Disputas - Master of Science ​​Samuel Kočiščák​

It is apparent from remote observations that the density of interplanetary dust is the highest in the near-solar region. Yet, there were little in-situ data available from measurements from within 1 AU before Parker Solar Probe (PSP) and Solar Orbiter (SolO) missions. They both venture close to the Sun on elliptical orbits, and they are equipped with electrical antennas capable of dust detection, but the detection is ambiguous, since the dust impacts leave intermittent and difficult to identify signals in the time-series. In this thesis, a novel convolutional neural network method for dust identification is presented and applied to SolO data, yielding the highest quality SolO dust data product to date. The data is analyzed statistically, in the Bayesian framework, using Integrated Nested Laplace Approximation, which allows for usage of more sophisticated models. Important characteristics of the dust cloud are revealed this way, such as the deceleration of the hyperbolic dust component. Dust detection with electrical antennas is complicated, and steps towards the full understanding of the process are taken as a part of this thesis. By closely inspecting the signals recorded by SolO, the importance of photoelectron sheath for the antenna measurements is investigated, and it is likely that the photoelectron sheath makes the antennas a lot more sensitive to the presence of free ions. Meanwhile, the data of PSP is compared to a phase-space distribution flux model and to the measurements of SolO. This way, an indication of a dust depletion zone is found in the PSP data. A lot remains to be done before the interplanetary dust environment is fully understood, but SolO and PSP are going to continue providing more unprecedented data, analysis of which is made easier and more rigorous by the tools developed in this thesis.