the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
06 Jan 2021
06 Jan 2021
Deciphering Organization of GOES–16 Green Cumulus, through the EOF Lens
- Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot, Israel
- Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot, Israel
Abstract. A subset of continental shallow convective Cumulus (Cu) cloud fields were shown to have unique spatial properties and to form mostly over forests and vegetated areas, thus referred to as green Cu. Green Cu fields are known to form organized mesoscale patterns, yet the underlying mechanisms as well as the time variability of these patterns are still lacking understanding. Here, we characterize the organization of green Cu in space and time, by using data driven organization metrics, and by applying an Empirical Orthogonal Function (EOF) analysis to a high-resolution GOES–16 dataset. We extract, quantify and reveal modes of organization present in a green Cu field, during the course of a day. The EOF decomposition is able to show the field's key organization features such as cloud streets, and it also delineates the less visible ones, as the propagation of gravity waves (GW), and the emergence of a highly organized grid on a spatial scale of hundreds of kilometers, over a time period that scales with the field's lifetime. Using cloud fields that were reconstructed from different subgroups of modes, we quantify the cloud street's wavelength and aspect ratio, as well as the GW dominant period.
- Preprint
(35582 KB) -
Supplement
(24208 KB) - BibTeX
- EndNote
Tom Dror et al.
Status: closed
-
RC1: 'Comment on acp-2020-1242', Anonymous Referee #1, 25 Mar 2021
The article "Deciphering Organization of GOES–16 Green Cumulus (Cu), through the EOF Lens" by Tom Dror et al. applies an Empirical Orthogonal Function (EOF) analysis to so-called green cumulus clouds, i.e., cumulus clouds observed over continental forested or vegetated areas, observed through the GOES-16 satellite dataset. The article is basically a proof of concept, that a carefully-done EOF analysis can reveal a number of well-documented patterns in such 2+1 dimensional data. EOF analysis is a widely-used tool in climate and atmospheric science, but, as the authors point out, is typically applied to much larger scales and usually much more poorly-resolved (both in space and time) data. The data at hand, the GOES-16 datasets, is special in that its spatial (.5 km) and temporal (5 minutes) resolution are comparably high.
The EOF analysis applied is fairly standard, but carefully carried out: the data (channel 2, 0.64 micrometers, which reflects visible solar radiation) is used for daytime detection of Cu. The data are first converted to reflectance using a referenced technique and are processed by a gamma correction (I suppose, to be able to better view the data with the bare eye?), and re-projected onto a lat-lon grid. They then focus on several subregions of the data within Eastern North America (including a section of the Appalachian Mountains) during a 2018 summer period.
The EOF analysis is standard, in that the temporal fluctuations are first computed and then the first 20 leading eigenvalues and corresponding eigenvectors of the anomaly matrix are computed. The authors perform standard statistical tests to assess the sampling error. The first 20 EOFs are found to account for roughly 2/3 of the variance.
Already without discussing the EOFs, preliminary analysis reveals a clear diurnal cycle, which is not surprising given the continental surface. The temporal mean of the reflectance field already shows cloud streets, which are fairly stationary in time. Focusing on two sub-regions, the authors further discuss the presence of a gravity wave, which appears to modify the local cloud size and structure. To assess the organization of the cloud field they compute the Iorg index, an index that essentially assesses the departure of the centroid-centroid distance distribution from random (Poisson distribution). The authors, perhaps as expected, find more regular (lattice-like) organization than random for both sub-regions.
They then turn to the EOF decomposition and use the EOFs, to further explain the pattern already discussed qualitatively. EOF1 is associated with cloud streets and the corresponding eigenvalue has a time dependence, where a peak is found around noon, hence an indication that cloud streets build up during the day under diurnal heating, but dissipate at night. The second principal component is mentioned to also show a diurnal cycle, but is later discussed as stemming from orographic clouds. The third principal compenent is considered to be associated with cirrus clouds. They then proceed to discuss the time dependence of all principal components analyzed (1-20) and find that, given the error bars, several of the EOFs should be seen as degenerate. They point out that the higher-frequency modes (EOF10-EOF20) are able to capture the period of individual clouds. They finally show that the actual field can be well reconstructed from the first 20 modes they analyzed.
The authors point out that the study is essentially a proof of concept and that it allows decomposition of the cloud field into modes of different frequencies and explicit detection of cloud streets, orographic clouds and gravity wave effects.
-
AC1: 'Reply on RC1', Ilan Koren, 08 Jul 2021
The comment was uploaded in the form of a supplement: https://acp.copernicus.org/preprints/acp-2020-1242/acp-2020-1242-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Ilan Koren, 08 Jul 2021
-
RC2: 'Comment on acp-2020-1242', Anonymous Referee #2, 04 Jul 2021
Shallow Cu account for much of the uncertainty associated with cloud feedback. Unlike trade Cu over subtropical oceans, the continental shallow Cu are less well studied in the literature. This study, by using high-resolution GOES-16 dataset, explored modes of organization of shallow Cu over forests and vegetated areas, also referred to as green cumulus by the authors. The identification, and the influence of GW on the organized patterns of Cu are also investigated in the paper.
While I generally find the manuscript suitable for publication in ACP, further improvements are needed before the manuscript is accepted. Below I have included a list of specific comments that I think should be addressed.
Comments:
Q1: L45, are there any reasons why LES fail in reproducing shallow Cu, as LES can explicitly resolve shallow Cu at a common resolution of 100m.
Q2: L100, the identification of cloud objects, and the definition of cloud fraction and number of clouds are not clearly stated in the paper. I suggest the authors put more words on this part. In Figure 3c, it seems that the cloud fraction and the number of clouds is inconsistent.
Q3: L164, what’s the criteria in judging an EOF mode degenerate or nondegenerate?
Q4: L197, why PC5 and PC6 instead of other PCs can be used to identify GW. Are there any objective standards?
Q5: L255, Grammar check “such that the lower the frequency is the larger is the spatial scale that dominates the corresponding EOF”.
-
AC2: 'Reply on RC2', Ilan Koren, 08 Jul 2021
The comment was uploaded in the form of a supplement: https://acp.copernicus.org/preprints/acp-2020-1242/acp-2020-1242-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Ilan Koren, 08 Jul 2021
Status: closed
-
RC1: 'Comment on acp-2020-1242', Anonymous Referee #1, 25 Mar 2021
The article "Deciphering Organization of GOES–16 Green Cumulus (Cu), through the EOF Lens" by Tom Dror et al. applies an Empirical Orthogonal Function (EOF) analysis to so-called green cumulus clouds, i.e., cumulus clouds observed over continental forested or vegetated areas, observed through the GOES-16 satellite dataset. The article is basically a proof of concept, that a carefully-done EOF analysis can reveal a number of well-documented patterns in such 2+1 dimensional data. EOF analysis is a widely-used tool in climate and atmospheric science, but, as the authors point out, is typically applied to much larger scales and usually much more poorly-resolved (both in space and time) data. The data at hand, the GOES-16 datasets, is special in that its spatial (.5 km) and temporal (5 minutes) resolution are comparably high.
The EOF analysis applied is fairly standard, but carefully carried out: the data (channel 2, 0.64 micrometers, which reflects visible solar radiation) is used for daytime detection of Cu. The data are first converted to reflectance using a referenced technique and are processed by a gamma correction (I suppose, to be able to better view the data with the bare eye?), and re-projected onto a lat-lon grid. They then focus on several subregions of the data within Eastern North America (including a section of the Appalachian Mountains) during a 2018 summer period.
The EOF analysis is standard, in that the temporal fluctuations are first computed and then the first 20 leading eigenvalues and corresponding eigenvectors of the anomaly matrix are computed. The authors perform standard statistical tests to assess the sampling error. The first 20 EOFs are found to account for roughly 2/3 of the variance.
Already without discussing the EOFs, preliminary analysis reveals a clear diurnal cycle, which is not surprising given the continental surface. The temporal mean of the reflectance field already shows cloud streets, which are fairly stationary in time. Focusing on two sub-regions, the authors further discuss the presence of a gravity wave, which appears to modify the local cloud size and structure. To assess the organization of the cloud field they compute the Iorg index, an index that essentially assesses the departure of the centroid-centroid distance distribution from random (Poisson distribution). The authors, perhaps as expected, find more regular (lattice-like) organization than random for both sub-regions.
They then turn to the EOF decomposition and use the EOFs, to further explain the pattern already discussed qualitatively. EOF1 is associated with cloud streets and the corresponding eigenvalue has a time dependence, where a peak is found around noon, hence an indication that cloud streets build up during the day under diurnal heating, but dissipate at night. The second principal component is mentioned to also show a diurnal cycle, but is later discussed as stemming from orographic clouds. The third principal compenent is considered to be associated with cirrus clouds. They then proceed to discuss the time dependence of all principal components analyzed (1-20) and find that, given the error bars, several of the EOFs should be seen as degenerate. They point out that the higher-frequency modes (EOF10-EOF20) are able to capture the period of individual clouds. They finally show that the actual field can be well reconstructed from the first 20 modes they analyzed.
The authors point out that the study is essentially a proof of concept and that it allows decomposition of the cloud field into modes of different frequencies and explicit detection of cloud streets, orographic clouds and gravity wave effects.
-
AC1: 'Reply on RC1', Ilan Koren, 08 Jul 2021
The comment was uploaded in the form of a supplement: https://acp.copernicus.org/preprints/acp-2020-1242/acp-2020-1242-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Ilan Koren, 08 Jul 2021
-
RC2: 'Comment on acp-2020-1242', Anonymous Referee #2, 04 Jul 2021
Shallow Cu account for much of the uncertainty associated with cloud feedback. Unlike trade Cu over subtropical oceans, the continental shallow Cu are less well studied in the literature. This study, by using high-resolution GOES-16 dataset, explored modes of organization of shallow Cu over forests and vegetated areas, also referred to as green cumulus by the authors. The identification, and the influence of GW on the organized patterns of Cu are also investigated in the paper.
While I generally find the manuscript suitable for publication in ACP, further improvements are needed before the manuscript is accepted. Below I have included a list of specific comments that I think should be addressed.
Comments:
Q1: L45, are there any reasons why LES fail in reproducing shallow Cu, as LES can explicitly resolve shallow Cu at a common resolution of 100m.
Q2: L100, the identification of cloud objects, and the definition of cloud fraction and number of clouds are not clearly stated in the paper. I suggest the authors put more words on this part. In Figure 3c, it seems that the cloud fraction and the number of clouds is inconsistent.
Q3: L164, what’s the criteria in judging an EOF mode degenerate or nondegenerate?
Q4: L197, why PC5 and PC6 instead of other PCs can be used to identify GW. Are there any objective standards?
Q5: L255, Grammar check “such that the lower the frequency is the larger is the spatial scale that dominates the corresponding EOF”.
-
AC2: 'Reply on RC2', Ilan Koren, 08 Jul 2021
The comment was uploaded in the form of a supplement: https://acp.copernicus.org/preprints/acp-2020-1242/acp-2020-1242-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Ilan Koren, 08 Jul 2021
Tom Dror et al.
Video supplement
Cloud field evolution – video Tom Dror, Mickael D. Chekroun, Orit Altaratz, and Ilan Koren https://doi.org/10.34933/wis.000007
Tom Dror et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 220 | 77 | 11 | 308 | 27 | 2 | 4 |
- HTML: 220
- PDF: 77
- XML: 11
- Total: 308
- Supplement: 27
- BibTeX: 2
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1