the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Oct 2024
Effectiveness of Cirrus Detection with MODIS Cloud Mask data
Abstract. All clouds influence the Earth's radiative budget, with their net radiative forcing being negative. However, high-level clouds warrant special attention due to their atmospheric warming effects. A comprehensive characterization of cirrus requires information on cloud coverage, obtainable from various data types. Active satellite sensors are presently the most accurate source for cirrus data, but their usefulness in climatological studies is limited. On the contrary, passive data, available for the past 40 years with sufficient temporal resolution for climatological research, were not specifically designed for cirrus detection. In this study, we assessed the utility of MODIS standard products for creating a cirrus mask by validating them against CALIOP data. Our objective was to determine if a MODIS product exists that detects cirrus with the same accuracy as CALIOP.
Using CALIOP data as the reference, we evaluated six tests for cirrus detection considered in MODIS cloud masking algorithm and their combination (ALL TESTS CONSOLIDATION, ATC). Additionally we applied two ISCCP-originating tests: ISCCP3.6 and ISCCP23 tests. All tests have been applied to MODIS radiances.
Study revealed that ATC test was the most effective resulting with the overall accuracy of 72.98 % during daytime and 59.50 % at night (probability of detection: 80.87 % and 25.46 %, false alarm rate of 34.86 % and 6.90 %, and Cohen's kapppa coefficient of 0.46 and 0.19 respectively). However, its effectiveness was notably reduced during nighttime compared to daytime. We conclude that the test is suitable for creating a mask of high-level clouds.
- Preprint
(1534 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on amt-2024-163', Anonymous Referee #1, 06 Nov 2024
reply
Referee Comment
Effectiveness of Cirrus Detection with MODIS Cloud Mask Data
by Nguyen Huu et al.
This paper aims to evaluate cirrus detections from MODIS by using data from the much more sensitive CALIOP lidar as ground-truth. It specifically tries to quantify how well cirrus can be detected with different spectral tests, and combinations of tests, when applied to MODIS data. The comparisons stratified by the various spectral tests are an interesting and useful breakdown of the various passive sensor cloud detection methods, including the ISCCP techniques. However, the paper is rather poorly written and difficult to follow because it isn’t clear how ‘cirrus’ is defined in this study or how consistent that definition is for the two data products being compared. This leads to some difficulty understanding the methods and interpreting the results (specifically the analysis with respect to the number of cloud layers). While the study has modest scientific merit, it seems to fail to address fundamental questions that arise when interpreting the results. For example, it does not attempt to clearly explain possible reasons that could lead to disagreements between the characterizations from the two sensors, such as quantifying (or at least remarking on, based on previous published works) the MODIS sensitivities as a function of the opacity of the cirrus or due to the presence of lower-level clouds. Relative to CALIPSO, the study does not address how well MODIS data can be used to describe the spatial variability and patterns of cirrus cloud cover, or to track regional changes throughout the course of the year of study. This would seem to be a simple and insightful aspect to add to the study. A major concern involves the contention that the MODIS cirrus detection performs poorly at night, compared to daytime. The contention is based on their findings that indicate CALIOP detects nearly twice as many cirrus clouds globally at night than during the daytime. The day/night differences for MODIS are not discussed and difficult to discern. There is no discussion as to the validity of this diurnal pattern as observed in the CALIOP data, no mention of potential day/night CALIOP sensitivity differences (which are known to be significant), and no discussion on how this influences their MODIS evaluations and conclusions. Overall, the paper could be published but it requires major revisions to address these flaws and improve its significance. In addition, the manuscript has too many grammatical errors should be professionally edited or at least heavily edited by a native English speaker.
Specific suggestions:
Suggest changing ‘Detection’ to ‘Identification’ in the title
The Figure 2 caption should indicate that (a) is daytime, and (b) is nighttime
Fig 3 and lines 258-259: The CALIPSO data presented here indicate that there are over 50% more cirrus clouds at night than during the day which is a remarkable diurnal cycle that has not previously been reported in the literature. If it has, please provide citations that indicate that this level of difference is reasonable. Is it possible that this difference results from the increased sensitivity of CALIOP to thin clouds at night compared to daytime?
Fig 4: The reader can’t easily distinguish ATC from ISCCP 3.6. Please adjust the line types accordingly.
Authors seem to be mistaking the results to indicate diurnal differences as being the fault of MODIS when in fact the differences shown in the CALIPSO data may be unrealistic and result from the day/night dependency of the CALOP sensitivity (CALIOP more sensitive at night).
Line 7: change ‘that detects cirrus’ to ‘that enables identification of cirrus’
Line 11: The study revealed that the ATC test…
Line 17: replace ‘All of them’ with ‘They’, and ‘radiative’ with ‘radiation’
Line 17: remove the word ‘for’ in ‘forcing for is’
Line 18: Replace “that means that…’ with ‘Thus, their overall impact are to cool the planet’
Line 29: 35.5 is a specific value that means something specific, not in ‘general’. Remove ‘general’ and replace with whatever meaning is implied in the citation (globally averaged?).
Line 66: regarding the statement “..to operate day and night with similar efficiency”, can you support this with evidence or citations? The lidar sensitivity is not the same during day and night, which could influence how you interpret the results in your study.
Line 74: change to “with temporal coverage adequate for climatological research”
Line 74: ‘not designed for cirrus detection’ is incorrect as the imager designs have matured over time to increase the likelihood for detecting cirrus. MODIS has a 'cirrus' channel! The imagers certainly are designed specifically for detecting clouds but their sensitivity to optically thin clouds depends on many factors. Perhaps you mean to say something about the varying capabilities of the imagers over 4 decades…
Line 75: Suggest restating your objective. It is already well known that passive sensors are not as sensitive to cirrus as active sensors. I suggest the following starting on line 74: “In this paper, we use cirrus characterizations from CALIOP data to explore the potential for creating a cirrus mask from the operational MODIS cloud data products. Our objective is to determine how well the MODIS products can be used to identify cirrus clouds compared to CALIPSO.” In addition, the readers would greatly benefit from a more thorough description of how 'cirrus' is defined for the two datasets being compared and how these definitions are consistent or inconsistent. Do these definitions lead to a fair comparison? Does the fact that CALIOP attenuates at low COT or the fact that the products are vertically resolved lead to any confusion with your comparisons with MODIS?
Line 80: active ‘sensor’ data
Line 81: The active sensor data was obtained from the CALIOP lidar…
Line 82: collocation ‘of’ those
Line 92: change what to which
Line 103: What are ‘middle’ thresholds? You should clarify this.
Line 120: The MODIS central wavelength is closer to 3.7 than 3.9 um and usually referred to as the 3.7 um channel
Line 166: It would help the reader if you could describe what the CALIPSO ‘cirrus’ subtype represents.
Lines 255-265: The data presented here indicate that according to CALIOP, cirrus coverage is nearly twice as large during nighttime than during daytime, yet no explanation for this phenomenon is given and no evidence if this is realistic. Please explain the reasons for this, whether this is a data artifact or not, and discuss the implications for your study. Also missing from this section, or elsewhere in the paper is a day/night evaluation of the magnitudes of the MODIS cirrus cloud coverage and a comparison between the two sensors with respect to the geographic patterns and their correlation. Such an analysis would also seem to be important for testing your hypothesis that MODIS can provide useful information on cirrus clouds. Also, it seems that it would be straightforward for you to examine how well MODIS tracks changes in cirrus cloud coverage during the course of 2015. This could be done in several different ways (seasonal monthly mean maps and/or difference maps, global and select regional monthly mean time series, etc.).
Line 268-269: regarding CALIOP ‘all cloud’ COT values near 4.2, considering that CALIOP is fully attenuated at higher values, are these numbers scientifically meaningful or somehow meaningful to your paper? If so, explain how, and if not, consider eliminating the sentence.
Line 274: Can you clarify what you mean by ‘precluded the use of the test’? Precluded the use of the test where?
Line 290: Please clarify what is meant by ‘respective radiation range’ and why its variation can be attributed to variations in cirrus detection statistics.
Line 325-331: This section is impossible to understand which points to a persistent problem trying to interpret the results in the paper related to a poor description of the experimental setup with respect to the definition of ‘cirrus’ as defined for the datasets obtained from the two sensors, and how these definitions differ or have been rectified to provide consistent information.
Citation: https://doi.org/10.5194/amt-2024-163-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 120 | 29 | 52 | 201 | 2 | 2 |
- HTML: 120
- PDF: 29
- XML: 52
- Total: 201
- BibTeX: 2
- EndNote: 2
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1