10:00AM Mountain Daylight Time
Members present: Allison Baker, Ben Johnson, Brad Klotz, Carol Ruchti, Eric Apel, Holger Vömel, Kelley Barsanti, Kevin Pham, Lulin Xue, Mariana Cains, Olga Wilhelmi, Peter Lawrence, Ryan Sobash
Also present: Glen Romine from the Directorate’s Office
Discussion with Glen Romine
Things are busy as always. We should start off with your request for funding social events. When we talk about requests for funding that supports catering or anything primarily for staff and not for external audiences, it has to come from discretionary pools and discretionary pools are very small. The requested amount was about half of the discretionary budget for the whole Center. We still would like to help the NSA get traction with staff as a whole so we should look for other options. There are other alternatives, such as the ECSA has a pool to fund travel for early career scientists. It’s not a dollar amount, it just comes from a different bucket. We can provide admin support or want to provide activities that don’t require discretionary support, we can do some, we can’t just fund the size of your original request. We also would like to engage directly with the EC. Have you reached out to Susan Chavez? The upcoming schedule is pretty atrocious but we should start a discussion for the NSA-EC to meet with the NSF NCAR EC.
Strategic Plan
We’re close to getting a written version out that we hope to be sharing with staff and external audiences in a week or two. The goal is that we’d have a rough draft for a general audience for feedback. We hope to have it up for comments for 4 weeks. There will be a Google Form that will be shared with staff and that’s how we’ll be collecting feedback on it. Google Forms are useful because you can export things to a spreadsheet and sort them. It helps to make sense of the responses. We’ll be sending it out to the UCAR Member Institutions and Packer. Packer is the super group of the UCAR members. We’ll also be sending it out to the National Science Foundation and the Board of Trustees. We’ll be sending it to the Board of Trustees in February and sending it to the National Science Foundation in March. The timeline is getting tight.
INFORM Program
INFORM is a fancy way about thinking about data assimilation. Except it’s with research observations and how we inform modeling activities and field campaigns. How do you collectively pull all the data together to make sense of all of the collective information that can inform the process. Part of the driver is if you think about the Phased Array Radar, you can put it on the C-130 but what other instruments could you also put on the C-130 to complement APAR?
Community Software Facility
If you look at our modeling ecosystem, we have a lot of different models that address different parts of the Earth System. Most of them are atmospheric, but we also have cryospheric, heliophysical models, etc, all integrated with CESM, from near the surface to the ionosphere. Each of them started as small models, when we think about Earth System Predictability Across Timescales, it becomes confusing because of all of these overlapping capabilities. Each of those tools are developed independently of each other. The computing environment is also changing where historically HPC centers had x86 processors. Now much of the demand is being driven by AI in industry. The problem is that our codes don’t run on modern machines. How do we ensure that we can still do computation in the future? It’s quite possible that our next super computer could cost the same amount and have the same computing power (whereas historically we had 3x computing power with each new generation). And it gets worse from there. How do we adapt to the future? We’ll start by hiring a consulting firm, but we can also kickstart things now. Two other things that we need to be intentional about:
- Artificial Intelligence. We have lots of efforts throughout the Center but it’s unclear where we should focus. We should step back and take an assessment.
- Data assimilation: Right now it’s a touchy and challenging space. If you go to the National Science Foundation, they don’t understand DA as an area of research. They view it as an operational capability. They don’t view DA as a tool that can improve the models or extract information, say from the atmosphere, and transmit it to the deep ocean. We have a complex ecosystem of tools, but they aren’t necessarily matched to these research problems. So we need to take a step back and see how they can address these research problems.
Question and Answer with Glen
Question: The project you were talking about is INFORM? What’s that stand for?
Integrating Field Observations and Research Models. It’s a pretty interesting project and I’m looking forward to see it kick off. We are looking for a potential lead for that project. If some of you were interested, I would encourage you to reach out. There are folks on this call that would be great at providing leadership in that space.
Question: You’ve talked about this issue we’re facing in the future with computer codes. Interested in learning more about the real detailed plans about that.
We’re trying to get more formalism and consistency across the Center. We have other efforts, in CISL for example, called the Research Data Commons. We have all of these data sets. One of the potential uses of AI is merging data sets, maybe combining social, ecological data. There’s opportunities to look at these intersectional datasets and seeing if we can develop observational capability to see if these are real relationships. But to do that work you need to have this data in AI-ready format. For example we have data that has metadata that isn’t CF Compliant. So we have to set some standards so this data is usable.
We also have to think about how we improve user support. You don’t want someone who is having a namelist issue talking to a Software Engineer 4 about that. We need to be more efficient with how we use resources for user support. Meanwhile you have someone in the university community that might be funded by NSF that develops new capability for a modeling system. Right now they just develop the capability, hand us the code and tell us to, “Get it in there.” We need to develop more formalism. Right now it’s endless rope. For example, CESM has a code doubling rate of every 5 years. Ignoring this is getting us into trouble fast.
Question: Are there going to be any constraints on the level of AI that we’re using? There’s more simplified and supervised learning versus the black box where you don’t understand what’s happening under the hood. Will there be constraints on what tools will be allowed? Are there concerns about security about how these tools are implemented in the frameworks?
We’re not going to constrain what people can and cannot do. The goal would be to establish guidelines with limited resources. If we look at Digital Earth. That would be a tough space to get into because private companies are going to kill us. We’re not going to be community leaders in that space. Maybe there’s other space that we can be community leaders in.
In terms of being cautious about how it’s used or applied, I’m not a technical enough person to give advice about what should be done. There’s also concerns about ethical AI – what it should and shouldn’t be used for. Do we have to pull in outside experts to understand this? It’s a known risk but at this point ethical AI is a grassroots effort that is scattered.
Question: Our current funding request is in the discretionary category. Do you have any advice on other formats that would tap into other pools?
If you wanted a travel fund, that would be an indirect cost, so that’s a larger pool that wouldn’t be as constrained. If you wanted administrative support to help you with planning meetings, that’s something that we could give you. If you want an account key to charge time for service activities or if you wanted to host career fairs, that’s something that we could help you with. If you wanted to provide food to staff, that’s challenging, it would have to come from discretionary funds. If you had external participants above a certain threshold, we could provide funding from a different pool. Food is expensive here, the rates went up about 25% about 4 or 5 months ago. If I were to guess, if you asked for something in the $2,000-3,000 range that has a reasonable chance of getting approved.
Glen departs the meeting and the NSA-EC representatives move on to other parts of the agenda.
Follow Up About Social Events
Glen continues that he wants to support in some way. We originally talked about doing something to bring scientists together to get the NSA more active or more connected. We were talking about a poster session. We also felt like we didn’t have a whole lot of connection with our own lab. That’s what this came out of. Still think about having these lab level events but maybe making it over coffee and cookies.
Glen also opened the door for us to ask for funds from a different pot. He did mention that it would be possible to have administrative time. Maybe it would be a way to garner more interest in chairing the committee. A timekey that everyone who attended the meeting with the NSF NCAR EC. Still majority support for lab-level events. We’ll split this out and make a request for other funds. If Glen gave the example of the ECSA travel fund. Would it be a good idea to have an NSA fund similar to that? ECSA has about $12,000 for visitors and $30,000 for travel to conferences and workshops. We could pursue something like that but perhaps not at that scale. Agreement on this idea.
Relationship of the ECSA and the NSA-EC
There are 2 ECSA representatives of this group. The ECSA has a diagram that we show at the town halls that shows the ECSA is part but not totally part of the NSA because the ECSA also has research engineers. They’re related enough. There’s definitely some overlap. There’s language in the Bylaws that the ECSA is nested in the NSA. There’s a perception of the ECSA being more visible. There’s a lot of enthusiasm with early career scientists. As careers progress, perhaps scientists want to pursue more of their research. It’s a good opportunity that speaks to more senior scientists.
Other ways NSA-EC can support scientists at NSF NCAR?
It would be nice to have targeted things, a way to poll staff so we have targeted feedback at these events. This would be a good way to let people know they have input.
Our Future, Our Space
Lulin is available for the 9-11AM focus group on October 29th, 2024. Eric is on the advisory committee already.
Day of Discovery
Question: Any feedback to pass along to Glen about the Day of Discovery?
There seemed to be an imbalance of representation among labs based on the scope and breadth of work. Some in RAL felt that it was underrepresented. There didn’t seem to be good feedback to the labs. People didn’t hear back until the day before so people were confused as to whether or not they were presenting. We need a better timeline and more clear communication about when we’ll hear back about the agenda. A similar communication problem existed last year. The agenda wasn’t released until a day or two before the event. There was no idea about who would be presenting. The organizing committee of the Day of Discovery should provide a more clear timeline.
Action items
- Attempt to find a time when the NSA-EC and the NSF NCAR EC can meet.
- Draft new funding request letter to Glen Romine. Identify which topics have the most support among representatives for the request:
- Scaled back social events in the $2,000 – $3,000 range
- 0.05 FTE support for NSA-EC cochairs
- Travel grants
- Event planning support for events that include external participants (Glen’s example was a career fair)