Introducing Code for Earth 2023 Selected Teams (Part 1 of 2)

At Code for Earth, our mission is to foster collaboration and support advancements in weather and climate research and initiatives like Copernicus and Destination Earth (DestinE). Since its first edition in 2018, our programme has brought together talented developer teams and experienced mentors from ECMWF and partner organisations to work on cutting-edge projects covering a wide range of areas within the ECMWF's scope of activities. These include data science, visualisation, machine learning, weather and climate analysis,atmosphere, and various other Earth sciences domains. By encouraging multidisciplinary collaboration and embracing open source principles, our program aims to facilitate the development of cutting-edge solutions and advancements in the field of Earth sciences.

In this context, up to ten developer teams join us every summer to explore a broad range of topics within the ECMWF's realm of activities. Code for Earth provides a platform for multidisciplinary collaboration, from data science and weather analysis to climate research and visualisation. Our participants bring their passion for the Earth sciences, expertise, and creativity to tackle complex challenges.

This year ten selected developer teams are at work to contribute their expertise, creativity, and problem-solving skills to address challenges in weather, climate, and related domains. Below is a short description of the selected projects:

Fire Forecasting: Enhancing Wildfire Prediction with Machine Learning focuses on developing a framework for forecasting European wildfires using machine learning techniques. By leveraging GFAS fire data and meteorological forecasts, the team aims to evaluate different machine learning tools for forecasting. The ultimate goal is to integrate these tools into the operational pipeline of the ECMWF, thereby improving wildfire prediction and response.

TropiDash: A Comprehensive Tropical Cyclone Hazard Dashboard aims to contribute to tropical cyclone data dissemination, enabling better population preparedness and resilience against extreme meteorological events. This project develops a platform on Jupyter Notebook that visualises key meteorological parameters in plots and maps. By applying interactive elements, TropiDash enhances the understanding of tropical cyclone hazards evolution. Furthermore, including sound documentation enables users to utilise and enhance the platform beyond the project's duration.

Sketchbook Earth: Democratising Climate Intelligence seeks to democratise the production of climate intelligence reports, traditionally limited by reliance on internal ECMWF tools. The project works on the development of Jupyter notebooks that illustrate our planet's climate stories in an accessible and engaging manner. Leveraging the cads-toolbox Python package and Copernicus climate data store, these notebooks transform raw data into expressive visual narratives. Sketchbook Earth offers a more visual, comprehensible, and reproducible approach to climate intelligence by providing meaningful climate insights and comprehensive training resources.

TesseRugged: Enhancing Spatial Resolution and Applicability of Reanalysis Data aims to enhance the spatial resolution of global reanalysis data sets, such as ERA5, for regions with heterogeneous terrain or renewable energy applications. By implementing model output approaches and deep learning techniques for post-processing and downscaling, TesseRugged utilises the higher-resolution CERRA data as a target/proxy. This step-wise approach allows for increased spatial resolution globally without significant computational resources, benefiting various applications such as renewable energy, agriculture, and climate control simulations.

Diffusion Models on WeatherBench: Exploring Machine Learning for Weather Forecasting focuses on exploring the potential of machine learning methods, particularly Diffusion Models, for weather forecasting using the WeatherBench benchmark dataset. By training the models to predict realistic future states based on current atmospheric variables, the project aims to advance the field of weather forecasting. The publication of code and trained models encourage replication and further development of the results.

Code for Earth is an innovation programme run by the European Centre for Medium-Range Weather Forecasts (ECMWF) with contributions by Copernicus, Destination Earth, the European Weather Cloud and WEkEO.