We are currently seeking candidates interested in applying machine learning and Bayesian inference to the Space Weather problem. Positions are open both at doctoral (PhD) and post-doctoral level. This project is funded through a 5-years NWO-VIDI grant. Applications will be accepted until the positions are filled. Application screening will begin on September 15th. For more info see: PhD position Postdoc position Synopsis of Research Project Satellites orbiting around Earth have become crucial for communication and a wide range of technology/sensing applications. They are, however, often exposed to harmful energetic electrons that can damage their electronic instruments, making them temporarily or permanently inoperable, with large economic and societal consequences. The physics behind the rapid energization of electrons up to a few MeV (“killer electrons”) in the Earth’s radiation belts is an open and debated topic in the Space Physics community. The standard way of approaching this problem is by means of quasi-linear diffusion simulations, that solve the non-adiabatic time evolution of the electron flux in the belts. In recent years the Space Weather community has been interested in the possible use of physics-based radiation belt simulations for the forecasting of relativistic electron fluxes. There are, however, a number of issues that make such predictions not very reliable, and therefore not yet up to the accuracy required to foster the interest of satellites operators. The aim of this project is to advance our forecasting capability of killer electrons by enhancing physics-based model with a new data-driven probabilistic framework. In particular, we will: (i) recast the reduced diffusion model as a probabilistic model (in a Bayesian framework) and carry out a data-driven estimation of input parameters; (ii) validate the radiation belt diffusion model against a first-principle model (based on the Vlasov equation) to determine when and why it is not usable; (iii) model and parametrize residual terms (non-diffusive), for cases when the diffusion assumptions do not hold; (iv) evaluate the performance of the new probabilistic model to predicting fluxes on a specific satellite orbit. The ultimate outcome will be a robust method and model that allows the forecasting of killer electrons along the satellites orbit. The code will be open-source and made widely available for the community.
If you are interested on working for your Master Thesis in our group, please contact us.