Machine learning in meteorology
The Meteorology group at the Faculty of Mathematics and Physics, University of Ljubljana, Slovenia, is looking for an eligible candidate to fill a PhD student position on the topic of machine learning in meteorology. The work would be done under the supervision of asst. prof. Gregor Skok and dr. Žiga Zaplotnik. The position is available from 1 October 2020 for a duration of four years. The candidate must fulfill some formal requirements for the position. These include finished bachelor and master’s degrees, high enough grades during the undergraduate and graduate studies (at least 70% or 8/10 on Slovenia’s grading system) and age under 28 years (junior researcher).
A candidate must also demonstrate sufficient programming skills (preferably Python, however experience with Fortran/C++ would be an advantage), knowledge of data analysis, mathematical modeling and must have a good background in physics and mathematics. A candidate must be able to work effectively in English. Specific knowledge of machine-learning tools such as TensorFlow would be an advantage but is not essential.
The candidates should submit a single pdf that contains a motivation letter, CV with all relevant information such as education and a description of previous work experience and a copy of academic transcripts (for bachelor and master’s degrees) where the grades for all finished courses are listed.
Outline of the PhD research topic:
The candidate would presumably focus on the use of machine learning methods to investigate the statistical relationships between different meteorological fields at distinct lead-times/locations. The candidate would also seek to physically interpret the identified couplings that affect the performance of the long-term forecasts (e.g. using the layerwise relevance propagation method, which can be used for the analysis of neural networks).
The results could potentially be useful for improving long-term weather forecasts and for improving the applications of these forecasts (e.g. long-term renewable energy production forecasts). The results could be also used to improve balance operators in variational data assimilation.
Although the topic is meteorological, a candidate from a different natural science field is also suitable.