Interactive comment on “ Climatology of free tropospheric humidity : extension into the SEVIRI era , evaluation and exemplary analysis ” by M .

has been shortened. Line 19, “The RHFT product is compared to computed relative humidity, obtained based on Analysed...” We changed this sentence into: “The FTH estimated from the Meteosat observations is compared to the FTH computed from . . .”. Lines 21-22, Suggestion “and normalized root mean square deviation (NRMSD)...are obtained. The NRMSD confirms......”. Introducing NRMSD will help when it is presented in the figures. e.g. Figure 1. Otherwise you have to write "relative NRMSD" in the figure. As answered earlier, we do not want to introduce this new abbreviation and prefer to use “relative RMSD” as it is already the case in figure 6. We inserted “relative” also in the abstract. We do not understand the reference to Figure 1 since no normalization is performed in this Figure. Line 26, Suggestion: "RH10FT". The latter abbreviation: "RH10" is in normal size and "FT" in subscript .

Major comments In the beginning of the introduction a motivation for the present study is presented, thus, stating the important role of water vapour in the Earth's energy balance and for the water cycle. In the present study, merely the relative humidity has been analyzed. It is not obvious how the present results of changes in relative humidity can be link to this important greenhouse gas. A change in RH due to the global warming could be due to either a change in water vapour or temperature, depending on the region of interest. In the manuscript, at many places, it is not clear that it is actually purely RH that has been investigated here (e.g. the title of the manuscript).
The reviewer is right in saying that FTH is a function of temperature and mixing ratio. This is mentioned in the abstract (p9694, ll 11+12.). We made this more clearly in the updated version by changing the following: We added "FTH is the mean relative humidity (RH) in a broad layer in the free troposphere." in the abstract. We also think that the understanding of the manuscript can be improved by removing the first two sentences in the introduction. We further slightly adapted the first sentence of the third sentence: "The importance of relative humidity (RH) in the free troposphere originates from the non-linear interaction between humidity RH and longwave radiation." Also, the impact of temperature and mixing ratios on RH and FTH was further discussed on p9607, ll22. This paragraph became now the second paragraph in the introduction.
The language is sometimes somewhat confusing and need to be improved, language is more clear in some chapters and less clear in others. Some suggestions are presented in "Specific comments" below, however, the full text needs a English proof-check.
The manuscript has been proof-checked by a scientific editor and many changes have C4960 been implemented.
Minor comments 1. Page 9614, paragraph 2 of Section 4. I wonder how accurate it is to use ERA reanalysis data here for the cloud screening, due to the poor horizontal resolution and uncertainties in estimated cloud fraction. Since at least the SEVIRI perform observations at visible wave lengths these data could be used for cloud screening.
The ERA-Interim cloud mask is used only to build the training set for the determination of the fitting coefficients of the retrieval and not for the generation of the FTH product. This way we have a consistency between the thermodynamic profiles and the cloud information. The ISCCP-DX cloud mask is used to screen out high and mid-level cloud cases in Meteosat observations for the generation of the Meteosat FTH data record. Using the visible channels on SEVIRI is not an option because SEVIRI observations do not cover the considered period nor is the visible information available during night.
2. I do not understand equation (1). It seems that the data before and after the break should be corrected with the same factor, but then the factor "a_before/a_after" is applied only on "b_after". Is it that the latter factor should be removed?
We added that the BTs are modified "after the breakpoint".
3. The treatment of the abbreviations is confusing. Since so many have been introduced please include a list, so it is easier for the readers to find what they stand for. Even so, if a no "important" abbreviation is used only few times after it has been introduced, e.g. ML and PL, it is not necessary to introduce it at all. If an abbreviation has been introduced use it consequently, e.g. "RH" instead of "relative humidity", in the remaining text. Define an abbreviation at the first place where it has been introduced, e.g. FTH and BT in the introduction, as well in the abstract. Note that the abstract is separated from the main text.
We agree and now include a list of abbreviations in the Appendix. Abbreviations are introduced at first occurrence and are used consequently.

C4961
Suggestion: the "FTH" should be "FTRH", but probably better with "RHFT". The latter abbreviation: "RH" should be in normal size and "FT" in subscript (also below).
Indeed the term "upper tropospheric humidity"/"free tropospheric humidity" and their abbreviations UTH/FTH have been used for more than 20 years in the scientific community, be it in conferences or workshops or in scientific publications. As our FTH product is a heritage product we do not want to change this. 4. A description of statistics used need to be included. For example, results of absolute and relative bias are obtained of the data analyzed. It is not clear how it is calculated, particularly the relative values.
We now provide a definition of relative bias, bias corrected RMSD and decadal stability in the Appendix. 5. End of page 9613, What is meant by "samples of the seasonal cycle with the 1st day of the months......" How representative is the obtained seasonal cycle based on a single day of the month?
We agree that the reader can misunderstand this point. The point is to include samples of various local times, seasons and years to cover the various temporal scales. This way, chances are enhanced to cover a large spectrum of different atmospheric conditions. We have thus rephrased the paragraph.
6. page 9618, lines 17 and 18, How does differences in absolute BTs minimize cloud contamination?
High level clouds strongly impact the observed BT at 6.3 microns. The observed signal will not be dominated by FTH but by the cloud. A high level cloud that has been classified as clear sky will therefore result in BT difference between observed BT (affected by a cloud) and simulated BT (not affected by a cloud) of typically more than 3 K. The main motivation to apply a threshold of 3 K is to minimise cloud detection uncertainties because it is our intention to characterise the quality of the FTH product and not of the C4962 cloud mask. 7. Page 9612, lines 4 -7, The two last temperature values, 4.5 and 0.8 K, seem not agree with the results in the figure.
We think that the definition of the "difference" was not clear. Thus we have rephrased the paragraph: " Figure 1 shows the deseasonalized anomaly of the original and the updated BT as well as their difference. The intensity of a breakpoint is the difference between the anomaly difference (black values) prior and after the breakpoint. The breakpoints in January 2001, in July 2006 and in May 2007 have the following intensities: 0.5 K, -4.5 K, and 0.8 K, respectively." 8. Page 9616, lines 3 and 4, Not completely convincing concerning the outliers. "Such outliers" could be marked in the figures." As suggested, we have changed the text to point explicitly at the outliers: "Exemplary outliers are observed in March 1992 (in the South Atlantic) and in April 1990 (over northeast Africa) and..." 9. Section 6.3. Figure 6a present relative bias with negative values. Please give the expression used to calculate the relative bias.
Following the comment #4, we now provide a definition of relative bias, bias corrected RMSD and decadal stability in the Appendix.
Line 9, What is meant by "The temporal correlation" ? If it is "R" then write "No correlation (R = -0.01) is found between N and relative bias." Done.
We do not wish to introduce this new abbreviation and prefer to say "relative RMSD". We thus inserted "relative" here and propagated the term throughout the manuscript.

C4963
Line 24. The unit "%RH" is confusing and since the word "absolute" is used it is actually clear what is meant. Suggestion: "The time series averages of absolute bias ("better with" differences ?) and RMSD (Fig.6, third panel) for the RHFT are -1.2% and 5.0%, respectively." The units have been defined on page 9615, l 19 and we consequently apply these definitions. Starting at line 24, we stop discussions of relative values (in %) and start the discussion of absolute values and thus think that it is helpful to emphasise this with the notation %RH to avoid confusion.
Lines 16 and 27, What is meant by "normalisation" ?, not explained in the text.
This is now explained in the new Appendix.
Lines 26 and 27, The explanation sounds realistic, but the increase in FTH is it real or bias? Maybe an explanation for what "normalisation" stands for will help me here.
Following the previous comment, the term "normalization" is now explained in the Appendix. Here we discuss the fact that the absolute bias and RMSD do not exhibit an increase, only the relative RMSD does. We normalize relative to the ARSA FTH values. Thus, ARSA might exhibit an increase in FTH which has not been explicitly analysed here -we focus on the FTH from Meteosat. In section 7.3 we discuss the temporal change in our FTH product.
10. Page 9625 and line 15, Could it be worth to investigate 1 month instead of 3 months period? For this, one particularly month could be chosen in a test.
Starting with Figure 4 the analysis consequently considers seasonal averages. In order to keep readability and in order to allow for a consistent interpretation we do not want to include Figures 8, 11 and 13 for a specific month. It is however true, that though the trends have been computed on basis of seasonal averages, intra-seasonal variability contributes to overall uncertainties. We included a corresponding statement at the beginning of section 7.
Specific comments Page 9604 Line 1, Suggestion: "A new free tropospheric relative humidity (RHFT) data..." As stated earlier we do not want to change the term "free tropospheric humidity (FTH)". Line 2, "with" instead of "of" Done.
Line 19, "The RHFT product is compared to computed relative humidity, obtained based on Analysed..." We changed this sentence into: "The FTH estimated from the Meteosat observations is compared to the FTH computed from . . .".
As answered earlier, we do not want to introduce this new abbreviation and prefer to use "relative RMSD" as it is already the case in figure 6.
We inserted "relative" also in the abstract. We do not understand the reference to Figure 1 since no normalization is performed in this Figure. Line 26, Suggestion: "RH10FT". The latter abbreviation: "RH10" is in normal size and "FT" in subscript .

C4965
The term FTHp10 has already been introduced in the literature (Roca et al., 2012). Therefore, we do not want to change it.
Page 9609 Line 8, Is it ok to write "Sects." instead of "Sections"? This was introduced by ACP.
Lines 15-18, Only radiance data are presented in this section. This part may be integrated into the last paragraph of Section 1, or has to be rephrased in some way.
We agree and have rephrased the first paragraph: "This section briefly describes the instruments and the radiance input data sets used to retrieve the FTH.". Lines 20-21, "Meteosat-7, which belong to the first generation of Meteosat satellites."
Line 26, "channels that cover the" Done.
Page 9610 Line 1, I do not understand why "while in operational mode" is written here.
It may happen that two satellites are available at nominal position (that is, over Africa). This is the case with Meteosat-8 and Meteosat-9. Then, one satellite needs to be slightly moved away from 0 • . Line 2, "present day." C4966 Done.
Has been removed.
Line 6, "Meteosat-2 -Meteosat-5". Suggestion: "Meteosat2 -Meteosat5", and then at all other places in the text. Note that "Meteosat" is in normal size and "2" and "5" could be in subscript Done and we keep the terminology introduced by EUMETSAT: the number of the satellite is thus kept in normal script. Line 7, "Cloud" Done. Line 8, "DX level"? Rephrase the content in the paragraph.
Now the abbreviation is explained at first occurrence.
This has been removed.
Lines 18 -20, A R2 value should be presented, and a suggestion: move"(not shown)" after "excellent linear behavior". Scatter plots of what?
The correlations are >0.99 which is now also mentioned in the manuscript. Changed into ". . . exhibit an excellent linear behavior with correlations >0.99 (not shown)." These are scatter plots between the BTs from various Meteosat platforms. We have included "simulated BTs".

Done.
Line 26, "The work by Picon" Done.
This term means that the BTs have been adapted to the Meteosat-5 spectral response function. As this has been mentioned in section 2, we drop "-like" here.
Line 12, "The magnitude of the observing period" ? Change to "are different, but" A suggested, the order has been changed.
Lines 10 and 11, "Equation (2) also correct for the satellite viewing angle ....." Changed into "Eq. (2) also corrects for the effect of the satellite viewing angle θ and includes...". in FTH at the Inter Tropical Convergence Zone (ITCZ) are evident during the boreal summer. Furthermore, Figure 4 illustrates the seasonal averages in FTH. The figure shows that the location, extent and strength of the humidified and dry areas highly depend on season." Changed into: " Figure 3 shows examples of instantaneous and monthly averaged products. Figure 4 illustrates the FTH seasonal averages featuring strong FTH minima over northern and southern Africa during boreal summer and strong FTH maxima in the Inter Tropical Convergence Zone (ITCZ). The location and the extent of dry and wet areas and the corresponding minimum and maximum FTH values strongly depend on the season." Page 9616 Line 1, "large" instead of "strong".
Lines 11 and 12, what is meant by "varies along the design of the algorithm." Changed into "depends on the details of the underlying algorithm." Further details are given in the next paragraph.
We mean "representativeness uncertainty" here.
We changed into: "Based on the tropical training a RMSD of 2% RH (8% when assuming an average FTH of 25%) and an average difference of 0.3% RH were estimated." Line 23, "20% yield a 10% relative"" ? The paragraph has been removed.
Line 14, It is not clear what is meant with "remaining". Suggestion: "night time only to avoid possible problem with radiosonde quality..." Change into "Night time only". Line 21, Suggestion: "contains radiosonde measurements from ships.." otherwise "...contains measurements from radiosondes launched on ships and at small islands..." The second suggestion has been implemented.

C4972
We have included more details in the appendix.
We changed into "Probability Density Function (PDF)".
Line 10, "more than 70%", Is this the cases really for the months September-May?
It means that more than 70% of the monthly means have valid FTHp10 values in the tropical South Atlantic, as defined in Figure 4. We dropped the term "all year through".
Line 21 and 22, This sentence is confusing.
We want to analyse variability on time scales ranging from interannual to decadal scales.
Page 9622 Line 28, R of +/-0.15 is not much of correlations. Suggestion: "of only around" Now we provide averages for positive and negative values.
Page 9624 Lines 1 and 2, "We tested two different methods for the analysis of linear trends: median of pair wise slopes regression (named "Theil-Sen slope estimator", Theil, 1950) and linear regression." Changed into: "Two methods to analyze linear trends are tested: the "median of pair wise slopes regression" method (named "Theil-Sen slope estimator", Theil, 1950) and the linear regression method." Line 11, remove "only" ? otherwise rewrite the sentence.

C4973
"only" has been removed. Figure 2, What is "RMS" and "RS" ? The first sentence is hard to understand, particularly when "and the "observed" FTH." at the end is included. Figure 2 has been updated and "...and the observed FTH" has been removed.
As mentioned earlier we will continue to speak of FTH. Figure 5. "Time series of regional RHFT averages" Change into "regional FTH averaged over. . .". We enhanced the font size of figures 6 and 8. We harmonised the labeling and consistently speak of bias/RMSD and relative bias/RMSD .