Abstract
In high energy astrophysics, several questions are still open and amongst them is the origin of Galactic cosmic rays. The first observational evidence of accelerated particles has only been revealed twenty years ago. Since then, supernova remnants, pulsars and their nebulae are known to efficiently accelerate particles but several questions still hold. In particular, evidence of accelerated protons (which consist on 90% of the cosmic-ray spectrum) is still elusive and several gamma-ray sources have unknown origin. Non-thermal emissions produced in Galactic accelerators provide insights about the nature of the accelerated particles. In particular, while radio and X-ray observations indicate the presence of accelerated electrons, gamma rays can be produced by both electrons and protons (or nuclei in general) but it may be difficult to assess the origin of the emission.The technical part of this thesis concerns the study of the impact on the reconstructed H.E.S.S. data when using atmospheric profiles measured with a lidar instead of a standard atmospheric model currently used. Very high energy gamma rays propagate into the atmosphere before reaching Cherenkov Telescopes and thus, the accuracy of our measurements depends on our understanding of the atmospheric composition. Using run-wise simulations based on lidar data, we study the impact on the instrument response functions and we analyse data to quantify the effect on the spectral reconstruction.The first scientific goal of this thesis is to understand the gamma-ray emission from two supernova remnants (G326.3-1.8 et RX J1713-3946) through detailed spectro-morphological analyses. The analysis of the composite supernova remnant G326.3-1.8, with Fermi-LAT data, has led to two major results: a new evidence of accelerated protons and the first morphological and spectral separation in gamma rays of two nested components. The study of RX J1713-3946, with H.E.S.S. data and using the Ctools package, confirms a significant gamma-ray extension beyond the X-ray emitting shell but its origin remains unclear.The second part of this thesis aims to constrain the nature of the unidentified TeV sources revealed in the H.E.S.S. Galactic Plane Survey. We thus present a generic code, based on a multi-wavelength approach, to find counterparts and estimate physical parameters like the radio spectral index and the mean magnetic field. We apply this code on five unidentified TeV sources and we put constraints on their nature. In particular, we present two high-confusion cases, for which the TeV emission is probably due to the contribution from different components, emphasizing the importance of multi-wavelength data to understand the origin of the gamma-ray emission.