These days abundant data is being gathered all over the world, ranging from weather stations to satellite images. It is vitally important to use this information in order to improve our knowledge about natural processes. However, what is the smartest way to combine the data and a model in order to decrease model uncertainties? This involves data assimilation. Dubinkina’s work is to develop such data assimilation methods.
Dubinkina says that model uncertainty will always remain because a computer model is still a bits representation. In other words, everything is converted into numerical points. Increasing the number of points allows a more precise solution to be produced. However, a different problem then arises: the computational time becomes longer, and it sometimes takes weeks before a computer model is finished with calculations. Dubinkina: 'That is an issue when a prediction, such as the weather forecast, must be available on time.'
With PhD researcher Bart de Leeuw, Dubinkina develops a data assimilation method for a simple weather forecast model. Meteorologists base their prediction on observations, for example from weather stations. However, as they cannot precisely measure what happens between weather stations, they use models. A small error in the interpretation of the initial conditions can culminate in wrong predictions. This is called the "butterfly effect", the classical explanation of chaos. The hope was that with a growing number of satellites, more observations would lead to more accurate predictions. 'However, it turned out not to be that simple', says Dubinkina. 'Mathematical challenges emerged: with more data, the original data simulation methods no longer worked. In our new method, we can use more data. The ultimate goal is to one day present the best working model to the Royal Netherlands Meteorological Institute (KNMI), but we still have a long way to go.’
Models for crude oil reservoirs
Dubinkina’s models are also used to explain assessments of energy-related problems, such as the structures of gas and oil reservoirs. Together with PhD researcher Sangeetika Ruchi, Dubinkina predicts when water-break in oil extractions occurs, thus mainly water instead of oil comes up. Dubinkina: ‘As you cannot drill soil everywhere, you must make assumptions about ground layers, which contain many errors. Researchers correct these errors by using computer calculations, that take a long time, or by making incorrect assumptions about the errors. My input in this specific research area is that I can correct the errors using a new data assimilation method that is fast and makes no assumptions about the errors.’
Oil and gas pipeline flows
A study of flows in oil and gas pipelines is planned for 2019. In pipelines slugs occur, which are isolated bubbles of fluid that can cause damage. Oil companies want to be able to estimate when these slugs arise. Dubinkina: ‘It takes a long time to determine that using a 3D computational model, but an existing 1D computational model fails to detect the slugs. Together with CWI colleague Benjamin Sanderse, I will develop a 1D computational model combined with data assimilation that can predict slugs formation.’
After 13 years, Dubinkina still enjoys working at CWI. ‘CWI gives researchers the time to write research proposals, and as a consequence to prosper their research ideas.’ Nevertheless, she would also like to hold a university appointment because then it would be easier to join in well-established collaborations between universities and institutions such as Medical Center, for example. Moreover, she also enjoys teaching and finds the exchange with students inspiring. Dubinkina: ‘I have taught at Utrecht University, one day per week, and I will teach there again in 2019.’ Her dream is to become a professor. As a role model, she wants to show other women that though challenging it is possible to combine a career in academia with family life. Dubinkina: ‘Women sometimes leave academia if they have to choose between a postdoc abroad and a family. It is a pity. Mathematics needs more female mathematicians because men and women view problems in different ways.’ Since September 2017, Dubinkina has also been a member of the CWI works council.
Dubinkina (36) is originally from Siberia. After her degree in applied mathematics in fluid dynamics at the Novosibirsk State University in Russia, a friend drew her attention to the possibilities in the Netherlands. ‘It was difficult to make ends meet as a PhD student in Russia.’ So in 2005, she went to CWI and gained her doctorate under Jason Frank and Jan Verwer supervision. In 2010, she pursued a postdoc at the University of Louvain. Four years later she returned to CWI for a tenure-track appointment.
NWO Institute CWI is the national research institute for mathematics and computer science and is located at the Science Park in Amsterdam. CWI has more than 200 employees. With new insights, these mathematicians and computer scientists contribute solutions to a wide range of areas including energy, healthcare, climate, communication, mobility and security.
Newsletter Inside NWO-I, October 2018