PhD position (C3)
Predictability of tropical and hybrid cyclones over the North Atlantic Ocean
This PhD position contributes to project C3 "Predictability of tropical and hybrid cyclones over the North Atlantic Ocean".
The Institute of Computer Science at the Johannes Gutenberg University (JGU) in Mainz has a strong research profile in the data sciences and is involved in numerous interdisciplinary projects with the natural sciences. In particular, there is a long-standing cooperation with the Institute of Atmospheric Physics (IPA) which in turn has strong links to the Institute of Meteorology and Climate Research of the Karlsruhe Institute of Technology (KIT). JGU is the largest university of Rhineland-Palatinate covering a wide spectrum of disciplines and located close to the city center of one of the most livable cities in Germany.
The statistical prediction of the occurrence and development of tropical and tropical like cyclones is a challenging problem. In project C3 of W2W we will develop statistical-dynamical forecast models using information about relevant and better predictable large-scale atmospheric waves or objects. One possible approach to obtain this information consists in the identification of repetitive patterns in the spatio-temporal meteorological data. We will develop a general framework for a low-dimensional, geometry-based description of wave-related features that influence cyclone development and apply this concept to large re-analysis and ensemble data. The ultimate goal is an improved predictability of cyclones by using these features in a neural network approach and by explaining their roles in analyzed forecast jumps in the evolution of cyclones.
The project is designed as a twin project of JGU and KIT. The PhD candidate will be co-supervised by Michael Riemer (JGU) and Andreas Fink (KIT), both renowned experts in dynamic meteorology and weather forecasting. There will be a close cooperation between the PhD students at KIT and JGU. In joint and regular discussions, they will share their knowledge on the relevant meteorological data sets and phenomena as well as on the computational aspects of data analysis and machine learning.
The PhD student at JGU will use deep learning methods for the detection and tracking of atmospheric objects such as PV streamers, troughs, and vortices. She or he will set up and train neural networks to search very large data sets for wave phenomena which are relevant for the genesis and evolution of cyclones. She or he will investigate the extent to which this collected statistical information can improve the predictability of cyclones.
The ideal candidate holds a MSc in Computer Science with a minor in Meteorology or Physics and has already some experience in data science (statistics, data analysis, machine learning) and high performance computing. Besides very good knowledge of English, experience with machine learning libraries like PyTorch or TensorFlow would be an advantage.
The position is remunerated according to TV-L E13 (75%), it is initially funded for 3 years with possible extension until June 2023, and should start as soon as possible. Applicants are asked to specify their desired starting date and to give the contact details of two academic referees.
W2W features an innovative program for the development of early career researchers based on self-government. In addition to self-organized activities such as workshops, trainings and a guest program, the successful candidate will have the opportunity, if desired, to pursue an international research visit of at least one week. The consortium conducts an ambitious program to gradually enhance gender equality on all career levels within the academic fields combined in W2W.
JGU is an equal opportunity employer. Women are especially encouraged to apply. Applicants with disabilities will be preferentially considered if equally qualified. The successful candidate will work in Mainz and will have access to extensive training and career development activities offered by the JGU.
If you are interested, please apply online here!