27 August 2021
on course website
Data Science and Beyond: Data Assimilation with Elements of Machine Learning
How do meteorologists forecast the weather and climate? Is there a way to predict the profit from a wind farm? These are some of the questions modern science addresses by using data assimilation. Many research institutes and companies (e.g. KNMI, Shell, US-NCAR or UK MetOffice) develop and employ data assimilation and the demand for trained personnel is constantly growing. The school will describe the theoretical foundation of data assimilation together with numerical tutorials, all the way to state-of-the-art methods, including modern machine learning approaches and their combination with data assimilation.
Data Science and Beyond: Data Assimilation with Elements of Machine Learning logo
Data assimilation is the science of combining measurement data and computational models. It encompasses a large portfolio of methods at the crossroad between numerical analysis, linear algebra, statistics, dynamical systems and optimal control. Data assimilation is crucial in all circumstances where one wishes to make sense of a model against data and is therefore ubiquitous in science and in real life applications.
The summer school aims at covering the mathematical foundations of data assimilation and at describing the existing methods up to the advanced approaches currently being developed. In particular, the school will address variational and ensemble methods, nonlinear Bayesian techniques for high-dimensional systems and the modern hybrid approaches emerging from the cross-fertilization of data assimilation and machine learning.
Together with overview and theory lectures the school will also provide tutorials with numerical exercises using the Jupyter notebook platform where the students can actively practice what they are learning.
The interdisciplinary character of the discipline, together with the broad class of scientific areas where data assimilation is used (climate science, neuroscience, biology, medicine, traffic control, energy production and power grid management, just to mention a few) makes the school a unique opportunity for students with very diverse backgrounds, such as mathematics, physics, environmental science or biology, and it suits ideally for the students of both the Master’s in Mathematical Sciences and in Climate Physics of the University of Utrecht.
A limited number of scholarships, in the form of course fee waivers will be offered to students coming from low income countries (see scholarships below).
If the course cannot take place face-to-face we will turn to a remote online version with lectures and tutorials streamed and recorded.
Prof. dr. N.A. (Alberto) Carrassi
The school welcomes students from a very broad portfolio of backgrounds. These include, but are not limited to, mathematics, physics, climate science, biology or neuroscience. The courses are primarily designed for students at level of advanced Master’s or PhD candidates but students at an early stage, with sufficient mathematical background, are also suitable as well as Postdoc or more senior scientists interested in data assimilation for their research.
The school aims to cover a broad spectrum of modern data assimilation methods: Kalman filters and smoothers and their first order nonlinear extension, ensemble approaches, variational methods, hybrid ensemble-variational, nonlinear particle filters and hybrid data assimilation and machine learning methods. It will also provide notable examples of applications of data assimilation to large climate systems such global ocean and sea-ice models. Lectures alternate with exercise sessions. The full detailed programme will become available soon.
EUR 230: Course + course materials
EUR 200: Housing fee (optional)
A limited number of scholarships in the form of fee waivers is available for students coming from low income countries. There are no travel grants or grants for housing costs. If you would like to apply for the fee waiver, please mention this.Register for this course
on course website