Retrieval of Cloud Liquid Water Using Microwave Signals from LEO Satellites: A Feasibility Study through Simulations
Round 1
Reviewer 1 Report
This paper proposes the use of LEO satellite to surface receiver links to retrieve the cloud liquid water content (LWC). Simulation studies based on model clouds and a tomographic retrieval algorithm indicate the potential of this approach. Although this paper can give some new insights for the global cloud studies, it is recommended that the authors reconsider the following points.
(Major)
- L50-52 “The approach of using the estimated signal-to-noise ratio (SNR) at the ground receivers for LEO satellites to achieve three-dimensional (3-D) tomographic rain field reconstruction was analyzed in detail in [6].”: Since the present paper is an extension of the rain-retrieval paper, the authors should explain the similarity and difference between the rain and LWC retrievals in more detail.
- L86-91 “The overpass time is determined by the minimum elevation angle of the receiver and the orbit height. The satellite has a direct spot beam to the ground receivers. The SNRs at the ground receivers are estimated for the purpose of measuring the path-integrated attenuation. For a single overpass, the signal link between a satellite and a receiver forms a curved surface, thereby the attenuation caused by water droplets on this surface can be detected by the SNR estimation at the receiver.” From this part, it is not clear how the 3D tomographic measurement of cloud LWC can be achieved. What is the intended surface coverage of the “direct spot beam,” and how many receivers are assumed within the beam spot along the satellite track? Is it assumed that the steering capability of each receiver is used for the attenuation measurement? It would be necessary that these points are schematically illustrated in Fig. 1.
3. L317 4.4 Partial Retrieval: Here it is assumed that a limited number of receivers are located in a limited portion of the region to be analyzed. I believe that a more important question is what happens if the spacing of receivers is 5 km, for example, instead of 1 km as originally assumed. If the tracking of the satellite beam can be made successfully, will it be possible to retrieve cloud LWC even for the case of more sparsely deployed receiver groups?
- L356 “nonprecipitating clouds”: What happens if rain events occur with thick clouds? The simultaneous retrieval is desirable from the practical point of view.
(Minor)
L19 “with the root-mean-square error around 0.2 dB/km” It would be better to add information on the range of liquid water content assumed for the simulated clouds, such as 0 -2 dB/km.
L30 “cloud liquid water (LWC)” should be “liquid water content (LWC).”
L45-46 “was first proposed in [3]”, L48 “In [5]”, L52 “in [6]”: these are not in line with the conventional way for citing references. Many similar citations are found in this manuscript. Please consider more natural ways to cite references.
L47 “the rotational motion of the LEO satellites makes ...” should be “the orbital motions of the LEO satellites make ...”
L49 “ the received signal level (RSL) measurements”: the abbreviation “RSL” is not used thereafter.
L108 In eq. (1) please reconsider the suffix attached to each attenuation factor. The suffix R and C are confusing for indicating the cloud and water vapor effects.
L109 “in which is the power ...” should be “in which P_r is the power ...”
L114 It is not clear why a receiver can obtain multiple samples (the k-th sample). A more understandable description is needed.
L126 In the integrals, d\theta should be d\Omega (solid angle).
L164 At the top of Eq. (21), V of V(k) is missing.
L169 F_n(k) is assumed to be zero. From Eq. (8), this implies that T_sky << T_sys. Is this condition satisfied in the usual case?
L173 Please explain how the values of parameters k and \Delta are chosen in the present simulation.
L192 Please explain how the value of \beta is assumed (or derived) in the present simulation.
L201 The unit K should be capital.
L203 The assumption of a constant lapse rate (-6.5 K/km) does not hold in the cloudy atmosphere.
L205 In Eq. (33), probably R is missing at the top.
L209 Fig. 3(a): Please explain the range of LWC as well as cloud type assumed in the present simulation.
L213 Here, it is assumed that the group of phased-array antennas is provided with a 1 km mesh, which seems to be too unrealistic due to the cost issues.
L226 The parameter T_m is not defined in the text. Why does “a sine curve” appear here?
L228 The acronym BPSK should be explained.
L255 “Because the high directivity ...” should be “Because of the high directivity ...”
L257 In relation to Eq. (34), is it possible to illustrate the temporal changes of A_R and T_sky for a single antenna as the satellite passes along the orbit? That would contribute to making the observation situation more understandable.
L264, L295 Please explain how the constants in Eq. (36) (2.8, 1.9, ..., 1.5) and Eq. (39) (3.0 and 2.1) are determined.
L281-282 From the current result, which of Fig. 3(c) and (d) is considered to be the best fit? Will it be acceptable just to judge based on RMSE?
L304 In Fig. 4(b) and (c) the field retrieved with Gaussian basis functions are not sharp enough. Does this mean that the number of tiles employed (160x25) is not sufficient?
L333 In Fig. 5(c), a vertical line is seen at around x = 46 km. What is the cause of this line?
L351 “the unknown sky noise”: Is it true that after the iteration, the value of sky noise is determined for each receiver for each observation direction?
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This short paper is well written. The idea is innovative and fits the scope of the journal. I recommend minor revision before acceptance. The comments are listed as follows.
- Line 127. θ is solid angle not dθ.
- Equation (19). V(k) is defined here but what the meaning?
- Equation (21). “(k)” should be “V(k)” at the very beginning.
- Line 165-170. It is very clear in the equation derivation to Equation (22), which is the final equation to solve using least square method. However, the sequence of your derivation follows mathematical order but is not logical enough. Till Equation (29), w can be numerically solved, but from the title of the paper, it aims at retrieval of liquid water properties using the Q, which is unknow from the observation, but known in the simulation. The authors should add a descriptive paragraph narrating the logic. In the real-world retrieval of liquid water, both Q and w are unknow. In this study, the authors want to quantify the w in their model framework, where the Q is known. Then go back to the real-world, with the numerically solved w, Q can be calculated to retrieve the cloud liquid water. My description is not accurate which just conveys the logic. Otherwise, readers without modeling experience can be easily disoriented. More importantly, by adding this description could pave the way for the authors’ subsequent papers.
- Line 274, other good indices like correlation, agreement index, etc. could be considered.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
General thoughts:
This work presents a novel and interesting approach for deducing vertical profiles of cloud water using a network of ground based receivers and microwave signals from orbiting commercial satellites.
The main concern I have with the results of the simulation is the lack of attenuation in the mid and low altitudes of the troposphere where it highest cloud liquid water would be expected. This concern is outlined in one of the comments listed below.
Another concern is that the technique requires a large number of ground receivers with phased array antennae arranged in arcs under the orbital tracks of the satellites. The cost and constraints in obtaining locations for stations are practical limitations to implementation of the proposed measurements.
I am unable to provide adequate feedback on the analytic details of the methodology presented at this time.
Below are suggests for improving the clarity of the text.
Lines 59-60. The statement is confusing. It implies that rain water is primarily closer to the ground than cloud water, and that rainwater is less than a few 100 meters above ground. These statements are not realistic.
In addition, the effects of ice particles (which make up a significant portion of high cloud cover and convective storms) should be considered.
Lines 86-87. This doesn't seem correct. The overpass time depends on the satellite orbital conditions, not by the receiver. Needs rewording.
Lines 90. The statement is unclear to me. Why would the signal link at any instant in time be curved instead of a straight line? Needs rewording.
Figure 1. I think it could be helpful to add a few receiving stations with signal link paths to the figure. As shown, the figure does add much to visualizing the problem.
Lines 99-110. Rain fields and clouds are mentioned here. Which are you retrieving and why?
Line 110. What is the Power of Noise? Specify: background (sky), instrument noise, or both.
Line 189: It would be useful to mention the typical duration of a satellite overpass for which the assumption of static attenuation is assumed.
What is the horizontal scale of the cloud where the cloud water profile is assumed to be uniform. (That would depend on cloud height and minimum viewing angle used in the retrieval). How does this scale compare to the cloud scale of 80 km? Is that size the minimum resolvable cloud in the NWP model?
Line 194: The assumption about ice crystals given here should be mentioned in the introduction so it is clear from the beginning that only cloud water is being addressed with the technique.
Fig.3. If I am interpreting the plots correctly, it appears that the maximum attenuation is at 7.5 km altitude or higher. This is the case in subsequent plots as well. Typically (at mid latitudes), there is mostly ice cloud (or mixed phase in strong updraft regions) at these altitudes since the temperature is in the -20-30 C range. If there are deep, precipitating clouds, one would expect significant cloud/rain water in the lower altitudes. Also, most clouds containing water or supercooled water are in the lower troposphere. The attenuation is zero below about 6km in the plots. This raises some concerns regarding the realism of the results. It would be help to know more about the cloud characteristics (cloud depth, cloud base, liquid and ice profiles) and the temperature profile in the NWP model.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have responded properly to most of my questions in the first review. Please consider the following points before preparing the publishable version.
L61 “The first objective of this study is to ...”: Please describe the second one more explicitly.
L64-67: “... the retrieval algorithm is extended and refined by introducing different basis functions to define the target attenuation field and by utilizing post-processing with a piecewise linear function instead of the iterative approach [6] to cope with the sky noise induced by rain or cloud LWC.” I wonder if this statement is describing the salient feature of the “piecewise linear function” since later (L324), it is stated that “To reduce the computing complexity of the retrieval process, we propose the following piecewise linear (PL) function to ...” (See also my question below regarding L387).
L127 “... as that in [6].”: This is not a recommended way of citing a reference.
L178 “Once the basis functions have been determined and the locations of the signal samples obtained, ??,? can be calculated offline and stored away for later use.”: Does this imply that the values of these parameters can be obtained solely from the geometrical conditions? If so, a clearer description is desirable.
L285 “After some experiments, parameter Δ for the differential approach in Eq. (21) is set at 3, which seems to produce the best retrieval outcomes.” Please explain why Δ =3 is better than Δ =1 or 2.
L319 Figure 4: It is good to show figures that exemplify the outcomes of the present analysis. Looking at these figures, however, I am still wondering why the curves of A_I (attenuation due to LWC) look like these. I suppose for both panels (a) (a receiver at 46 km) and (b) (at 66 km), the data at sample number 200 represents the data when the satellite position is at the zenith. The change of sample number 0 to 400, then, “covers the entire course of the overpass” (L315). If so, the number of peaks and their strengths of the A_I curves do not seem to match what is expected from the change of the observation direction from each of the specified receivers. Moreover, the assumption that a single receiver can cover the entire overpass seems inconsistent with the -3 db footprint depicted in Fig. 1. Therefore, I think the authors should explain how the downlink microwave is transmitted and detected in a more understandable manner.
L387 Figure 6: Comparison between panel (a) and (b) indicates that for the retrieval of specific attenuations, it is obvious that the method for (a) (from the third and fourth iterations) leads to much better accuracy as compared with (b) (using the PL function). If so, what exactly is the advantage of introducing the PL function approach in the first place?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
