Successful third Crash Course on Data Assimilation
In the beginning of June, NERSC researchers ran their third Data Assimilation summer school, together with NORCE colleagues, and this time digitally. Over 70 people participated and learned both about the theory of data assimilation and its applications!
Data Assimilation – what is that and what is it used for?
Data assimilation is the scientific field of combining real-world observations, such as measured ocean temperature, with theoretical structures (a numerical model), respectively an ocean model. This combination relies on not-so-complex statistical processes. The goal of assimilating data is to produce the best combination of both components to be able to reliably estimate and even predict changes in a system, such as the ocean, over time.
Data assimilation is relevant for several fields in science. It has been originally developed to improve weather prediction. Other fields where data assimilation can help to improve our knowledge are for example climate prediction, ecosystem investigations, and hydrological forecasting... and last but not least forecasting the COVID-19 epidemic. These applications were all covered in the summer school.
A distant cousin of data assimilation is machine learning, a relatively new research field compared to the prior. Machine learning also uses similar mathematical processes to produce predictions, but with a focus to learn from the real-world data more than satisfying theoretical considerations.
Combining both data assimilation and machine learning is a relatively new concept with very promising perspectives. For the first time, machine learning is included in this year’s summer school curriculum!
Screenshot of several of the participants on Zoom during the all-digital summer school, taken by Laurent Bertino.
Summer school organization
The summer school went on for five full days, all digital on Zoom and for free. By now, over one year into the pandemic, everyone was comfortable with that solution. NERSC and NORCE researchers organized the summer school in the framework of two Research Council of Norway-projects, DIGIRES and REDDA, and supported by the research school CHESS.
The aims of the summer school this year were to introduce PhD candidates and other early career scientists to the basic concepts of data assimilation, to familiarize them with ensemble methods (a popular type of data assimilation techniques), to introduce operational applications where data assimilation is used, and to educate on machine learning and its relevance for data assimilation.
The organizers planned five days of course work with a mix of short student presentations, lectures by ten researchers, six of them NERSC-affiliated, and three tutorials on applying what the participants have learned in the lectures. The following scientists were teaching in the DA crash course: Laurent Bertino (DA group), Yue (Michael) Ying (DA group), Julien Brajard (DA group), François Counillon (CDP group), Annette Samuelsen (OM group), Geir Evensen (NORCE, NERSC), Patrick Raanes (NORCE), Alberto Carrassi (University of Reading, UK), Rosella Arcucci (Imperial College London, UK), and Mohamad el Gharamti (National Center for Atmospheric Research, USA).
Outcomes
Around 70 participants joined the course from all over the world and they actively participated via an anonymized Questions & Answers page during lectures and tutorials. This enabled lively discussions around the current topics, and resulted in a successful summer school implementation, even in these trying times. Being flexible and going with the digital flow has proven useful once again!
Now, the PhD candidates and other early career scientists who joined the summer school know more about data assimilation than they did before the course started. Hopefully they can build on their newly acquired knowledge and use it for their research.
All Zoom sessions have been recorded, with the participants’ permission, and are available on YouTube for everyone to follow!
Link to presentations:
Link to the YouTube Playlist containing all lectures:
https://www.youtube.com/playlist?list=PLmn0ffSg-S8pKdJObkh_cshevEs58dqr8
Add comment