GRWG/GDWG IR Sub Group Web Meeting 2021-01-13
GSICS Web Meeting on investigation of the bias at 3.9µm channel in IR sub-group
This web meeting is to investigate shortwave bias at 3.9µm channel when comparing IASI with geostationary imagers (especially for AMI and AHI). Basically, both the JMA and KMA teams independently found that there is a large bias at 3.9µm channel when comparing AMI/AHI with IASI. This bias however cannot be identified from the inter-calibration of GOES16-IASI inter-calibration . The related background information can be found in the presentations by Minjun Gu
(slide 6) and Fangfang Yu
1. Likun Wang (UMD & NOAA Affiliate): Introduction
2. Minju Gu (KMA): Status of AMI 3.8um anomaly analysis at cold scene using IASI
3. Arata Okuyam (JMA): AHI/B07 (3.9um) Tb bias for cold scene
4. Tim Hewison (EUMETSAT): Issues with use of IASI to inter-calibrate 3.9µm band (20 min)
5. Hui Xu (UMD): Investigations of IASI Negative value at SW channels: Inter-comparison with CrIS
6. Fangfang Yu (UMD & NOAA Affiliate): G16 ABI B07 cold scene bias to IASI
7. Likun Wang (UMD & NOAA Affiliate): wrap-up and acctions (5 min)
Guest Chair: Likun Wang (UMD & NOAA Affiliate)
CNES: Laura Le Barbier, Clemence Pierangelo
JMA: Hideaki Tanaka, Kazuki Kodera, Kazutaka Yamada, Arata Okuyama
KMA: Minju Gu, Won Hyeong, Hyelim Yoo
UMD: Hui Xu
UW: David Tobin
EUMETSAT: Dorothee Coppens, Jérôme Louis, Tim Hewison , Vincent Debaecher, Ali Mousivand, Alessandro Burini, Sebastien Wagner
NOAA: Manik Bali, Fred (Xiangqian) Wu, Likun Wang, Fangfang Yu
NASA Langley: David Doelling , Conor Haney, Rajendra Bhatt, Ben Scarino
NASA GSFC: Tiejun Chang
WMO: Heikki Pohjola
Universität Hamburg: Martin Jörg Burgdorf
Likun Wang (UMD & NOAA affiliate)
By introducing the web meeting, Likun Wang reviewed the challenges and current status of inter-calibration of 3.9 µm channels using the IASI because of low radiance values at SW spectral regions. He also pointed out that the inter-calibration results between IASI-AHI and IASI-AMI are consistent, which however are different from and IASI-ABI.
Comment (Tim Hewison): 1) IASI does not fully cover SEVIRI SW bands; 2) what’s the relationship between general IASI data quality flag and IASI negative radiance values.
Response (Dorothee Coppens, Laura Le Barbier, and Dave Tobin): The general quality flag (GQisFlagQual
) is linked to the general quality of the data (no underflow/overflow of the coding table + no spike detected + good radiometric and spectral calibration + good post-radiometric calibration to ensure that all the processing chain was applied correctly, and the data is good to be use). The general quality flag does not indicate there are no negative radiance values at the IASI spectra. More importantly, the negative radiance values are normal, which are caused by the instrument noise on low radiances often found in the SW spectral region. If one changes it or replaces it with positive values, it will introduce a false bias into the IASI spectra.
Comment (Fangfang Yu): Glad to see that the independent analysis for IASI-ABI agrees well with previous analysis.
The IASI SDR instrument team from EUMETSAT and CNES is recommended to draft a memo on how the IASI general data quality flag is generated and explain why IASI have negative radiance values at the SW region. Based on this, a best practice on how to handle IASI negative radiance values will be recommended to the team. (Circulated by email 2021-01-14
Minju Gu (KMA): The status of AMI 3.8µm anomaly analysis at cold scene using IASI
Minju presented the detailed analysis for AMI 3.8 µm bias at cold scenes using IASI. First, the location of these cold bias pixels was found at the top of deep convective clouds. The examination of the AMI dynamic range indicates that the AMI has an ability to detect the lowest cold temperature of 150K at this band, which is also clearly demonstrated by the lunar image. For these cold scenes, she found that IASI spectra do have negative radiance values.
Dave Tobin: In Slide 4: Are these AMI radiance values? All the data points stay together, and it is probably caused by quantization. How many bits are used for AMI quantization?
Response (Fangfang): Not sure for the AMI. But for the ABI (the same instrument as AMI), 14 bits is used to process the data.
Likun Wang: Based on the slide on lunar analysis, what is the dynamic range of AMI for this channel.
Response (Fred Wu): Ideally it should be 80K. But from the figure. It seems 150K (later corrected to 197K for ABI). Usually one bit corresponds to the lowest detected temperature.
Martin Jörg Burgdorf : it is possible to Use the Moon image for nonlinear response check.
Response (Tim) It may need an accurate lunar radiance model.
KMA is recommended to check how to handle the IASI radiance negative values before the convolution.
Arata Okuyama (JMA): AHI/B07 (3.9um) Tb bias for cold scene
Arata presented the study on AHI 3.9um band (B07) Tb bias for cold scenes. He indicated the bias appears in both of inter-calibration results using IASI-A and -B but does not show using AIRS and CrIS
. He also found that the uniformity criterion should be relaxed at this band in order to include enough cold scenes. With detailed analysis for these cold scenes, he found that there are negative radiance values and fluctuations on some IASI channels at colocation data in cold scenes. In JMA’s approach, the IASI negative radiance values are replaced to positive values computed by the gap filling method before the convolution.
Likun Wang: For CrIS
inter-calibration for 3.9µm channel, which GAP filling method used? NOAA Hui Xu’s PCA method or the JMA method.
Response: We use the JMA method.
Tim : With and without the gap filling, it seems that the bias range and patterns are both changed.
Response: Yes. When the gap filling is applied, the IASI based “super channel” Tb (x-axis) is changed. So, the bias range and patterns also changes.
Fangfang Yu: Comment the uniformity check. It agrees with her findings.
Hui Xu: For your LBL simulation output the constant values for cold scenes. It may be caused by model accuracy. I also got the similar values when using LBLRTM.
: JMA is recommended to present the inter-calibration results with and without gap filling to the team .
Tim Hewison (EUMETSAT): Issues with use of IASI to inter-calibrate 3.9µm band
Tim presented several issues with use of IASI to inter-calibration 3.9µm band. He suggested that 1) doing all processing of IASI data as radiances and only converting to BT at the last step for communication purpose; 2) including negative radiances for convolutions because they are caused by the noise and should be canceled out after band average. He proposed to optimize the current GSICS inter-calibration algorithms for very cold scenes, e.g. by characterising the bias in different BT slices to give more uniform coverage of the dynamic range, then fitting a regression through the results
Question for all: How to handle the IASI spectra with negative values.
Recommendation: Using IASI QA flags first, which indicate IASI spectral measurements and calibration has some issues. Should not change or fill negative radiance values using any methods. Otherwise, it will introduce the false bias into the IASI spectra.
Tiejun Chang: How to use IASI or CrIS
to find the DCC FOVs because they have relatively large FOV size.
Response: Using collocated imager pixels (e.g. IASI’s integrated imaging system, VIIRS for CrIS
). This is going to be an interesting topic for CrIS
and IASI DCC FOV identification.
: Tim is recommended to lead the study in the IR sub group to optimize the current GSICS inter-calibration algorithms for very cold scenes.
Hui Xu (UMD): Investigations of IASI Negative value at SW channels: Inter-comparison with CrIS
Hui Xu investigated IASI negative channel statistics by analyzing IASI data in different periods. He used the values BT at 900 cm-1 to classify the statistics of IASI SW negative radiance values. The spatial location of IASI spectra with SW negative radiance values are presented. By comparing with CrIS
, IASI agrees with CrIS
well in the dynamic range based on the mean spectra. In the end, he also showed before and after the gap filling IASI negative radiance values, the bias pattern is changed because the warm biases are introduced into the IASI spectra.
Dorothee Coppens and Laura Le Barbier: Surprised to see so many negative values in the IASI spectra.
Response (Hui Xu and Dave Tobin): 1) Use window channel 900 cm-1 as a classifier; 2) the most negative values are from the SW CO2 absorption regions and the radiance values are extremely low here.
Arata Okuyama: I will check before and after gap filling methods.
: Hui Xu is recommended to update the statistical figures on IASI negative values (separated with CO2 absorption and window regions) and report it back to the team.
Fangfang Yu (UMD & NOAA Affiliate): G16 ABI B07 cold scene bias to IASI
Fangfang Yu presented an updated study for G16 ABI B07 cold scene bias to IASI, which was presented in the 2019 GSICS annual meeting. Basically, she suggested that the uniformity check should be relaxed for 3.9µm band to include enough cold scene samples. Yet the resampling process at ABI L1b image process may introduce artificial bias at heterogeneous areas at the relaxed uniformity check. She proposed using the lunar image to check band 7 and 8 relation for diagnosis. No apparent non-linear bias can be found between ABI B7 and 8 at very cold scenes at the non-illuminated moon pixels.
Each agency provides the information related to the dynamic range of each band/channel.
- Meeting link:https://umd.webex.com/umd/j.php?MTID=mc2cbc68128b9914dcdb1ee58009819ab
- Meeting number:120 087 0292
- Host key: