Weather Enterprise Study – The Front Page

Community Modeling and the Future of Numerical Weather Prediction

December 4, 2024

A 2024 AMS Summer Community Meeting highlight

The AMS Weather Enterprise Study will provide a comprehensive picture of the shifting landscape of weather-related fields to inform our joint future. At the 2024 Summer Community Meeting, working groups discussed what they’d found about key issues facing the enterprise.

Here are a few takeaways from the Community Modeling working group, as reported by Gretchen Mullendore of the NSF National Center for Atmospheric Research (NCAR). Community modeling employs Earth system model software developed by public-academic partnerships. Community models have open-source components and are freely available for use by anyone with the computing power to run them–for example researchers, students, and private companies.

Photo: Gretchen Mullendore

How has the community modeling landscape changed in recent years, and where are we now?

First, artificial intelligence and machine learning (AI/ML) have become huge players in numerical weather prediction (NWP) model development. Second, a cultural change in weather research and forecasting is taking place; we’re beginning to collaborate much more closely across agencies and industries than we used to, and many people are invested in deepening those collaborations.

What were the main themes that came out of your working group’s discussions?

The NWP community is collaborating more than ever before. However, the community remains spread among many institutions, with each research group working on small pieces of the overall weather prediction challenge. Having many research groups can be a strength in terms of encouraging innovation, but it is a weakness if research isn’t coordinated effectively to fully realize collective benefits. Limited funding and resourcing is an additional barrier to community model development. As a community, we need to continue to prioritize modularity and interoperability across NWP systems and work towards more effective shared governance.

Another major theme is the role of the private sector in NWP. Big tech companies are increasingly getting into NWP and there is a concern that public forecasting efforts are not able to keep pace. The private sector brings agility and innovation to the field, and working to leverage unique contributions across public, academic, and private research entities is valuable. However, if the growing role of the private sector in NWP leads to more observations, simulations, and software being behind proprietary walls, there is risk to accessibility and collaboration.

The NWP community is also facing challenges in workforce development. Universities are teaching people the right skills to work in data assimilation and analytics, but many of those people are being scooped up by private sector companies in other fields offering salaries that employers in the weather industry cannot compete with. We need to better communicate the value of our missions and our work to attract and retain talented early career professionals.

What preliminary recommendations or future directions have you discussed?

We can and should continue to build on community efforts to coordinate across public, academic, and private developers. This coordination should include planning for the appropriate use of AI/ML tools in NWP research and applications. We can also build on efforts to leverage social science research to prioritize our limited resources, e.g., by learning what type of forecasting improvements will most benefit stakeholders. Finally, we need to recognize the importance of the legislature in resourcing model development. It’s important to communicate our successes and the value of a thriving NWP community. In summary, we should strategize to develop intentional communication among ourselves, across disciplines, and most importantly, with legislatures and the public.

What did you hear from the community at the Summer Community Meeting?

My pick for the most important question asked at the SCM is, “What does success look like in NWP development?” The goal that motivates us all in the NWP community is for no more deaths to occur as a result of weather hazards. In order to achieve breakthroughs in prediction that stand to move us closer to that goal, we need to invest in innovation, which requires risk. However, much of the work in NWP development is funded by federal agencies, which tend to be risk-averse. More broadly, the systems in which our scientists work can be an impediment to innovation. For example, the pressure to publish often incentivizes incremental progress over new ideas. Collectively, as an NWP community, we need to build systems that allow researchers to take risks without fear of failure or negative consequences.

What are the main challenges, conflicts, or points of discussion identified by the group (or at the SCM)?

AI/ML could possibly improve the skill and speed of all parts of the NWP system. That said, the challenges are also great. Challenges include a lack of AI/ML expertise in NWP community leadership; a need to invest in AI/ML without additional resources; and a need to keep up with the latest AI/ML research, which is moving incredibly rapidly. The lack of clear AI/ML plans from U.S. institutional leaders in NWP led some to ask at the SCM if leaders were skewed against it. My perception is instead that the community is feeling overwhelmed by these challenges. We can overcome these challenges through innovation and collaboration, leveraging our respective expertise and investments to more efficiently take advantage of the great opportunity that is AI/ML in NWP.

Want to join a Weather Enterprise Study working group? Email [email protected].

About the Weather Enterprise Study

The AMS Policy Program, working closely with the volunteer leadership of the Commission on the Weather, Water, and Climate Enterprise, is conducting a two-year effort (2023-2025) to assess how well the weather enterprise is performing, and to potentially develop new recommendations for how it might serve the public even better. Learn more here, give us your input via Google Forms, or get involved by contacting [email protected].

About the AMS Summer Community Meeting

The AMS Summer Community Meeting (SCM) is a special time for professionals from academia, industry, government, and NGOs to come together to discuss broader strategic priorities, identify challenges to be addressed and opportunities to collaborate, and share points of view on pressing topics. The SCM provides a unique, informal setting for constructive deliberation of current issues and development of a shared vision for the future. The 2024 Summer Community Meeting took place August 5-6 in Washington, DC, and focused special attention on the Weather Enterprise, with opportunities for the entire community to learn about, discuss, debate, and extend some of the preliminary findings coming from the AMS Weather Enterprise Study.

What’s the Future of Weather Decision-Making?

November 12, 2024

A 2024 AMS Summer Community Meeting highlight

The AMS Summer Community Meeting drew exceptional attendance and engagement this year as people across sectors helped inform an upcoming report on the Weather Enterprise. The AMS Weather Enterprise Study will provide a comprehensive picture of the shifting landscape of weather-related fields to inform our joint future. At the 2024 SCM, working groups discussed what they’d found about key issues facing the enterprise, and asked for feedback from the community.

Here are a few takeaways from the Decision Support Services working group, as reported by Matt Corey (pictured at left) of Microsoft Weather. Decision support services (DSS) help stakeholders make weather-related decisions that are informed by the best available knowledge across fields. They are crucial for emergency managers and many other decision makers, as well as members of the public.

How has the decision support landscape shifted in the last decade or so?

Stakeholders for DSS range from an emergency manager making critical decisions about an entire community to an everyday citizen making a decision for themselves or their family. For decision support services, the last two decades have seen an abundance of technology changes which have allowed stakeholders easier access to information. However, this can be both a benefit and a challenge, as misinformation has also become more readily available.

What were the main themes that came out of your working group’s discussions?

The themes that emerged for us included:

The different sectors of the Weather Enterprise have become coupled, with less well-defined boundaries when it comes to providing decision support.
New players are entering the enterprise, with growing AI and novel ideas.
Developing and maintaining the necessary workforce is a concern.
There are increased opportunities for translating forecasts into easily understood language in order to support decisions.
There is a need for increased funding for quality observational datasets for many applications, especially in AI.
In a complex, misinformation-rich environment, there is still room for all sectors to tailor communications to stakeholders, but there is also concern about maintaining consistency in order to maintain trust.
Embracing user centric design to understand stakeholder concerns, technical levels, and understanding is important, including the use of probabilistic information. Example: “There is an 80% chance the flooding will happen this afternoon.”

What are the main challenges you have identified?

In our group, the discussion continues to be about who should be providing decision support services. As the NWS gets more involved in DSS, one concern is for increased friction from some private sector entities. Another key point is that DSS is not limited to a specific stakeholder type. DSS is important to all citizens who need to make decisions involving weather every day, thus there is a shared dimension and need for responsible and clear messaging to all stakeholders (including the tactical use of probabilistic information).

A final recurring theme is around the workforce itself. Forecasters need to be taught communication skills, and social science is critical in helping to understand the needs and problems to be solved for the end users. With the focus shifting to newer tools including AI-infused capabilities, there is a concern that the new workforce will lose the necessary skills critical in conveying adequate decision support services.

What preliminary/tentative recommendations, solutions, or future directions have you discussed?

Some of the recommendations we’re working with right now focus on:

Integration of weather, water, and climate information with socioeconomic and biosphere information for earth system forecasts.
Cross-sector support of ecological forecasts and environmental early warning systems (for example, warnings of fishing industry impacts due to warmer water) to benefit society and facilitate impact-based action.
Improved communication about weather impacts, especially in a changing climate, using common terms and learnings based on stakeholder’s decision needs.
Embracing AI as a way to increase the velocity of forecasts, integrate probabilistic information into forecasts, and increase efficiency for both short-term services like nowcasting and long-term climate solutions for all.
Helping meteorologists to become the communicators that they should be. Leveraging AI solutions and tools to help make them more efficient at helping stakeholders with their decisions.
Expanding opportunities for smaller businesses/individuals to obtain specialized DSS.
Increased public awareness of changing weather patterns stimulating the need for better accuracy, earlier warnings, and long-range projections.
The need to smartly integrate probabilistic information to help stakeholders better understand forecasts and limitations.

Want to join a Weather Enterprise Study working group? Email [email protected].

About the Weather Enterprise Study

About the AMS Summer Community Meeting

How is Weather Research Changing?

September 24, 2024

A 2024 AMS Summer Community Meeting highlight

The AMS Summer Community Meeting (SCM) drew exceptional attendance and engagement this year as people across sectors helped inform a major upcoming report on the Weather Enterprise. The AMS Weather Enterprise Study will provide a comprehensive picture of the shifting landscape of weather-related fields to inform our joint future. At the 2024 SCM, working groups discussed what they’d found about key issues facing the enterprise, and asked for feedback from the community.

Here are a few takeaways from the Research Enterprise working group, as reported by Daniel Rothenberg of Brightband.

Photo courtesy of Daniel Rothenberg.

How has the weather research landscape shifted in the last decade or so?

Two of the most important shifts have been a movement of exploratory and applied research from the public to the private sector, and the rise in importance of “data science” and other hybrid roles blending a mixture of domain expertise and broader engineering and technical skills.

Possibly the biggest example of these shifts coming together has been the advent of AI-based weather forecasting tools, although it also shows in trends such as the rise of private companies operating earth observation platforms.

What were the principal themes that came out of your working group’s discussions?

One major theme we discussed was the balance of responsibilities across the traditional weather enterprise. Initiatives such as building and launching satellite constellations or developing new weather models were at one point solely within the remit of the public sector (due to complexity and cost), but are now commonly undertaken by the private sector – sometimes even at start-up companies.

This re-balancing opens as many opportunities as it does challenges, and leads to another major theme: how we can best prepare for the workforce needs of today and tomorrow. Meteorologists will increasingly need to apply technical skills such as software development and data science alongside ones from the social sciences; preparing our current and future workforce for these demands will be a challenge in its own right.

A third major theme is that the weather enterprise is getting bigger. We’re not just a community of meteorologists anymore. Increasingly, critical work related to weather, water, climate, and their impacts on society is being undertaken beyond the traditional boundaries of our enterprise. There is a significant opportunity to improve society’s resilience if we as a community are able to build relationships with the new institutions working on these issues in a collaborative, interdisciplinary manner.

What are the main challenges you have identified?

Better accounting for how we ought to invest limited – and declining – federal resources will be a significant and contentious challenge, only complicated by the shifts in priorities and capabilities across the enterprise.

Those shifts motivate a second key challenge, which is clarifying who in the enterprise is accountable for, or has ownership over, certain areas. For example, NOAA makes available nearly all of the observations used in its operational forecast models, with some exceptions for proprietary data from commercial entities. But as more private companies try to sell data to NOAA, how will this balance hold? What if those private companies move towards selling actual weather modeling capabilities or services – perhaps a proprietary AI-based weather model – to the government? In the case of expanding commercial data purchases, who is responsible for maintaining and improving our data assimilation capabilities?

Coordinating many actors across the enterprise, in a manner that most effectively serves our mission to society, will be a key challenge we must navigate in the coming years.

What preliminary recommendations or future directions have you discussed?

Our tentative recommendations revolve around building robustness. We encourage academic organizations who train our future meteorologists to consider how to prepare these students to work in a multidisciplinary capacity, and to embrace data science skills. Not everyone needs to be an interdisciplinary scientist, but it’s vital that our students learn how to apply their deep domain knowledge as part of a team of such individuals.

We also acknowledge that the rise of AI/ML techniques is changing the demands of our computing and data infrastructure. Not only must our workforce learn to adapt to these technologies, but we must consider how the enterprise will support enabling them: for example, by ensuring that in addition to large, traditional high-performance computing resources, we provide access to GPUs and similar tools. As part of this re-evaluation, we must evolve the ways in which we as a community define our priorities for federal research funding

What did you hear from the community at the SCM?

We thank the community for the warm reception to our assessments at the Summer Community Meeting. Many of the themes we touched on – the re-balancing of capabilities across the enterprise, the emergence of AI/ML and its implications, as well as core workforce development concerns – were echoed across many other working groups, underscoring their importance.

Within our group, we also discussed the growing importance of convergence science, which was echoed several times throughout the meeting. Convergence science, which involves coordinating diverse, interdisciplinary research teams with real stakeholders to solve societally relevant problems, is likely to be an important mechanism of translational research in the future, but we (and others at the meeting) identified a need for federal agencies to devote more resources earmarked for this sort of work in order to complement traditional, siloed funding programs.

Want to join a Weather Enterprise Study working group? Email [email protected].