GRWG/GDWG Web Meeting 2021-04-08

GSICS IR sub-group Web Meeting in lieu of Annual Meeting Session II - GEO-LEO Inter-Calibration

In lieu of 2021 GSICS annual meeting , the IR group will host two(2) web meetings to review all the presentations of the annual meeting. The second one will be held at 12:00-3:00 GMT on Thursday April 8 2021 (check you time here) and focused on GEO-LEO Inter-calibration. Should you have any urgent issues related to the IR group, please bring them for discussion.

Agenda

1. Tim Hewison(EUMETSAT) Introducing the Hot Land Bias issue

2. Chengli Qi (CMA) FY-3E/HIRAS-II pre-launch calibration and instrument performance

3. Arata Okuyama(JMA) Updates on inter-calibration results at 3.9 µm band

4. Minju Gu(KMA) AMI IR inter-calibration results and issues

5. Dorothee Coppens( EUMETSAT) Another look at the SEVIRI/IASI comparison in the 3.9 µm band

6. Likun Wang(NOAA Affiliate) NOAA ICVS/GSICS Recent Updates for GEO-LEO IR Inter-Calibration

7. Fangfang Yu(NOAA Affiliate) A review of GSICS GEO-LEO IR Inter-Calibration Algorithm

Attendees

Guest Chair: Likun Wang (NOAA Affiliate)

CMA: Chengli Qi , Yong Zhang

ESA: Stefano Casadio

ISRO: Pradeep Thapliyal

JMA: Kazuki Kodera, Arata Okuyama, Kazutaka Yamada

KMA: Minju Gu, Eunkyu Kim, Jiyoung Kim

UMD: Hui Xu

UW: David Tobin, Bob Knuteson

EUMETSAT: Dorothee Coppens, Tim Hewison, Ali Mousivand

NOAA: Fred (Xiangqian) Wu, Likun Wang, Fangfang Yu, Banghua Yan, Erin Lynch, Flavio Iturbide, Hyelim Yoo, Zhi-Peng Wang

NASA GFSC: Yonghong Li, Tiejun Chang

Universität Hamburg: Martin Jörg Burgdorf

Summary

Tim Hewison(EUMETSAT) Introducing the Hot Land Bias issue

Tim presented the hot land bias issue inter-calibration results. There are large biases over hot scenes in SLSTR-IASI inter-calibration results. These excess biases are visible not only over desert areas, but other land - e.g. Brazil. When checking GEO-LEO inter-calibration results by expanding the dynamic range, the larger bias is also identified over deserts. Preliminary analysis indicates that the biases could not be explained by the variations in NDVI/slope/aspect and elevation and further analysis is on-going. Finally, by introducing the comments from Tim, he left this topic for discussion and comments.

Comments (Dave Tobin): When I first looked at the slides that you presented. I think when you collocate two sensors with heterogonous scenes, the sensor with small footprint has tendency to look at hot scenes because they are not blurred much as the sensor with large footprint.

A: When we do the collocation, we do the spatial average for the smaller footprint pixels. So, each pair of observations should represent the same spatial area. But motion compensation has different blur effects on the sensors.

Comments (Likun Wang): I do not see it when comparing VIIRS and CrIS. However, it is on the same platform so there are no angle match issues. I suspect that it is caused by the collocation caused by the angle issue. What is the zenith angle difference used for collocation? The best way is to present a case study, which shows both the sounder and imager image side-by-side and then check the collocation.

A: The zenith angle ratio is 2%.

Comments (Ali): These biases are not always at heterogonous scenes. We also found it in the Libyan Desert and Brazilian forest. These sites are spectral invariant and supposed to be homogeneous.

Q: Did This bias always happen over land? IS this related to terrain correction issue during collocation?

A: This happened over land. However, we do not see bias only over mountain regions. The biases can be found at other places as well.

Comments (Fred Wu): First, this bias happen mostly at 3.9 µm band. When one averages radiance instead of brightness temperatures, this effect is smaller for heterogeneous scenes. Another question is, have you plotted the bias along the relative azimuth angle? If two instruments look the scenes from the opposite directions, this orographic effect probably is stronger than from the same direction. We see this pattern before especially for visible channels over mountain regions. One side is sun shined and the other side is shadowed. With time passes, you can see this pattern changes. This clearly indicates orographic effects on inter-calibration.

Chengli Qi (CMA) FY-3E/HIRAS-II pre-launch calibration and instrument performance

Chengli presented the pre-launch calibration and instrument performance of FY-3E/HIRAS-II. First, the instrument FOVs have been changed from 2x2 (16.0 km) to 3x3(14.0 km). Second, FY-3E/HIRAS will provide the whole continuous spectrum of 650~2550 cm-1. Based on the prelaunch test, the performance of FY-3E/HIRAS has great improvements in spectral calibration, noise, polarization, nonlinearity.

Q: Do you have the plots on the nonlinearity for different FOVs (detectors)?

A: I only show the plots after nonlinearity correction here.

Q: The FY3E will be put on the Early morning orbit. Generally speaking, what would you expect the instrument performance for the instruments on the early morning orbit?

A: We have challenges for Spectral calibration, polarization correction, and non-linearity correction. These calibration or correction coefficients needs to be further adjusted during the orbit testing?

Arata Okuyama(JMA) Updates on inter-calibration results at 3.9 µm band

Arata presented the updates on inter-calibration results at 3.9µm band to respond the recommendation item following the web meeting on Jan 13 2021. He compared the inter-calibration result between two different ways on how to handle the IASI negative radiance values. One way is to integrate IASI negative radiance as normal values, and the other way is to replace the IASI negative radiance values with positive values. It was found that the bias patterns are greatly changed. He also compared the ways on how to handle GEO negative radiance values. It was found the impacts is very small with and without including GEO negative radiance values for spatial average.

Comments and Questions (Fangfang): Slide 5: 1) what spectral regions do IASI negative radiance values appear? 2) For your plots, maybe it is nice to make the plots in a radiance domain.

A: For question one, on the IASI negative radiance value, please refer to Hui Xu’s talk on Jan 13 2021 and as well as the talk presented by Dorothee.

Comments (Fangfang): Slide 6, you see negative GEO radiance values at channel 7 because this channel effects more by bit depth issue for lower radiance value. The lower radiance value is close to the noise level at this channel, which is different from other channels.

Comments (Fangfang): Slide 7 How do you select the cold scenes in your plots? Suggest playing different standard to see how these plot changes?

A: I use standard deviation of radiances as threshold to select cold scenes.

Comments (Likun Wang): Should one use GEO negative radiance values for channel 7 for spatial average? The good news is there is no large difference here.

Comments (Fangfang): This is what I commented before. This channel is different, and we should use GEO negative radiance values at this band.

Minjun Gu (KMA) IR inter-calibration results and issues of AMI

Minjun discussed the AMI IR inter-calibration results and issues. She first reviewed the inter-calibration for IR bands between AMI and reference instruments (IASI-B and -C, NOAA20 and SNPP CrIS). She then presented the investigation of the AMI TB anomaly at Hot Land (Australian Desert) regions. It was found that after applying the threshold of azimuth angle ratio, the mean and standard deviation of the BT differences are reduced. Second, she presented the investigation of TB bias anomaly (AMI-IASI) at cold scene for SW038 band. The results found that ∆ TBAMI-IASI biases are converted from large negative biases to large positive biases at cold scene in TB domain after integrating negative radiance value. Finally, she presented the AMI and AHI GEO-GEO inter-comparison results.

Comments (Tim): It is very to see azimuth angle ratio can remove the bias at hot land? Do you have any idea why this works? Do you have any algorithms (e.g. orographic heating rate) to apply here?

Comments (Likun Wang): Based on the findings in slide 6, do we need to add the azimuth angle constraints for the IR algorithm. Apparently, right now, we only use the zenith angle constraints in current algorithms.

Comments (Fangfang Yu): Do we need collocation for daytime? Or we only need collocation during nighttime. To me, this issue is not caused by the instrument but caused the collocation itself.

Comments (Tim): We need to add the dynamic range in the future. This is going to be an issue that we must address. We want to add as many collocations to cover the dynamic range so we can characterize the instrument nonlinearity.

Comments (Likun Wang): On one side, we agree with Tim, we need to include many samplings to cover the dynamic range. On the other side, the inter-calibration procedure should not bring uncertainties into inter-calibration. Otherwise, this error will be attributed to instrument calibration errors. This is not the purpose of the inter-calibration.

Q (Likun Wang): For the GEO-GEO inter-calibration, the AHI and AMI have larger difference at Band 7 at the cold scene range. Would you like to explain it?

A: At this point, I do not know the root causes yet, but I will further investigate it.

Dorothee Coppens( EUMETSAT) Another look at the SEVIRI/IASI comparison in the 3.9 µm band

Dorothee presented the investigation the SEVIRI/IASI comparison in the 3.9µm band. She firstly showed the statistics of IASI negative radiances at the 3.9µm band. She found that the use of the operational IASI L1 PCS products drastically reduces the negative radiances in filtering out the noise but does not change the inter-calibration results.

Q (David Tobin): For the CrIS, we are now considering developing CrIS PCA data by following the IASI similar procedures. The PCA product not only reduces the instrument noise also data volume. My question is, does the IASI PCA dataset provide noise estimation? What is that, for example 10% of the instrument noise?

A: I cannot answer it right now. But I can find the answer after the meeting and reach the person who handles this. I would also like to mention the hybrid PC data is not operational yet.

Q (Fangfang): Just like to make sure the inter-calibration results between original IASI data and PC data are consistent.

A: Yes. They are consistent if you see the table. It is only slightly different. We try to make sure the PC IASI data do not bring any uncertainties into the inter-calibration.

Likun Wang (NOAA Affiliate) NOAA ICVS/GSICS Recent Updates for GEO-LEO IR Inter-Calibration

Likun Wang presented recent updates of GEO-LEO IR Inter-Calibration under the NOAA/ICVS GSICS portal. He tried to explore the possibility of evolving the GSICS generic IR inter-calibration algorithm by using it as an example. He presented the new collocation algorithm that carries out the collocation over space instead of on the Earth by taking advantage of the K-D tree structures. He also mentioned that the CrIS gap filling method worked well when comparing the results between CrIS and IASI for ABI Band 11. Finally, he updated the double difference between SNPP and N20 CrIS, IASI-B and IASI-C, and CrIS and IASI.

Q (slide 11): Yaw slip has effects on different channels. Is it only for IASI-B? How about other inter-calibration?

A: No. It is also be found for GOES-17 IASI-C data. You can check the results from NOAA GSICS ICVS portal website. It also surprised me that not all the channels are the same. It only impacts on channel 8, 11, and 12.

Comments (Fangfang): We know the yaw slip impacted on the calibration. It is interesting to see the bias patterns during day and night are changed.

Comments (Fangfang): We know that the day and night difference at Band 13 is caused by the heat pipe issue. But it is interesting to see it is consistent.

5. Fangfang Yu(NOAA Affiliate) A review of GSICS GEO-LEO IR Inter-Calibration Algorithm

Fangfang reviewed the GSICS GEO-LEO IR inter-calibration algorithm and proposed several potential issues for future improvements. Her talk discussed each step of the GSICS IR inter-calibration algorithm, including sub-setting, collocating, transforming, filtering. She finally compared three methods on uniformity check and outlier rejection, which is very important for the GSICS correction generation.

Comments (Likun): sub-setting of dataset is suggested to use the orbit predict model for accelerating the process.

Comments (Likun): Using a relatively large sounder field of view may not be necessary once the collocation is accurate.

Comments (Tim): There is a tradeoff when relaxing uniformity check. When reaching to the cold scenes, we do see some bias patterns for cold scenes (e.g. large offset), which may be physically reasonable though the instrument detectors may not response in that way. What I would like to suggest is that one should also check the change of the slope and intercept uncertainties.

Q: Are the slope that you showed here based on a linear or a weighted linear regression?

A: They are calculated from the linear regression because current GSICS algorithm ATBD suggests using the linear regression for correction coefficient generation.

Comments (Arata): JAM used a different threshold for uniformity check, which is the standard deviation of radiances. Fangfang’s presentation encouraged me to further investigate it.

Outcome

Action: ??

A.GIR.20210408.1 Action to all the team members to further investigate the bias at hot lands for LEO-LEO and GEO-LEO inter-calibration and report the results back.

A.GIR.20210408.2: Action to Dave Tobin to circulate his paper on how to process the collocation data. *

A.GIR.20210408.3: Action to Likun Wang to write a short paper or a technical note to document the findings on the investigation of 3.9 µm band inter-calibration bias as well as the best practice to handle the IASI negative radiance for inter-calibration.

A.GIR.20210408.4 Action to all the team members to further investigate how uniformity check and data quality control impacts on the bias estimation and further on GSICS correction. The team members should report the results back.

R.GIR.20210408.1: It is recommended that the GSICS IR inter-calibration algorithm ATBD should be revised and include a part on how to correctly handle IASI negative radiances.