Code for Earth returned in 2025 with ten exciting projects that united developers, scientists and innovators internationally to create open-source solutions addressing today’s most pressing environmental and climate challenges.
Organised by ECMWF, including support from the Copernicus Climate Change Service (C3S), the Copernicus Atmosphere Monitoring Service (CAMS) and Destination Earth (DestinE), the programme invites each year selected external teams to collaborate with expert mentors from both ECMWF and partner organisations. The aim is to deliver solutions that improve how we use, visualise and interpret Earth science data.
A programme built on strong partnerships
Collaboration is at the core of Code for Earth. Each year, the programme works with partner organisations – so-called 'challenge partners' - from academia, research and industry, on selected projects, some of which are co-mentored by the partners. These collaborations broaden the network of Code for Earth, fostering knowledge exchange and supporting meaningful innovation.
This year's edition involved four distinguished partner organisations, working with ECMWF experts to design and mentor three unique challenges:
International Foundation Big Data and Artificial Intelligence for Human Development (IFAB) (Italy)
Based in Bologna, IFAB is a non-profit foundation operating in the heart of Italy's 'Data Valley'. With strong links to high-performance computing, AI and sustainable innovation, IFAB acts as a bridge between research, technology and society.
Forschungszentrum Jülich (FZ Jülich) (Germany)
FZ Jülich is a German interdisciplinary research centre tackling societal challenges in energy, climate change and information. It is also a leader in supercomputing and Earth system science.
In this year's edition of Code for Earth, IFAB and FZ Jülich joined forces to co-mentor the Physics-Aware Consistency Evaluator (PACE), a tool designed to evaluate how realistic machine learning-based weather forecasts are, addressing the crucial question of whether fast, data-driven models for weather forecasting can remain physically consistent, ensuring reliable outputs for both scientific and operational use.
Helmholtz Centre Hereon (Helmholtz-Zentrum Hereon) (Germany)
Hereon is one of Germany's leading research centres in environmental and coastal systems, with expertise spanning atmospheric and aquatic pollution, climate change and machine learning applications.
Hereon co-mentored the 2025 project AI4AirQuality: High-Resolution Air Pollution Downscaling. The project explored advanced machine learning architectures to downscale CAMS global data into higher-resolution insights. By focusing on feature selection, interpretability and scalability, this work contributes to more accurate, actionable urban air quality information.
European Network of Transmission System Operators for Electricity (ENTSO-E) (Europe)
ENTSO-E represents 39 electricity transmission system operators from 35 European countries, working to ensure reliable and sustainable electricity across the continent.
ENTSO-E co-mentored the project Weather-Energy Analysis & Visualisation for Extremes (WEAVE). WEAVE focused on developing an interactive platform to visualise the impact of extreme weather events on energy systems, using the Pan-European Climate Database and Jupyter Notebooks.
These collaborations exemplify ECMWF's role as a hub for scientific partnerships at all levels – international, interdisciplinary, and open-source.
2025 Code for Earth projects at a glance
Here's a snapshot of this year's projects, spanning across three key streams – Data Visualisation, Machine Learning, and Software Development – each addressing challenges in their fields:
| AI4AirQuality: High-Resolution Air Pollution Downscaling | Focuses on refining air quality predictions via advanced ML-based downscaling techniques. |
| EarthReach Agent: Dual-LLM Framework for Validated Meteorological Chart Descriptions | Uses Large Language Models (LLMs) to generate validated textual descriptions of meteorological charts for visually impaired scientists. |
| EVALKIT: Model Error Detective | Produces tools to diagnose and visualise errors in modelling outputs. |
| Fire Front Radar | Designs an interactive web application for dynamically exploring fire-related data worldwide. |
| MLCosting | Produces an ML API plug-in that predicts Data Store System (DSS) request costs at ECMWF, improving efficiency. |
| OPEN DATA VISIO | Helps users access, download and analyse ECMWF real-time forecasts. |
| Physics-Aware Consistency Evaluator (PACE) | Develops metrics to assess how well ML models respect physical laws in meteorological predictions. |
| PolyShell | Offers a fast, efficient way to simplify complex polygonal data while preserving essential features. |
| Visualising CAMS Data in the Browser | Creates intuitive web-based tools for exploring Copernicus Atmosphere Monitoring Service atmospheric data. |
| Weather-Energy Analysis & Visualisation for Extremes (WEAVE) |
Offers visual tools linking weather extremes to energy sector impacts. |
More details on the projects, the participants, the mentoring teams and the programme in general are available on the Code for Earth website (https://codeforearth.ecmwf.int/projects/?_filter_by_date=2025).
Partnerships at the heart of ECMWF's mission
Code for Earth is more than an innovation programme - it is a demonstration of how ECMWF fosters impactful collaboration. By bridging expertise from multiple sectors in Earth sciences, ECMWF nurtures tools and solutions that benefit science, policy and society. The programme continues to embody ECMWF's vision of open, inclusive and future-ready scientific advancement.