Science blog

Photo of clouds with light blue leading to dark blue sky

Separating the signal from the noise

Separating forecast errors into information and noise reveals true predictive skill, improves how weather models are evaluated, and enables fairer comparisons between traditional numerical and emerging machine learning approaches.
AI Weather Quest blog banner

The AI Weather Quest: spotlighting the best of machine learning sub-seasonal prediction

At the midpoint of its first competitive year, the AI Weather Quest brings together a global community advancing machine learning methods for sub‑seasonal forecasting.
Forest fire

Build your own probability of fire model

The PoF-Toolbox is a lightweight, reproducible framework for building local machine learning probability of fire models using user-defined data, helping translate global fire signals into locally meaningful risk information.
Satellite image of a hurricane

The eye of the storm: how EarthCARE is helping to evaluate hurricane physics in ECMWF forecasts

EarthCARE’s overpass of Hurricane Humberto captured the first space-based view of vertical winds and precipitation inside an eyewall, offering unprecedented insight into storm structure and updrafts that will help ECMWF improve hurricane understanding, intensity prediction and forecasts.
Blue banner with swirls and a cloud graphic on

Simplifying AI for weather forecasting with the European Weather Cloud

Rising demand for AI and ML workloads on the EWC has driven ECMWF to develop ready-made templates that democratise access by simplifying setup and lowering technical barriers. Available through the EWC Community Hub, they help users focus on science over infrastructure.
A green landscape in Indonesia with a dark grey sky above

Forecasting of convective weather in Indonesia

Ruth Mahubessy explores how Indonesia’s complex weather patterns challenge forecasters and how a new ECMWF–WMO Fellowship project aims to improve predictions of extreme rainfall and landspouts in the tropics.
Rain pouring down on a rainforest in Ecuador

Evaluating rainfall forecasts in Ecuador: Observations from AI and physical models

WMO Fellow Angela Iza Wong explores the challenge of forecasting rainfall in Ecuador.
blog banner crop from new sub-seasonal chart

Getting the most out of ECMWF sub-seasonal forecasts

Tim Hewson introduces a new type of forecast map which enables users to visualise the information content of sub-seasonal forecasts like never before.
AI Weather Quest blog banner

The AI Weather Quest: uniting international expertise to advance sub-seasonal forecasting

As the competition phase of the AI Weather Quest starts, the organisers introduce the competition framework and establish a baseline by assessing the skill of current dynamical sub-seasonal forecast models.
ESA Arctic Weather Satellite

Small-but-mighty Arctic Weather Satellite now assimilated at ECMWF

ECMWF has started operational assimilation of data from a trail-blazing small satellite launched by the European Space Agency (ESA) that provides new microwave sounding measurements from space. The new observations give a robust forecast improvement, complementing similar sensors from larger platforms.
Los Angeles fires January 2025

2025 California wildfires: insights from ECMWF forecasts

On 7 January 2025, a series of devastating wildfires started in Los Angeles. ECMWF, which has been producing fire danger forecasts since 2018, has been exploring new approaches to fire forecasting in recent years.
Flooded rice plantation in the Pacific coast region near Guayaquil, Ecuador, in 2017

Forecasting extreme events in Ecuador

The challenge of forecasting extreme events across Ecuador is the focus for Angela Iza Wong, from the Ecuador Hydrometeorological Service, during her year at ECMWF as a WMO Fellow.
ESA-ECMWF machine learning workshop, card

Reflections on the 4th ESA-ECMWF Workshop on Machine Learning for Earth System Observation and Prediction

As one of the organisers of the joint ESA-ECMWF Workshop, Massimo Bonavita discusses some of the highlights and considers the future role of machine learning for weather and climate prediction.
Excerpt from figure 1 in the article showing different coloured squares denoting number of fires forecast or observed

Machine learning ignites wildfire forecasting

Joe McNorton reports on recent success in forecasting wildfires using machine learning at resolutions as high as 1 km and work to incorporate a wildfire component into ECMWF’s Integrated Forecasting System.
Card image, hot sun

Heat stress and the European heatwave of 2023

Claudia Di Napoli looks back at the extreme heat in parts of Europe this summer and describes her work to develop forecasts of heat stress and other weather-related health hazards.
2m temperature forecast 18 July 12UTC

European heatwave July 2023

Europe is experiencing some of the hottest temperatures of summer 2023 so far, with parts of Greece, eastern Spain, Sardinia, Sicily and southern Italy seeing temperatures above 45°C at the start of this week.
Machine learning card (Getty)

The rise of machine learning in weather forecasting

ML-based weather prediction models have developed rapidly over the last year with exciting results. A group of our scientists discuss developments and their potential implications for the future.
Blog card, cloud-clearing waves

The strange phenomenon of large-scale cloud-clearing waves in the atmosphere

While looking at kilometre-scale simulations, Philippe Lopez and Inna Polichtchouk spotted unusual cloud-free areas in the tropics which also appeared in observations. Here they describe their investigation of this large-scale cloud-clearing wave phenomenon.
International Women's Day

One woman's journey to a career in Earth system modelling

To mark International Women’s Day, Birgit Sützl, an early career research scientist at ECMWF, shares her experiences of studying maths at university and her journey to find her perfect career path in Earth system modelling.