Data Assimilation Training Course

We will deliver an intensive training course on Data Assimilation (DA), providing the environmental science community with the skills to exploit large observational data sets to enhance environmental prediction and understanding. DA provides a practical framework for combining models and observations using robust mathematical theory. It has been proven to be essential for forecast accuracy and enables the creation of reanalyses by reconstructing Earth system components from limited historical observations. DA is increasingly playing an important role in the explosion of machine learning activities, providing both a source of training data and a natural framework to interface data-driven and physically informed techniques for inferring the Earth system’s past, present and future. DA also underpins the use of observational data across the NERC remit for the development of digital twins, designed to tackle some of the most challenging environmental problems. The course will therefore fulfil the objectives of NERC’s Digital skills priority area by ‘supporting the development and use of new technologies such as digital twins and artificial intelligence’, helping to fulfil a national and international critical digital skills gap identified in recent scoping studies (DOI 10.5281/zenodo.11191276, https://www.gov.uk/government/publications/space-sector-skills-survey-2023/space-sector-skills-survey-2023-report). The course will target scientists working within the NERC remit who would benefit from integrating DA into their research. Since DA research and its use are rapidly evolving, we will focus on both early career scientists and those at later stages who need to update their skills. The course will be divided into two stages: foundational and advanced, to accommodate scientists with varying levels of confidence in mathematics and different desired outcomes from the course. The foundational stage will cover the basic theory of DA, practical algorithms, and its role in forecasting and reanalyses for various environmental applications. The advanced stage will expand on these concepts, delving into more mathematical theory and exploring the diverse ways that machine learning is being integrated into DA. The course will consist of interactive lectures for 30 students, including group discussions, anonymous polls, and breaks allowing for one-on-one discussions. The theoretical concepts covered in the lectures will be reinforced through computer practicals, offering additional opportunities for questions. The practicals will be conducted using a cutting-edge DA software package called JEDI (Joint Effort for Data Assimilation Integration), which is being adopted by operational centres like the Met Office and supports academic research. By utilizing the JEDI software, students will have the skills to access the latest DA developments even after completing the course. At the end of the course, no matter if they have attended only one or both parts, the participants will have increased confidence in how to incorporate DA into their own research. They will be provided with a cache of resources to allow them to continue their studies independently and will have access to a network of potential collaborators. The material developed will form the foundation for a future annual training course, allowing DA to continue to address new environmental challenges and drive the ambitious development of digital twins.