Evaluation of VIIRS, GOCI, and MODIS Collection 6 AOD retrievals against  ground sunphotometer observations  over East Asia

Xiao, Q.; Zhang, H.; Choi, M.; Li, S.; Kondragunta, S.; Kim, J.; Holben, B.; Levy, R. C.; Liu, Y.

doi:https://doi.org/10.5194/acp-16-1255-2016

Articles | Volume 16, issue 3

https://doi.org/10.5194/acp-16-1255-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

Special issue:

Meso-scale aerosol processes, comparison and validation studies...

https://doi.org/10.5194/acp-16-1255-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 16, issue 3

Research article

|

03 Feb 2016

Research article |

| 03 Feb 2016

Evaluation of VIIRS, GOCI, and MODIS Collection 6 AOD retrievals against ground sunphotometer observations over East Asia

Q. Xiao, H. Zhang, M. Choi, S. Li, S. Kondragunta, J. Kim, B. Holben, R. C. Levy, and Y. Liu

Download

Final revised paper (published on 03 Feb 2016)
Supplement to the final revised paper
Preprint (discussion started on 03 Aug 2015)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

AC C5747: 'Correction of author affiliation', Qingyang Xiao, 07 Aug 2015
RC C6437: 'Review', Anonymous Referee #1, 31 Aug 2015
- AC C6712: 'Response to two major comments', Qingyang Xiao, 09 Sep 2015
RC C6795: 'Commnets', Anonymous Referee #2, 12 Sep 2015

Peer-review completion

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

AR by Qingyang Xiao on behalf of the Authors (20 Oct 2015) Author's response Manuscript

ED: Reconsider after major revisions (01 Nov 2015) by Yinon Rudich

ED: Referee Nomination & Report Request started (09 Nov 2015) by Yinon Rudich

RR by Anonymous Referee #1 (03 Dec 2015)

Suggestions for revision or reasons for rejection

Main review points
==================

The authors have addressed many of my original concerns as well as
points raised by reviewer 2. However, the following major items remain
to be resolved:

1. The motivation of the study is still unclear:

- Why would the spacing of ground-based measurements make a
difference for satellite procuct evalution, in which each
individual retrieval is compared to only one ground-based
measurement at a time?

- While particulate matter, including PM2.5, will certainly
contribute to atmospheric aerosol loading, there is no way to
directly infer PM from AOD. Also, smaller particles in
particular tend to travel far, reducing the utility of
high-resolution aerosol retrievals for PM2.5 in particular. I
agree that monitoring air pollution is important, but I do not
see how particulate matter distribution would *directly*
motivate this study.

2. If this is not a communication problem, reprojection and data loss
thus incurred does not seem necessary and is very likely detrimental
to the reliability of the evaluation.

3. I still have strong concerns regarding the statistical analysis of
data sets not normally distributed. If this cannot be resolved, at
least it needs to be explicitly discussed, and all results need to
be carefully appraised in this light.

Detailed comments
=================

- 4-8, 4-14 and 23-29: I do not see from the manuscript how PM2.5
characterization at the ground would be helped by having a
high-resolution AOD product available. AOD is a very different
parameter and represents a vertical integral rather than information
pertaining to ground level only. There may of course be links
between particulate matter (of whatever size fraction) and AOD, but
this link is complex and not self-evident.

- 6-22: In your response to one of the comments in the previous review
round you argue that you require spatially concentrated ground-based
observations, and proceed to calling this “intensive” observations.
I have two queries: 1) I don’t think “intensive” is any clearer than
the previously used term “high resolution”. I would suggest
“spatially concentrated”, “closely spaced” or something similar, or
possibly a description instead of an adjective. 2) I am not
convinced that the evaluation of a 3km product requires closely
spaced observations. Just like a 10km, the performance of the
algorithm can be compared to ground-based observations. For the
quality of the individual retrieval, any potential spatial
heterogeneity hardly seems to be relevant. Or is there a particular
reason to suspect that the adjoining 3km pixel would have reduced
quality? Since this continues to be the central motivation of the
study, I would suggest some further explanations.

- 9-18 (was: 16-27 in previous round). Yes, I understand that
reprojection is needed to create a grid.

1. But why do you need a NEW grid to compare gridded satellite data
to non-gridded ground-based data? The quality of the satellite
data set will be reduced.

2. In your second response you claim that reprojection does not
include averaging of any kind. Every reprojection includes
changes to the original data, I would think. Can you please
elaborate?

Reviewer 2 also pointed out a possible lack of clarity regarding the
’remapping’ strategy employed. While you have addressed this point
in principle, I am still not sure I fully understand the procedure
from the manuscript.

- (was 19-28): I agree that there is little physical meaning in slopes
and intercepts in log-transformed data. In the same way, however, if
the condition required for the numerical method (i.e. normal
distribution) is not fulfilled the physical meaning of your results
will be equally limited. It does not matter in the least whether
other studies in the past made the same mistake, as you argue. In
your response you imply that the actual distribution of the values
is closer to a normal distribution than the log-transformed
distribution (although this cannot be seen from Figure 4, where
values on the horizontal axis are unevenly spaced). Can you support
this with a test for normal distribution? Are there other
transformations that would get us closer to a normal distribution?
In any case, I think this problem needs to be explicitly addressed
in the paper. Also, and accordingly, all discussion based on the
results of the numerical analysis need to be treated and interpreted
with great care on this basis.

- Figure 5: Are the regression lines shown here statistically
significant at a certain level? If so, please state this in the
figure caption. Otherwise I suggest removing the regression lines.
By the way: The first and third rows clearly show how the AOD
distribution is dominated by small values – there does not seem to
be a normal distribution, so linear regression will not yield
reliable results here.

- Figure 6: Reviewer 2 noted that color scale limits vary among
figures, impairing comparability. I agree with reviewer 2 and think
the sclaes should use identical limits.

Technical comments
==================

- 6-26: ...which temporarily deployed additional sun photometers in...

- 8-18: You changed the wording from “measure” to “observe” - still,
the station itself does not measure or observe anything. The
measurement occurs AT the station. Please increase precision in
wording here.

- 29-2: marking both slope AND intercept values with significance
flags seems redundant

- Figure 5: labels on horizontal axes are inconsistent – this may lead
to confusion on whether there is a difference between AOD and
AOD550. Suggestion: decide on one and use consistently

Hide

ED: Reconsider after major revisions (05 Dec 2015) by Yinon Rudich

AR by Qingyang Xiao on behalf of the Authors (13 Jan 2016)

ED: Publish as is (14 Jan 2016) by Yinon Rudich

AR by Qingyang Xiao on behalf of the Authors (19 Jan 2016) Author's response Manuscript

Download

Article (5091 KB)
Full-text XML

Short summary

Using ground AOD measurements from AERONET, DRAGON-Asia Campaign, and handheld sunphotometers, we evaluated emerging aerosol products from VIIRS, GOCI, and Terra and Aqua MODIS (Collection 6) in East Asia in 2012–2013. We found that satellite aerosol products performed better in tracking the day-to-day variability than the high-resolution spatial variability. VIIRS EDR and GOCI products provided the most accurate AOD retrievals, while VIIRS IP and MODIS C6 3 km products had positive biases.

Evaluation of VIIRS, GOCI, and MODIS Collection 6 AOD retrievals against ground sunphotometer observations over East Asia

Download

Interactive discussion

Peer-review completion

Evaluation of VIIRS, GOCI, and MODIS Collection 6 AOD retrievals against ground sunphotometer observations over East Asia