Weathernews Joins "Using ECMWF's Forecasts 2025" — Where Europe's Cutting-Edge AI Weather Technology Meets Global Innovation

From September 15 to 18, two forecasters from our Weathernews Forecast Center—Kohei Sakamoto and Jumpei Fujino—traveled to Bologna, Italy, to participate in the international conference "Using ECMWF's Forecasts 2025 (UEF2025)," and gave a presentation. In this article, Sakamoto and Fujino share the key insights they gained from this experience and explore how what they learned will shape our future approach to weather forecasting.

What is UEF2025?

The Using ECMWF’s Forecasts (UEF) is an annual event organized by the European Centre for Medium-Range Weather Forecasts (ECMWF), one of the world's leading institutions in numerical weather prediction technology.

This year’s conference took place in Bologna, Italy, home to ECMWF’s cutting edge data center.

The event provides a valuable opportunity for ECMWF developers and users from around the world to come together — developers share upcoming development plans, users provide feedback, and participants exchange real-world use cases and experiences.

This year was particularly special, as ECMWF celebrated its 50th anniversary, with UEF2025 held as part of a series of commemorative events.

Purpose of Participation

Our primary goal in attending UEF2025 was to learn about the latest global trends in AI-based weather forecasting models.

As Weathernews continues to expand our global presence, it's crucial that we understand what leading public institutions and private companies are doing in this space—especially ECMWF, which is pioneering much of the innovation we see today. These insights directly inform how we develop our own forecasting technology and shape our service strategies worldwide.

We also wanted to strengthen our working relationships with the ECMWF developers who are spearheading AI innovation in weather prediction. Since we already use several of ECMWF's models and tools in our daily operations, having direct lines of communication makes it much easier to collaborate on technical challenges and explore potential improvements together.

On a personal level, it was genuinely inspiring for us as meteorologists to connect face-to-face with fellow engineers who are working on similar AI forecasting challenges. There's something invaluable about exchanging ideas and experiences with peers from around the world who share the same passion for advancing weather prediction technology.

Weathernews’ Presentation and On-site Feedback

Our presentation focused on a case study evaluating ECMWF’s AI forecasting model “AIFS” and how it performed in predicting extreme weather events in Japan. We focused on two particularly significant cases from this past summer: • The widespread frontal heavy rainfall that impacted large areas of Japan in mid-August 2025 • The record-breaking temperature of 41.8°C recorded in Isesaki, Gunma Prefecture, on August 5, 2025 Results showed that while AIFS produced generally reasonable forecasts, it underestimated the intensity of extreme events, particularly heavy rainfall — highlighting the need for higher-resolution AI forecasting models.

Specifically, for the rainfall event, AIFS successfully predicted the geographical location of the precipitation peak but significantly underestimated rainfall amounts — around 160mm of rainfall over 24 hours, while actual observations exceeded 400mm in some locations. For extreme heat events, AIFS predicted temperatures would stay below 40°C throughout the region, falling short of the record-breaking temperature of 41.8°C that we actually observed. You can find our full presentation materials and video on the ECMWF official website.

Kohei Sakamoto and Junpei Fujino from the Weathernews Forecast Center giving his presentation (Photo: ECMWF official website)

The response from conference participants was incredibly encouraging. Many attendees came up to us afterward with comments like "Your presentation was fascinating!" They were particularly interested in seeing AIFS evaluated specifically for the Japan region, which is not typically featured in other research studies.

Insights Gained and Future Outlook

Participating in UEF2025 was an extremely stimulating and rewarding experience for our team.

One of the most striking takeaways was seeing just how rapidly European meteorological organizations are embracing and implementing AI technologies—the pace of adoption there is noticeably faster than what we're seeing in Japan. While Weathernews is already making significant strides in developing and applying our own AI-based forecasting models, this conference gave us fresh motivation to accelerate our efforts even further and approach innovation with greater urgency and creativity.

Looking ahead, we're committed to continuing our participation in global forums like UEF. These platforms allow us to showcase Weathernews' technological capabilities to the international community while also contributing to the collective advancement of weather forecasting technology worldwide. It's this kind of collaborative exchange that drives the entire industry forward.

A Few Travel Notes

This trip also gave us a wonderful opportunity to experience different weather and climate conditions firsthand. Bologna felt noticeably more comfortable than summer in Tokyo—slightly cooler and much less humid, so even when the sun was strong, it remained pleasant.

What caught our attention as meteorologists was how daylight saving time shifted the timing of solar noon and peak temperatures compared to what we're used to in Japan. After the conference, we traveled to London for meetings with our partners there. Our colleagues mentioned how unusually beautiful the weather was during our visit, which gave us a pretty good hint about what London's weather is typically like!