|This is a well-written manuscript that presents a comparison of retrospective versus NRT CCI soil moisture products generated during 2013. From an applications point of view, this paper provides a valuable assessment of the decline in quality expected in NRT soil moisture products (versus a retrospective benchmark). However, because of the weakness described below, the paper still lacks a clear methodological contribution. A number of these issues were raised in my original review and, unfortunately, were not adequately addressed in this round of revisions. |
Therefore, I’d (still) recommend revision focusing on the following points:
1) The paper still offers no clear description of what specific methodological differences (e.g. calibration issues, screening difference and algorithm parameterizations) exist between the retrospective and NRT CCI soil moisture data sets and how these differences may potentially lead to retrospective versus NRT differences in SM products. Without a clear understanding of these source differences, it’s difficult for the reader to gain any real insight from the SM comparison results presented in the paper. Towards the beginning of the paper the author’s need to summarize these differences and present (at least) some cursory discussion of their expected impact. I understand that some of the differences might require a tedious level of detail/explanation…in these cases it would be fine to keep the discussion at a relatively high level.
2) While there is some attribution discussion in Section 5, it is presented in a cursory and unsatisfying manner. For instance, lines 381-382 say that “Since most of these regions are covered by AMSR2, the most likely error sources are the GLDAS-based rescaling parameters.”
There are two issues here. First, both AMSR2 and ASCAT soil moisture products are rescaled via by “GLDAS-based rescaling parameters”, so it’s unclear why the use of AMSR2 in these regions points to a re-scaling problem. Second, the “problems” being referred to are associated with poor temporal correlations. This is odd since correlations should be minimally impacted by rescaling (i.e., correlation is not impacted by any kind of linear scaling).
3) Along the same lines, in lines 392-394 it is unclear how a 2013 (retrospective versus NRT) bias can be attributed to a AMSR-E/AMSR2 cross-calibration issue given that AMSRE stopped functioning in 2010 and obviously played no direct role in the generation of any 2013 soil moisture product. I suspect that there is a subtle calibration/scaling issue at play here - whereby AMSRE does, in fact, end up impacting the calibration of the 2013 soil moisture results. However no explanation is given on exactly what this issue/connection is.
4) Same issue with line 396 – which mentions the potential benefits of a “dynamic snow map for ASCAT.” The sounds plausible but it’s also not clear how snow mapping errors may be affecting the observed retrospective versus NRT differences. Is snow masking applied differently in the two products? I couldn’t find any discussion on this issue. The same issue with the RFI masking mentioned later in the paragraph…how exactly does this issue lead to retrospective versus NRT differences (i.e. the core issue examined in the paper)?
5) In contrast, Lines 400-404 are very good…they clearly describe how ASCAT calibration issues may be driving retrospective versus NRT differences. I’d really like to see more of that in a revised version.
So, in summary, I’d still urge the authors to spend more time: 1) describing/listing the underlying sources of retrospective versus NRT soil moisture differences and 2) providing a better attribution discussion which clarifies the connection between these sources of differences with specific inter-comparison results presented in the paper (see above for specific advice on how to do this). I don’t think this would require a significant amount of re-writing…just a modest amount of additional text in Sections 1 and 5 or the revised manuscript.