what this paper adds. It appears that the paper is a continuation of a discussion in the paper by Ridolfi et al (2022) which is more complete than the current document.
The terminology is confusing: the author is using a term data fusion. maybe i do not understand this but if i look at how the Ridolfi et al is defining data fusion it looks more like a data combining - averaging than fusion. I would understand data fusion more as a combining two different data sets from totally different sources (e.g. model and observations) than combining two different observational datasets. When combining observations with model simulations, reference is made to data assimilation, which is a different topic then discussed here.
Here the author discusses a method to combine two different observational datasets.
The abstracts suggests something new. However, already for decades there are instruments mounted on the same satellite platform, making observations in different spectral regions of the same atmospheric properties: vertical profiles of T, q, O3, etc. For iinstance IASI-MHS or HIRS and AMSU etc. So there is already a wealth of real data where new ideas can be tested. We do not have to wait for future missions like FORUM or IASI-NG, before the value of a new approach can be demonstrated.
The use of averaging kernels to mitigate the difference in vertical resolution and to mitigate issues with the mean state of the background has been reported. Rodgers has published several papers on this. And this is thus an accepted procedure to compare retrievals from different instruments. And for sure can used to derive new products which enables an more comprehensive way to combine products from different sources as discussed by the author.
But i believe there are some more fundamental problems then vertical resolution when combining similar products derived from semi simulateous observations by different instruments. Each instrument has different characteristics, viewing geometry, spatial sampling, Point Spread Functions etc. How these problems affect the final product is not addressed in the paper.
The author should demonstrate the approach he is advocating on real data.
My read from the paper is that the data producers should not only distribute the products, but also averaging kernell information. This is only possible if the product is generated using the methods based on OE.
If observations are not nearly coincidence in space and time, some sort of assumptions about the time/space evolution of the derived products needs to be made. It can be expected that the more complete the assumption the better the combined product will be. In that case Data Assimilation is around the corner and we would not need the new products proposed by the author as level 2 data assimilation introduces different problems, in particular correlated errors. Level 2 data assimilation is possible, but requires a different set of products.
In summary: to bring the proposed methodology to the attention of the target user community, the author should spend time to apply the method to some practical examples and then re-submit.
Citation: https://doi.org/10.5194/amt-2023-42-RC2 - AC2: 'Reply on RC2', Simone Ceccherini, 04 Jul 2023
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 510 | 139 | 48 | 697 | 43 | 42 |
- HTML: 510
- PDF: 139
- XML: 48
- Total: 697
- BibTeX: 43
- EndNote: 42
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1