Mark: I just wanted to share some information with you. Jim Hawk and I had an email communication. For the parallel RTTOV the parallel parts are only used for postprocessing. But for the DA part they still use the single process.

Ben: I don’t know if they told you but there is a plan for me to go to UK MetOffice to help them get their parallel work going in July. I don’t know if it’s going to happen or not.

Mark: In 2010 I was invited as well and they paid the lodging cost. They invited me for one month to work on the FASTEM model. But I only staid two weeks because of the Soumi NPP launch.

Mark: They have a lot of programs to invite people from other countries. The JCSDA has also programs.

Ben: I try to encourage that, but so far, it’s been only French students. I tried to get a visitor from Goddard, but that didn’t work out.

Mark: For the profile loop adding parallelization that is very fast, but at the channel level that takes longer. I have the code at the channel level, it works fine. But for the channel loop there is some overhead because you need to define some private variables. The overhead is about 5%.

Ben: I think we are kind of ahead of the curve already.

Mark: Yeah, that’s different from ATS. At that time John Derber asked to have a parallel part. The operational system at the time had threads. If you use a serial implementation you waste a few threads. Today I talked to IT people and in today’s machines each thread is treated as an individual CPU. But still, going parallel has some advantages, such as saving some memory.

Ben: Nick, did you want to give an overview of your ISSI talk?

Nick: I joined a little late, did you talk about the ISSI already?

Ben: Yes, it’s a team under the lead of Steve English to create a reference ocean emissivity model with most of the expertise in MW. We had some guests already to work from L-band upwards. We realized we had a gap in the IR and so I suggested to invite Nick.

Nick: Right, Steve English is the team leader/coordinator and they’re focused on MW for the ocean emissivity. They want this good, reference model treatment across the entire IR realm, so Steve asked me to give a 20-minute briefing about the state of IR modelling. I overviewed it at a high level with a historic perspective. The only reason we have IR under wraps is because of high-accuracy Airy instrument observations. We have these instruments looking at the ocean surface and this is the only reason why IR has become an established part of the forward operator. Right now, the challenge is really the consistent treatment of the surface itself. If you put a detector above the surface you are looking at the radiance, which is a convolution of the reflectance part as well. That’s where the real problem with the IR part is. If we just model the emissivity itself the Suder 2006 model is good enough but you also have to take the BRDF into account.

The MW is very focused on modelling the foam. In the MW spectrum the foam has a large impact. In the IR it doesn’t change the refractive index, but it changes the slope. They brought my attention to some papers. In the MW they also have the same concern in the CRTM/RTTOV world, all these models are fast models. If we really want to do the full-blown double-integral radiation we can do that but it’s not fast enough for a practical model. That is the concern right now. Back to you Ben.

Ben: Thanks Nick, I put a link in the chat about the notes of our first meeting. I hope we can publish at this high level so that we can work at this referene model level.

Mark: Ben, I think for the next meeting maybe Ming Chen can attend the surface meeting.

Ben: I agree.

Ming: I think that it’s certainly very important to have a reference model. But the problem is what the reference model looks like. If it’s a physical model can we be sure that the physics are accounted in the current model, or are we formulating it in an accurate way? For a physical model there are some issues already, notwithstanding a practical model. We need something that at least discusses the physics, let’s say the reference data from observations. So, in this case we have cross-validation with satellite. In this case we can extend that. Over land we have more desire to have some high-quality data. We can already see some improvement but it’s very challenging on a global scale. This will be hard. I don’t know if they have a plan to bring in a reference model and bring it back to DA. So, this is not a simple problem for a small group. All the members are certainly experts in their sphere but if it’s possible and if they really want they should invite more people. Even within the CRTM you provide the framework for cross-validation. I know that Mark was working on this (or Paul). But there is some slight difference in the spectral regions. But also, in the computational assimilation. Many pieces to work on. Also, we needed to combine expertise to include e.g. wind speeds. I have been focusing on a model framework. Also, Mark raised some very good point with the wave spectrum. We didn’t consider any development in ocean waves.

Mark: Ming, I just wanted to check, did you get the version 3.0 beta version?

Ming: No, I think Kevin said Yingtao should do some testing.

Mark: If Ben has time he can share it.

Ben: Mark, you deleted the REL-3.0 from the ftp site?

Mark: I deleted the REL-3.0 tarball because the ftp site of STAR is public.

Ben: Ming, on the sea surface emissivity reference model we are starting with the dielectric constant but if you have that perfectly characterized there is still the issue of the interface with issues such as wave age. They are very focused on getting the physical base model very well characterized.

Mark: Two weeks ago Steve English and Thomas Meiser contacted me and it looks like for the operational ocean surface emissivity model there are two groups. One model is FASTEM and the other is the remote sensing system that Thomas Meiser has. Thomas contacted me because before he coded the FASTEM part based on a paper, he mentioned one coefficient is a Sine and the other is square and he got some results that look different when he followed the paper. When he changed the sine and the predictor from square to cubic he gets very good agreement. He is right, it was a typo in the paper. But the code is correct. Only in the paper is a typo. In this sense RSS and FASTEM agree well. But he also pointed out for the salinity their model is probably more realistic. I haven’t got time to check the low frequency part.

Ming: I did a comparison between RSS and FASTEM in the L-band.

Mark: Right now, RSS and FASTEM agree well.

Ming: True. Two, three years you told me that they need some standalone version of FASTEM. For example, the JPL model they use a different calibration.

Mark: RSS is a company. They are very strong. Frank Weights has lots and lots of expertise in the MW.

Ming: But RSS separates into L-band and not-L-band. When I coded their TL/AD it’s very challenging.

Mark: But there is some physical background as Ben already mentioned, which is called the boundary layer. Currently all the physical models depend on the sea-surface speed. But they never consider the development of these things. Another issue is foam. For the MW the foam impact is very large.

Ben: Foam can persist even if the wind slows down. At high frequencies foam becomes more of a scattering problem. For RSS they only go up to 89 GHz. We definitely want to go up to 183. The surface has a significant impact. There is an effort in this group to improve the high-frequency surface properties. The Aquarius mission was very helpful in this regard. So far, the RSS model has only been compared to few other models. The frontier in MW models is in high-frequencies.