SAGE - TMR Joint Workshop on AI/ML Application on S2S

Enhancing Sub-Seasonal Predictions with AI: Insights from the Joint SAGE–TMR Workshop 

 

Enhancing sub-seasonal to seasonal prediction and advancing tropical meteorology research were recognized as vital for strengthening early warning systems and improving climate-sensitive services in agriculture, health, energy, and disaster risk reduction — essential steps toward building resilience amid increasing climate extremes. At the same time, integrating Artificial Intelligence (AI) and Machine Learning (ML) into forecasting was seen as offering powerful new opportunities to boost prediction skill, operational efficiency, and sectoral service delivery. 

To explore these emerging frontiers, a hybrid joint workshop was hosted on 26 May 2025 on "Enhancing Sub-Seasonal Predictions with AI for Early Warnings" was convened by the Sub-seasonal to Seasonal Applications for Agriculture and Environment (SAGE) project and the Working Group on Tropical Meteorology Research (WGTMR), under the World Weather Research Programme (WWRP) of the World Meteorological Organization (WMO), with Shandong University as the local host. 

Invited webinars and panel discussions have been included, focusing on:  

S1- AI&ML application in S2S and tropical related predictions 

S2 - S2S application in serving sectors 

S3 - AI/ML for DRR and Early Warning  

Agenda and links to the available presentations can be found below:  

Time (UTC+8) 

Theme 

Presentation and Speakers 

09:00-09:30 

 

Opening Ceremony   

09:30-09:45 

 

Introduction about EW4All, WWRP, TMR and SAGE. Presentation 

 

Kunio Yoneyama (WWRP Scientific Steering Committee)  

09:45-10:15 

S1 

FuXi Weather: A Machine Learning System Surpassing Traditional Global Forecasting in Data-Sparse Regions. Presentation 

 

Xiaohui Zhong (Fudan University)  

10:15-10:45 

S3 

A multi-hazard early warning system framework for the SADC region. Presentation 

Dewald van Niekerk (African Centre of Disaster Studies/North-West University) 

10:45-11:15 

 

Coffee Break 

11:15-11:45 

S3 

Text as Data: LLM-Driven Disaster Warning Dataset and Optimization  

Chao LI (China Warning Center of CMA) 

11:45-12:15 

S1 

Statistical and ML approaches in air pollution prediction: Interpreting the multiscale time-series characteristics and physicochemical processes  

Tao LI (Shandong University) 

12:15-12:45 

S2 

Automate, anticipate accelerate – harnessing machine learning to mitigate climate risks in agriculture. Presentation 

 

Catherine Jones (FAO Regional Office for Asia and the Pacific) 

12:45-14:00 

 

Lunch Break 

14:00-14:30 

S2 

Monthly to seasonal forecast of Global Wind and Solar Power Generation Capacity  

Yunyun Liu (National Climate Centre of CMA) 

14:30-15:00 

S2 

Integration of S2S predictions in early warning systems for climate-sensitive health risks. Presentation 

 

Chloe Fletcher  

(Barcelona Supercomputing Center) 

15:00-15:30 

S1 

Developing data-driven S2S prediction at ECMWF. Presentation 

 

Steffen Tietsche (ECMWF)  

15:30-16:00 

 

Coffee Break 

16:00-16:30 

S1 

Using interpretable gradient-boosted decision-tree ensembles to uncover novel dynamical relationships governing monsoon low-pressure systems. Presentation 

 

Kieran Hunt (University of Reading)  

16:30-17:00 

S2 

Statistical and AI/ML methods to improve subseasonal to seasonal (S2S) wind speeds over India. Presentation 

 

Aheli Das (University of Reading)  

17:00-17:45 

 

Panel Discussion: Gap and advances in Science to Application (In-person) 

  • Emma Hudson-Doyle, Massey University, New Zealand  Presentation 

  • Victor Marchezini, Cemaden, Brazil 

  • Ziqiang Han, Shandong University, China 

Closure 

 

 

Looking Ahead 

As science and technology evolve, the workshop emphasised the importance of international collaboration—underpinned by the SAGE Project and WG TMR—to ensure that AI-enabled S2S forecasts and services contribute meaningfully to the "Early Warnings for All" initiative. Continued dialogue across disciplines will be key to unlocking the full potential of S2S predictions in a changing climate.