Impacts of active satellite sensors' low-level cloud detection limitations on cloud radiative forcing in the Arctic

Liu, Yinghui

doi:https://doi.org/10.5194/acp-22-8151-2022

Articles | Volume 22, issue 12

https://doi.org/10.5194/acp-22-8151-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/acp-22-8151-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 22, issue 12

Research article

|

23 Jun 2022

Research article |

| 23 Jun 2022

Impacts of active satellite sensors' low-level cloud detection limitations on cloud radiative forcing in the Arctic

Yinghui Liu

Download

Final revised paper (published on 23 Jun 2022)
Supplement to the final revised paper
Preprint (discussion started on 23 Feb 2022)
Supplement to the preprint

Interactive discussion

Status: closed

RC1:
'Peer-Review Comment on acp-2022-140', Anonymous Referee #1, 17 Mar 2022

This is an interesting application of GV data to understand how satellite data cloud masks could lead to uncertainty in radiation forcing estimates. The author develops an application of the Quickbeam simulator to the high-temporal resolution ground validation. Uncertainties in the radiative fluxes due to the missed detection of clouds from simulated CloudSat and CALIPSO cloud masks were estimated. They found that the seasonal cycle has a large impact on the magnitude of CRF biases due to changes in cloud microphysics, temperature structure, and surface albedo characteristics. For monthly averages, the combined CloudSat-CALIPSO mask provides the best match to surface observations, however, errors can get large when comparing 1-minute case data. Uncertainties in the cloud detection methods are described in the final section. This work is useful for the community as uncertainty estimates in Arctic regions are needed; the work, however, has numerous issues that need to be tackled first before publication.

More significant comments:

1.

Results in Table 4, Table 5, Figure 10, and Figure 6 are difficult to interpret because of the sampling strategy presented. The author mentions that many cloud types are omitted from the study (snow, drizzle, liquid cloud+drizzle, rain, haze, or uncertain retrievals). It is important to understand how often these cases occur and how their exclusion impacts the results presented here. The CRF results are needed, however, if the sampling represents a small fraction of the total CRF they would not be representative of the total population of clouds. It would be helpful to understand the frequency of occurrence of each type mentioned here to get an idea of what is being sampled.

Further, radiative calculations are performed once per hour. Why not compute all profiles? One of the main advantages of ground validation data is the high temporal resolution and a once per hour sampling lowers the accuracy of the data. If a subset of clouds is being examined, clouds could be missed and random errors could become large. It would be helpful to demonstrate that increased sampling does not impact the results (i.e. 6 times per hour or 10 times per hour).

2.

As stated in the uncertainty section, the results presented are based on a single-shot cloud mask (on profile at a time). The combined CloudSat and CALIPSO cloud masking products as well as the cited 2b-FLXHR-lidar products use averaged data along the satellite track to better detect clouds that might be missing in a one-shot case. The cloud mask results derived in the current study would represent a worst-case scenario for cloud detection errors from the satellite perspective. As demonstrated in figures 20 and 21, a large portion of clouds could be detected if a small change is made in the retrieval. This uncertainty, however, is not translated back to the radiation. It is, therefore, difficult to relate how this uncertainty impacts results and makes it hard for the reader to interpret. It would be helpful to take a month of data, such as when errors are large, to demonstrate a range of CRF uncertainty due to changes in cloud mask thresholds.

3.

Something seems off about the optical depth calculations shown in Figure 1. Previous studies using SHEBA data display column optical depth ranges up to ~30, with most clouds having column optical depths < 10 (Turner 2005; Zuidema et al 2005). If the vertical resolution is 63 m., Figure 1 shows optical depths greater than 5 for each vertical bin in the 15-20 bins below cloud top. This would lead to a column optical depth > 100 and does not seem physical for such a geometrically thin cloud. Given that the optical depth is used in the radiative transfer calculations as well as the CALIPSO cloud mask it would also lead to large uncertainty in the results as it prevents the CALIPSO cloud mask detections. This needs to be investigated to make sure the correct calculations have been made throughout the study.

Turner, D. D. (2005). Arctic Mixed-Phase Cloud Properties from AERI Lidar Observations: Algorithm and Results from SHEBA, Journal of Applied Meteorology, 44(4), 427-444

Zuidema, P., Baker, B., Han, Y., Intrieri, J., Key, J., Lawson, P., Matrosov, S., Shupe, M., Stone, R., & Uttal, T. (2005). An Arctic Springtime Mixed-Phase Cloudy Boundary Layer Observed during SHEBA, Journal of the Atmospheric Sciences, 62(1), 160-176

5.

There are a large number of tables and figures in the paper. Figures or tables that are not necessary could be removed or added to supplemental. For example, the surface albedo figures are not needed as their source is described in the text. Further, the tables showing the raw numbers for vertical profiles could be included in the supplemental.

Specific Comments

Paper Title: The title is too vague and suggests a global study when the spatial sampling is limited to ground validation sites in the Arctic. I would include the Arctic in the title to make it more specific or mention the use of SHEBA data.

Pg 1 Line 22: Change “modulator of the radiation flux” to “modulator of radiation”.

Pg 4 Line3: I would make this clearer that phrases such as “with every 63 m from 150 to 1050 m” is talking about changes in vertical resolution.

Fig 1b: A log-scale color bar would make some of the finer details of the cloud optical depth pop out a little more and limit the saturated COD above 5 (see comment 3 above as well).

Pg 7 Line 5. I would move the description of the radiative terms being calculated after the description the two sets of profile experiments.

Pg 7: Line 18: The use of “cloud layers” or “layers” gets jumbled here. I would describe cloud layers as layers with cloud data and total layers (50 or 125) as just “layers”.

Pg 9: Line 21: Please introduce Figure 6 first.

Pg 9 Line 28: remove “very” from high optical depths.

Pg 11 Line 2: Should this reference Fig 6e?

Fig 8: Is “all” the surface observations? This should be made clear.

Pg 12 Line 12. With a difference of only a few percent, I would mention that the combined retrieval captures the majority of clouds above 1 km.

Fig 6. The Julian Day labeling is a bit confusing and would be easier to interpret if months could be added.

Pg 16 Line 10: This is mostly due to the lack of sunlight during many months leading to the LW heating being the dominant term.

Pg 28 Line 5: This caveat along with the one for CloudSat needs to be listed in the cloud mask description in section 2.

Citation: https://doi.org/10.5194/acp-2022-140-RC1
- AC1: 'Reply on RC1', Yinghui Liu, 29 Apr 2022
  
  The comment was uploaded in the form of a supplement: https://acp.copernicus.org/preprints/acp-2022-140/acp-2022-140-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/acp-2022-140-AC1
RC2:
'Comment on acp-2022-140', Anonymous Referee #2, 21 Mar 2022

This study examined the potential biases in the joint CloudSat/CALIPSO cloud mask and radiative fluxes retrievals. Due to the difficulties in distinguishing the cold surface from clouds by passive sensors in the polar regions, the joint CloudSat/CALIPSO observations are critical to study polar clouds and their radiative effects. Although the limitations of the CloudSat CPR and the CALIPSO cloud observations and retrievals are well known, this is the first study to systematically quantify the potential biases in the CloudSat/CALIPSO cloud cover and radiative forcings due to such limitations.

This study uses the ground-based MMCR retrievals during SHEBA as input to the QuickBeam radar simulator to simulate cloud profiles from the perspective of CloudSat CPR and the CALIPSO CALIOP. The effects of ground clutter of the CPR signals are carefully approximated. Adding the approximate ground clutter together with the simulated CALIOP integrated signal attenuation, the author identifies the missing clouds in the simulated profiles using a simplified version of the joint CloudSat and CALIPSO cloud mask. Indepth and detailed analysis of the contribution to the missing clouds and the cloud radiative forcing (CRF) from clouds of different phase are presented. The author finds monthly CRF uncertainties up to 2.7 W/m2 at the surface and 4.0 W/m2 at TOA and the uncertainties are up to 30 W/m2 in individual cases.

The findings of this research provide a more solid basis for the studies of the Arctic cloud and cloud radiative effects. This manuscript is well suited for publication at the Atmospheric Chemistry and Physics. At this point, it still requires careful revision before publication, especially for Section 3.2, mostly with clarification and improvements to presentation. Detailed comments are listed below.

General comments:

Maybe consider moving redundant information to the supplement materials, especially the long tables and Figures from previous publications. I do appreciate the author having all the data in the table for easy and quick reference but having them in the manuscript sometimes disrupts the flow of the presentation.

Specific comments:

P2, Line 23: “CALIPSO” is capitalized here. Please use consistent abbreviations throughout the manuscript.

P2, Line 23: “high-level clouds”, this sentence is ambiguous because the signal attenuation issues can happen for clouds at any level.

P3, Line 30: It should be Shupe et al. 2006.

P4, Line 6-7: Is there an estimate of how many profiles are excluded? Maybe adding a qualitative estimate of how such selection will affect the total CRF in later sections as well? Including such assessments in the manuscript will help the readers to interpret the results of this study more accurately.

P4, Line 16: It is mentioned on P7 that solid hexagonal column is used for ice particles in radiative transfer code. Is that consistent with the Quickbeam setting here? By default, Quickbeam uses ice spheres. If solid column is used here, are other parameters adjusted as well, such as for area/mass ratio?

P4, Line 18: Haynes et al. 2007 for the BAMS paper about Quickbeam.

P5, Line 12-13: Does this statement refer to subtracting the estimated mean clutter from the received power of the lowest four bins? Marchand et al. (2008) mentioned it as preliminary. Was this approach shown to be effective in later studies or used in later version of the CloudSat cloud mask processing?

P8, Line 8-11: These two sets are referred to frequently later in the manuscript. Maybe it is useful to have an abbreviation/code name for each of them? The efforts to describe them accurately make sentences long, convoluted, and confusing. Similarly, it might be worthwhile to find a more succinct way to address cases when clouds of a specific phase are missed by the combined CloudSat/CALIPSO mask.

P9, Line 8-11: Maybe specify the sign of the difference here as well? I can tell from the captions of Fig. 10 and 16 but it will be easier for the readers to follow to clarify.

Fig. 6 and 9: Maybe use thicker lines?

P16, Line 10: The time period of negative CRF from SHEBA is much shorter than the results by Kay and L’Ecuyer (2013). The SHEBA CRF is likely biased because most of its albedo observations were taken mostly locally from ice. It is true that they used a 200 m line to sample more surface types, but it was not likely going to represent the extent of open water over a larger area. The length of negative CRF may vary with latitude and the time span of surface melt/open water. So, the SHEBA CRF does show the typical seasonal cycle of the CRF in the Arctic Ocean but might not be an accurate representation of the Arctic Ocean in summer.

Fig. 10 and 16: Maybe change to two panels, one for CRF for all profiles, and one for the difference of CC from all? The differences are difficult to tell in a single plot.

P18, Line 6: How about adding the last sentence of caption in Fig. 11 here too? It helps to remind the reader that CRF differences within 2 W/m2 are not included in the following analysis.

P18, line 8-9: This sentence is confusing, please rephrase.

P19, Line 14: Does this mean the low-level clouds around 1 km is still opaque enough that the effective emitting temperature are close to when the clouds below 1 km are included?

P19, Line 15: Figure 11, S6?

P19, Line 15-17: this sentence is long and confusing. Please rephrase.

Fig. 13 15, and 18: These scatter plots are pretty noisy. Because the optical depth is related to the fraction of reduction of incoming radiative fluxes and not with the absolute values of the fluxes, normalizing the CRF differences with the solar zenith angle for SW and Ts^4 for LW may produce a cleaner fit.

P20, Line 1: I think the temperature of clouds below 1 km should be compared to the effective emitting temperature of the clouds above 1 km instead of to the surface temperature. The LW CRF being small in winter suggests the clouds above 1 km emit at a temperature cloud to the temperature within 1 km. Fig. S1 shows a liquid layer above but close to 1 km, which is consistent with LW CRF being small. The question is how representative this case is in winter.

P22, Line 9-10: This statement is not exactly accurate. The clouds are brighter than the surface in the summer mainly because the surface has more open water, more ponds, and less snow-covered ice rather than clouds being brighter.

P22, Line 12-13: Is it possible that the LW effective emitting temperature at TOA is much higher in the atmosphere that the BL inversion is not relevant at tall?

P23, Line 9-10: Maybe this is another sign that the LW emitting temperature is above 1 km?

P24, Line 7: “... determine the CRFs” to “... determine the CRF changes”.

P25, Line 11-12: Add comma between “Cloud especially ...” and add comma between “… near the surface help brighten …”.

P25, Line 12: Remove the second “larger”.

P25, Line 13: Remove “more”.

P25, Line 14: Remove “larger”.

Fig. 20, 21: use thicker lines and larger fonts.

P26, Line 22: Is there a reason for Quickbeam to perform better in ice clouds? Comparing Fig. S1 and Fig 19, the large reflectivity differences are in a slightly shallower layer than the retrieved liquid layer in Fig. S1. Maybe the larger differences in reflectivity of liquid clouds are related to the uncertainties in the height of phase transition? Using the refractive index of water for ice clouds will increase reflectivity significantly. If you set the places with large liquid reflectivity differences to ice phase for the case in Fig. S1, would it get rid of the large reflectivity differences? If so, it might point to phase retrieval error.

Citation: https://doi.org/10.5194/acp-2022-140-RC2
- AC2: 'Reply on RC2', Yinghui Liu, 29 Apr 2022
  
  The comment was uploaded in the form of a supplement: https://acp.copernicus.org/preprints/acp-2022-140/acp-2022-140-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/acp-2022-140-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Yinghui Liu on behalf of the Authors (29 Apr 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (04 May 2022) by Matthew Lebsock

RR by Anonymous Referee #1 (19 May 2022)

Suggestions for revision or reasons for rejection

I would like to thank the authors for a clear response to my questions. I think the new additions significantly improve the uncertainty discussion. Further, thank you for clarifying my misunderstanding of the optical depth. I think the revised figure that includes the integrated and raw values is useful. I only have a few minor clarifications that I would suggest before the final submission. Specifically, I am still a bit confused by response related to the omitted cloud types. My comments are below.

Page 11 line 1: Fix “the cloudsat”

Line 15 pg 20 :

“Of all the profiles in every month from October 1997 to September 1998, the profiles with snow, drizzle, liquid cloud+drizzle, rain, haze, or uncertain retrievals account for 11.6%, 17.0%, 7.3%, 9.0%, 7.3%, 10.5%, 9.1%, 4.9%, 10.1%, 22.3%, and 20.8%. Majority of all profiles have been used in deriving the results in this study.”

The percentages for these categories do not match. There are eleven numbers and only six categories mentioned. Further, the summation exceeds 100%. Please clarify this part and provide the proper respective categories. It would be helpful to include the percentage of the 10 categories and also add an all-inclusive statement, such as “XX% of all profiles were included in the CRF analysis and XX% were omitted due to uncertain retrievals”. Finally, I would provide these percentages on Page 4 line 5.

Page 17 Line 6. Add Fig 10 reference.

Fig 10 and 11: “Please note that cases with absolute differences less than 2 Wm-2 are the majority while excluded in the histogram in Figure 10”.

Could you include a percentage as “ the majority” is vague.

Section 3.3 “It should be noted that a more sophisticated detection scheme….”

You could combine the CloudSat and CALIPSO statements to reduce redundancy of this statement.

Hide

RR by Anonymous Referee #2 (20 May 2022)

ED: Publish subject to minor revisions (review by editor) (04 Jun 2022) by Matthew Lebsock

AR by Yinghui Liu on behalf of the Authors (05 Jun 2022) Author's response Author's tracked changes Manuscript

ED: Publish as is (07 Jun 2022) by Matthew Lebsock

Download

Article (7639 KB)
Full-text XML

Short summary

Cloud detection from state-of-art satellite radar and lidar misses low-level clouds. Using in situ observations, this study confirms this cloud detection limitation over the Arctic Ocean. Impacts of this detection limitation from combined satellite radar and lidar on the monthly mean radiation flux estimations at the surface and at the top of the atmosphere in the Arctic are limited but larger from only satellite radar or satellite lidar in monthly mean and instantaneous values.