2020-08-27

Agenda Item 1:

CRTM 2.4 Tasks (Johnson)

Discussion:

The goal is to have an intermediate release up to CRTM 3.0. A lot of the elements come from CRTM 2.3.1, which was never released.

(Ben is showing a whitepaper on his screen)

The key features are that we want to keep backwards compatibility and we add OpenMP as for many people a wall clock speedup would be beneficial but we don’t want to break things. Several things are already supported and I am coordinating with Kevin’s team to get some sensor coefficients they have.

The other element is NetCDF support. We are looking to have that at least for the aerosol coefficients, but hopefully also for the cloud coefficients. We will still keep the binary files in a transition period.

We also want to increase the maximum size in the cloud tables. This has already been done and needs to be implemented.

We also wanted to implement automated testing similar to JEDI. For new CRTM developments there need to be unit tests for each element.

The goal is also to have consistency with other JCSDA projects.

The other issues are just to increase the community outreach and support. We also have a specific CRTM forum in the JCSDA forum for general and technical support questions.

We are trying to coordinate this release with the JEDI release to have it unified.

We have this time table here (shown).

Any questions?

Isaac: When we generate coefficients we first have a NetCDF and then convert them to binary.

Ben: Even though these coefficients are not in the release we will have aerosol and cloud coefficients. We can also add additional information such as meta data and backscattering coefficients. With internal documentation in the NetCDF we also will know where the data came from, which is not the case with the binary files right now.

Yingtao: So only the aerosol coefficients will be in REL-2.4?

Ben: Patrick can talk about this.

Patrick: As Isaac mentioned, the transmittance coefficients are first created as NetCDF but reading them so far hasn’t been a requirement.

Yingtao: Are there any modifications in the CRTM reader part?

Cheng: Only the CRTM initialization routine needs some additional arguments. All the NetCDF I/O will be in a separate function.

I’m not sure what the concerns is.

Yingtao: The new NetCDF reader has groups. You need to modify it. I don’t know if that has already been done.

Cheng: The current task is to test aerosol properties but I should look into that.

Isaac: For the transmittance coefficients I have switched to NetCDF4 and I can give a file to Patrick.

The reason that I split those instruments is that I wanted to submit a lower number of channels.

Sarah: So, you actually wanted to convert to the new table?

Cheng: We won’t use other agencies’ table directly because some of them have information we don’t need, e.g. for NASA. For the NetCDF look up table you also need to rearrange the data.

Tom: So, there was a question during the last MOB meeting. Will all the features be selectable features and will we have backwards compatibility? Are there differences?

Ben: There will be differences due to bugfixes, such as the cloud fraction. Emily has submitted a bugfixes for ATMS snow cover. There was a bug in those channels when you pass bad data. These things will change and we will document all that. I wanted somebody to be able to download that from GitHub and test it without any changes. The other complication that hasn’t been tested is the availability of NetCDF4 on the system.

Cory: We have NetCDF4 and also OpenMP.

Ben: Minimum compiler versions are also always an issue. We haven’t tested on unique compilers. Daniel has tested it on a PGI compiler.

Daniel: Yes.

Result:

Tasks for Release 2.4 have been clarified.

Tasks:

Responsible People:

Ben, Cheng

Deadline:

October 2020

Agenda Item 2:

Cloud Fraction Testing Update (Johnson)

Discussion:

Ben: I just wanted to talk about the cloud fraction bug.

(shares screen)

It’s basically if you set cloud fraction to zero, there is some minimum cloud fraction based on guidance from RTTOV. What I found was that setting the cloud fraction to zero in IR channels doesn’t give you the same result as when setting the water content to zero, only when aerosols are present. So somewhere in the CRTM there is a bit of code that, when clouds are present, there is something done differently. You might think why would someone do that?

Communicating with RTTOV folks, specifically Alan Geer, they are running into the same error. Based on the conversation they will probably go into the same direction as we are. How frequent are small clouds in models? Hard to say, but we will implement a smooth transition to zero now. You are welcome to look at this issue on GitHub with a lengthy discussion.

Tom: Is this actually for small fractional content or for small water content?

Ben: There is also a water content threshold but that is not a concern. The issue appears only if you set the cloud fraction less than 10^-6.

Tom: A common trick that is used for the simulation of all sky radiances. If you have no clouds your Jacobians are zero. So, the trick is to set cloud values to a minimal value so that you have some sensitivity.

Ben: This change wouldn’t brake that ability. CRTM would still do the cloudy computations and this change will allow you to go to very small cloud fractions that might be unphysical.

Hui Christopherson: What about the OMS GEMS-1 transmittance coefficients?

Patrick: The issue is still in the inbox. 

Result:

-       The cloud fraction bug has been identified

-       A solution for the cloud fraction bug was agreed upon.

-       The cloud fraction bug solution will probably also be implemented by the RTTOV team.

-       The work on OMS GEMS-1 transmittance coefficients hasn’t been started yet.

Tasks:

-       Implement cloud fraction bugfix

-       Look into OMS GEMS-1 transmittance coefficients

Responsible People:

Ben, Patrick

Deadline:

October 2020

Agenda Item 3:

Aerosol Update

Discussion:

Cheng: I am cleaning up my code and will push my code to Github. Maybe you and Patrick can review the code maybe this or next week.

Ben: Are you able to create any branches on Github, Yingtao?

Yingtao: Yes, I will create a new branch soon.

Result:

Tasks:

Responsible People:

Cheng Dang

Deadline:

October 2020

Agenda Item 3:

Transmittance Coefficient Package Update

Discussion:

Patrick: I will go through the list: So far, I have been working on the IASI-NG coefficients and I have obtained the IRF from EUMETSAT. Furthermore, I have been working on the VIIRS comparison with Yingtao and I have asked him to provide his Transmittance_production.defaults file.

Haixia: Are you computing a channel subset for IASI-NG?

Ben: We will generate the same subset as for IASI.

Result:

-       Work on IASI-NG calculations has started.

Tasks:

-       Finish IASI-NG calculations

Responsible People:

Patrick

Deadline:

September 2020

Agenda Item 4:

General questions and comments

Discussion:

Ben: There’s a BAMS paper I was involved in where we will have a high-quality emission model. I post the link in the chat.

Link: https://journals.ametsoc.org/bams/article/doi/10.1175/BAMS-D-20-0085.1/353866/REFERENCE-QUALITY-EMISSION-AND-BACKSCATTER

Ben: I think that covers the main things out of our group. Are there any other issues to discuss?

Yingtao: I am wondering if there will be a testing profile set for the new release?

Ben: Yes, I have been working on a comprehensive testing profile set with thousands of profiles. The goal is to provide a comprehensive case. All the tests will be based on cmake and ctests. We want to move to a nightly build system like for JEDI with continuos integration.

Yingtao: That also will include the surface emissivity data?

Ben: Yes.

Isaac: That was NOAA-funded work I did for Ming Chen. CRTM did pretty good in all ATMS channels. In some cases, CRTM is more accurate than RTTOV. I can put together some slides.

Ben: That may be timely because there will be a talk about an RTTOV implementation as a UFO.

Isaac: So, we only compared the models in terms of spectroscopy. I might be able to compare CRTM and RTTOV within a month or so.

Ben: Is this with cloud scattering?

Isaac: No, this is just for clear sky.

Ben: With the clouds there are the most differences.

Isaac: With the clouds this is not really an apples-to-apples comparison because RTTOV is doing all-sky assimilation much longer.

Ben: This is a multi-year effort not only for RTTOV and ARTS, but also ARMS in China.

Ben: Any other questions?

Nick: You mentioned a BAMS article. What kind of article is that?

Ben: It’s more of an overview, not the level of detail you might want. It is just an outline of developing the emission model.

Nick: When the final version comes out I’d be interested. Then in addition a little update. I am still working with Bob Knudsen. He wants to give me a validation set from ship in the Pacific. Right now I have a very warm water and a very cold water data set and the Mary dataset will be somewhere in the middle. And we will probably publish it in IEEE as two articles. Yim Jung is working on the assimilation side where he has encountered some difficulties. In addition to that there is the JPSS proving ground work and this is for snow emissivity model. In this project I want to develop a snow emissivity model. In addition, we want to adapt the CAMEL-based approach from RTTOV. We want to adapt both. After that I will have to be in contact with you and Kevin.

Ming: CAMEL has already been used in CSEM. It’s already there in the CRTM REL-3.0. We included all the RTTOV surface models.

Nick: I am not a simulation guy. This is something I was approached about.

Ming: CAMEL has some updates.

Nick: No doubt the focus will still be on the modelling side. The CAMEL part was just something that didn’t seem to be in the CRTM. I just thought this was something to be good to work on since this was on the NEWCAPS retrieval. If CAMEL is already in your module that’s a good thing because that will facilitate my work.

Ming: Ben, your reference model is MW ocean only?

Ben: Yes, MW ocean only with atmospheric correction coefficients from L-band to 8mm.

Ming: This new reference model is quite different (unintelligible) You have to have the atmospheric correction.

Ben: The reference has to have the polarized BRDFs. We want to take the state of the art right now and want to create an accurate reference model including foam etc. All of the key players are on board because it benefits all of us. There’s a lot of interesting work going on.

Ming: You can include me in the discussion.

Ben: This model will be hosted on the JCSDA GitHub without restrictions. Any other updates?

Ming: Another thing is, I am working on the implementation of L-band RSS model (unintelligible)

Ben: Could you share some slides with us?

Ming: (unintelligible)

Ben: I am curious to see that because Hamideh is working on something very similar with Guillaume. Have you shared these results with Thomas Meissner at RSS?

Ming: Yes. (unintelligible)

Ben: Anything else? Tom Greenwald?

Tom: I really don’t have any updates.

Result:

N/A 

Tasks:

N/A 

Responsible People:

 N/A

Deadline:

 N/A