Optimal Estimation MSG-SEVIRI Clear-Sky Total Column Water Vapour Retrieval Using the Split Window Difference
Round 1
Reviewer 1 Report
The article presents a new algorithm for determining the total water vapor mass (TCWV) based on the results of measuring the Earth's radiation in the infrared range by means of the geostationary satellite MGS-SEVIRI. There are no comments on the theoretical part of the work. The algorithm is built according to the classical scheme for solving the inverse problem based on the radiation transfer equation (RTTOV model) using a priori error covariance matrices. A feature of the proposed method is that the brightness temperature of the surface is measured from the satellite in the split window difference at 11 μm and 12 μm, the spectral dependence of the absorption in water vapor is taken into account, and its contribution is distinguished depending on the altitude. Comparison of the obtained SEVIRI data with the AERNET, GNSS and MWR data has shown their good consistency at different times of the year for the territory of Germany. High absolute accuracy of TCWV measurement from 0.11 to 2.85 kg / m2 has been achieved.
Only minor remarks can be made on the article.
- A number of questions arise when comparing SEVIRI data with MWR. In the first case, the upward radiation of the atmosphere is measured with a spatial resolution of 3 km, and in the second, the downward radiation at a given point. At the same time, when it comes to high measurement accuracy of about 0.1 kg / m2, spatio-temporal fluctuations of TCWV caused by tropospheric turbulence can have a great effect. It would be desirable for the main characteristics and measurement methodology of the MWR device to be presented in the paper.
- It would be useful to test the effectiveness of the proposed TCWV retrieval algorithm in other climatic zones, for example, in Italy or Spain.
The article contains new results, and it can be recommended for publication in the Remote Sensing as presented.
Author Response
We would like to thank for the reviewer's comments. In the following, we will reply to the remarks. The line numbers given are in reference to the manuscript provided with the reviews. In the attachement, a revised version of the PDF with changes colored accordingly (red = removed, blue = inserted) can be found.
Point 1: A number of questions arise when comparing SEVIRI data with MWR. In the first case, the upward radiation of the atmosphere is measured with a spatial resolution of 3 km, and in the second, the downward radiation at a given point. At the same time, when it comes to high measurement accuracy of about 0.1 kg / m2, spatio-temporal fluctuations of TCWV caused by tropospheric turbulence can have a great effect. It would be desirable for the main characteristics and measurement methodology of the MWR device to be presented in the paper.
Reply 1: The measurement accuracy of MWR is between 0.5 and 0.8 kg/m2 as given in Steinke et al. (2015). The accuracy of MWR TCWV from MOLRAO (Lindenberg Observatory) is assumed to be in the same order. RMSD between SEVIRI TCWV and MWR reference is at 1.68 kg/m2. What are the 0.1 kg/m2 you the reviewer is refering to?
It is a very good comment to go into more detail for the reference datasets as in the case of MWR we compare a spatially very narrow observation of TCWV with a large field of TCWV. This aspect will be clarified in the revised version of the manuscript, changes were done in the description of MWR in section (2.7.3). In the validation a reduction of the area (3x3=10x30km2 vs 5x5=16.5x30km2) did not alter the statistics significantly for any of the validations, including the MWR.
In Steinke et al. (2015) within time intervals less than 30 minutes standard deviations higher than 0.5 kg/m2 were observed. Nonetheless, the time intervals used in this validation were half of that and limited to cloud-free conditions. Concerning the spatio-temporal fluctuations of TCWV, Carbajal Henken et al. (2020) conclude that sampling effects of spatio-temporal collocation in comparing satellite-retrieved TCWV with reference TCWV are low compared to the overall retrieval uncertainty. In Carbajal Henken et al. (2020), however, satellite TCWV was compared with GNSS-TCWV, not MWR.
The magnitude of fluctuations at this spatio-temporal scale are not negligible and will be discussed in more detail. A study to repeat the intercomparison done in Steinke et al. (2015) with the MSG SEVIRI TCWV would be insightful.
Point 2. It would be useful to test the effectiveness of the proposed TCWV retrieval algorithm in other climatic zones, for example, in Italy or Spain.
Reply 2: Valid point. Validation studies in other climatic zones are a good idea and indeed we are planning full disk retrieval and subsequent validation, this is mentioned in the manuscript, at l. 515.
References:
Steinke, S., Eikenberg, S., Löhnert, U., Dick, G., Klocke, D., Di Girolamo, P., and Crewell, S. (2015) Assessment of small-scale integrated water vapour variability during HOPE, Atmos. Chem. Phys., 15, 2675–2692, https://doi.org/10.5194/acp-15-2675-2015
Carbajal Henken, Cintia & Dirks, Lisa & Steinke, Sandra & Diedrich, Hannes & August, Thomas & Crewell, Susanne (2020): Assessment of Sampling Effects on Various Satellite-Derived Integrated Water Vapor Datasets Using GPS Measurements in Germany as Reference. Remote Sensing. 12. 1170. 10.3390/rs12071170.<
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Please see the attached pdf.
Comments for author File: Comments.pdf
Author Response
First of all, we want to thank the reviewer for the valuable remarks and questions. Our replies to the questions/comments are below. The mentioned line numbers refer to the original version of the manuscript provided with the first reviews. In the attachement, a revised version of the PDF with changes colored accordingly (red = removed, blue = inserted) can be found.
Point 1. How the proposed method is used for the meteorological applications?
Because precipitable water is a precursor of heavy rainfall, the weather is not always fine in that case. Even if precipitable water in a fine clear day can be precisely obtained by the method proposed in this study, how do the authors use the method for meteorological applications?
Reply 1: On the one hand satellite-retrieved total column water vapour (TCWV) is investigated to be assimilated into numerical weather predictions (NWP) (i.e. Saunders2020). On the other hand this exact question (utilisation of TCWV for improving weather nowcast) is an on-going topic of research in the RealPEP project (https://www2.meteo.uni-bonn.de/realpep, accessed: 2021-09-05). E.g. preceeding the formation of clouds, spatial variabilities and/or specific patterns in clear-sky water vapour fields can be observed (Carbajal Henken et al., 2015 (convective rolls)), Lindsey et al., 2018 (they use direct SWD)). In the future, this aspect is tried to be exploited in order to enable or improve predictions of cloud formation.
More specifically, within the RealPEP project, statistical relationships between observed spatio-temporal variabilties of TCWV and observations of clouds and precipitation at later time steps are studied in order to advance the monitoring and nowcasting of clear-sky and cloudy convective initiation (Troemel et al. 2021, see. https://www2.meteo.uni-bonn.de/realpep/doku.php?id=project_c1, accessed: 2021-09-05).
Beyond the use in weather related studies, the algorithm can be used to create long time series of globally retrieved TCWV for climate research.
The use of TCWV for convection initiation studies is briefly discussed in the introduction at l.102-106. However, it is a good point to also elaborate on this train of thought in the discussion.
Point 2. Why satellite data are necessary for estimating precipitable water in Germany?
Because three independent ground data are densely distributed in the ground for estimating precipitable water (Fig. 1), why satellite data are necessary in addition in Germany?
Reply 2: Valid question. To begin with, Fig. 1 shows all available stations where data are available for any given day in the validation. Not all stations provide TCWV 24/7, thus, gaps would appear if only available stations are shown for a specific day. Furthermore, the stations only provide point measurement fixed in space. The stations by themselves would not provide sufficient observations for the investigations mentioned in reply 1). Even if assimilated into NWP, the subsequent TCWV-analyses are not capable to capture the spatial variabilities a geo-stationary TCWV could.
Germany was chosen as a first test bed since funding and framework is provided by RealPEP. This means that within the project there are some case studies for specific extreme precipitation events over Germany for which we want to provide satellite-retrieved TCWV in the preceeding conditions in high spatiotemporal resolution. This could be made more clear in the manuscript, it is briefly mentioned in l.107.
A positive side-effect of Germany is that its dense network of ground-based TCWV observations can be utilised to assess SEVIRI-TCWV performance. Additionally, satellite products (and their spatiotemporal resolution) bring the advantage that they can be synergized or combined quite easily with other (future) satellite observations, e.g. cloud physical properties or soil moisture.
Eventually, this retrieval can and will be used for any region covered by MSG SEVIRI or similar instruments.
Point 3. What is the merit of this method in comparison with the traditional split-window method? In L.485-487, the authors stated that “A direct, quantitative comparison against the performance of other TCWV retrievals from split window observations is not possible, since their validation were either limited to a few case studies, focused on other study domains or used different TCWV references”. The reviewer cannot understand the reason at all. Why the authors do not carry out case studies in German in 2016-2017 by using traditional split-window method? Because both methods use the data of 10.8 and 12 micrometers, the reviewer thinks that the comparison with the traditional split window method is necessary to highlight the merit of the new method proposed in this study.
Reply 3: Good point. Certainly, a well-tuned traditional split-window retrieval of TCWV yields similar values. The goal of the paper is not to compare our algorithm against the other approaches but rather validate against independent references. However, there are some advantages to the combined method of optimal estimation and radiative transfer:
From the full optimal estimation (OE) we not only get estimates of the state but also estimates of the uncertainty, averaging kernels and cost functions which provide insights on the performance on a pixel by pixel basis. Additional sources of uncertainty can easily be integrated when identified and quantified. Furthermore, this framework can seamlessly be extended with additional measurements, i.e. the NIR or other TIR bands as it will be provided by MTG. In this case OE can combine the information from these two sources of information in a mathematically consistent way. OE also allows us to quantify further sensitivities (e.g. dust, inversions). With additional measurements it follows that more geophysical parameters can be estimated. One example could be a parameter describing the vertical profile of humidity which could be valuable to diagnose atmospheric stability.
Another reason for the chosen approach is that we want to develop a retrieval in the TIR that builds on the radiative transfer model (RTM) Radiative Transfer for TIROS Opverational Vertical Sounder (RTTOV). RTTOV is a robust RTM and is in operational use for a wide range of products retrieved from MSG SEVIRI and other EUMETSAT instruments.
A direct comparison between the various retrieval algorithms is a splendid idea. However, we lacked the time to both focus on algorithm development, validation and the processing of TCWV with one of the various split-window TCWV alogrithms.
Thus, the advantages will be clarified deeper in the introduction and the comparison of performance indicators from other split-window methods will be removed from the manuscript.
Point 4. L.125: What is rpm?
Reply 4: Revolutions per minute (rpm), the abbreviation will be added.
Point 5. P.11, Fig. 3: What is the scatter plot around “GNSS Germany (27-49 mm) vs SEVIRI (10-33 mm)”? The authors did not explain these outliers in the text.
Reply 5: This is actually addressed in the discussion at l.441-445, albeit with a mistake: it should read "higher", rather than "lower". However, we will add a remark in the caption, too.
Overall:
The misspelling and minor mistakes were corrected, especially concerning the references. However, some sources do not provide a DOI, in these cases we added a link to the online pdf.
References:
Saunders, R. (2020): Exploiting satellite total column water vapour, JCSDA-18th-Workshop, url: http://data.jcsda.org/Workshops/JCSDA-18th-Workshop/Thursday/7.Saunders.pptx (accessed: 3.9.2021)
Carbajal Henken, C. K., Diedrich, H., Preusker, R., and Fischer, J. (2015): MERIS full-resolution total column water vapor: Observing horizontal convective rolls, Geophys. Res. Lett., 42, 10,074– 10,081, doi:10.1002/2015GL066650
Lindsey, D. T., Bikos, D., & Grasso, L. (2018). Using the GOES-16 Split Window Difference to Detect a Boundary prior to Cloud Formation, Bulletin of the American Meteorological Society, 99(8), 1541-1544. Retrieved Sep 6, 2021, from https://journals.ametsoc.org/view/journals/bams/99/8/bams-d-17-0141.1.xml
Trömel, S., Chwala, C., Furusho-Percot, C., Henken, C. C., Polz, J., Potthast, R., Reinoso-Rondinel, R., & Simmer, C. (2021). Near-Realtime Quantitative Precipitation Estimation and Prediction (RealPEP), Bulletin of the American Meteorological Society, 102(8), E1591-E1596. Retrieved Sep 6, 2021, from https://journals.ametsoc.org/view/journals/bams/102/8/BAMS-D-21-0073.1.xml
Author Response File: Author Response.pdf
