WMO is reaching out to the private and academic sectors as it seeks to leverage Artificial Intelligence to improve the forecasting capabilities of National Meteorological and Hydrological Services, whilst also recognizing that major technological companies are now entering the weather prediction space.
An Open Consultative Platform on Artificial Intelligence on 16 June, set the scene for WMO Executive Council discussions on how to be prepared for the changing industry landscape - and the opportunities and challenges this entails for Earth System prediction.
"The rapid and transformative development of AI is truly astonishing. The convergence of a new generation of AI systems linked to the availability of the next-generation reanalysis will likely drive improved accuracy of AI-based predictions," said Michel Jean, President of the WMO Commission for Observation, Infrastructure and Information Systems (INFCOM) .
He said that the emergence of AI does not render past and current efforts obsolete. The value of observations and physics-based forecasting remains undiminished. Indeed, the WMO's World Weather Watch, established in 1963, and its open data policy paved the way for the technological revolution now under way.
"In light of the roles played by NMHSs and the WMO's global infrastructure, it is essential that we continue to deepen our mutual understanding and strengthen our collaboration with the private sector and academia. To get the best out of all this, we need to combine expertise in data science, observing systems, and earth-system science - hence working together. We are in this together," he said.
Private sector and Academia
Public private cooperation is a win-win for the world and has the potential to improve the lives of many millions of people, and to contribute to the Early Warnings for All initiative . But it is essential to ensure basic ethical and scientific standards are respected and that nobody is left behind.
The private sector is agile and can embrace innovations faster. Public institutions invest in establishing infrastructure that creates the huge data sets needed for the AI algorithms and serves a wide range of different users.
Speakers from Google Research, Microsoft Research and Accuweather (representing the HydroMeteorological and Environmental Industry Association), and the Shanghai Academy of AI for Science showed the huge potential which lies ahead.
They stressed their commitment to collaborate with - and not compete against - NMHSs as the authoritative providers of essential services and warnings - and in fact there are already many examples of this happening in different countries around the world.
Thus, for example, Google is partnering with the US National Hurricane Center on AI-based tools for hurricane forecasting and is working with WMO and meteorological services in other countries including Viet Nam on efforts for flood forecasting. It is using AI for flood forecasting at global scales.
Microsoft has partnered with the UK's Met Office on supercomputing. Through its AI for Good Lab, Microsoft is committed to the UN's Early Warnings for All agenda - empowering National Meteorological and Hydrological Services (NMHSs) in the detection, observation, monitoring, analysis, and forecasting of hazards.
The Shanghai Academy and the China Meteorological Administration promote an all-in-one service on a cloud platform which is a potential tool to help assist countries without modern data centres.
Roar Skalin, permanent representative of Norway to WMO , presented highlights of a joint pilot project between ECMWF, the Norwegian meteorological service and the Malawi meteorological service which combines global expertise and local know-how.
The training is performed on extensive computational resources of the global North and the operational run will be done through modest resources in a developing country.
Arlene Laing, Permanent Representative of the British Caribbean Territories with WMO , showed how AI models are already improving tropical cyclone track forecasts.
The Open Consultative Platform was the sixth in a series of annual events bringing together the public and private sectors.
Here are some of the key takeaways:
Open Principles
- The principles of openness, transparency, and traceability are foundational for the effective integration of AI-based Earth System Prediction (AI-ESP) systems into the WMO Integrated Processing and Prediction System (WIPPS).
- Access to open data has unleashed innovation in the academy and private sector, and public databases and reanalysis data sets are feeding the AI revolution. The research and industry ecosystems are in need of open, reliable data feeds, which are a benefit from the public investments.
- The private sector has to strike a balance - providing sufficient detail to demonstrate the depth of their research and the effectiveness of their solutions, while protecting the intellectual investments that drive continued innovation.
Supporting small NMHSs
- AI must help small NMHSs and developing countries, rather than jeopardize their activities and undermine their mandate.
- The democratization promise: AI technology is an opportunity for the least developed/developing world to "leap frog" to more advanced prediction capabilities.
- Reduced Computational Burden: Once trained, AI-ESP models need smaller computational resources than traditional Numerical Weather Prediction (NWP) models.
- Filling Observational Gaps: AI-ESP models can utilize diverse data sources, including reanalysis data and satellite observations, which can help address significant observational data gaps, especially in developing countries.
- Cost-Efficiency: NMHSs can gain access to advanced observational data at no or low cost through partnerships involving non-NMHS entities. Collaborative networks with commercial partners can also provide observations in a more cost-effective manner or with less implementation/operational risk.
- Knowledge transfer/ training is essential to enhance the capabilities and allow proper local adaptation and appropriate use of the new technologies;
Public-private sector roles
- AI reshapes how and in some cases by whom forecasts are produced, and there is a need for proper framework to optimize this.
- AI-based systems should be harnessed to support authoritative extreme weather warnings services provided for the protection of life and property.
- Single Authoritative Voice: NMHSs will continue to be recognized as the trusted public service and single official and authoritative voice in the provision of weather and hydrological forecasts and warnings for public safety. This role is critical for supporting decisions related to natural hazards and disaster risks.
- Capacity Builders: NMHSs will continue to engage in and benefit from capacity development initiatives to adopt new technologies and respond effectively to the rapidly evolving data landscape and user demands for more diverse and precise services.
- Private companies can help amplify the warnings from NMHSs by further distributing those alerts to consumers and businesses within their countries.
- AI/ML holds the possibility not only to improve the prediction, but also the decision-making process, that is - one can design end-to-end AI/ML solution. This holds a promise on one hand but also may conflict with the Single Authoritative Voice principle. Ethical considerations need to be discussed together to address these challenges;
- It might be beneficial to all sectors to establish a joint coordination body to have open continuous discussion and elaboration of joint principles.
