Low oxygen concentrations remain a global concern for the ecological health of lakes. High nutrient inputs and climate-induced changes in stratification and mixing are anthropogenic threats which largely impact aquatic oxygen budgets and overall ecosystem health. In this thesis, the relevant processes for hypolimnetic oxygen depletion were investigated on different temporal and spatial scales in the deep perialpine Lake Geneva and the corresponding total oxygen budget was estimated. The short-term variability of sediment oxygen uptake (SOU) and its dependency on the bottom boundary layer currents were investigated using microprofile measurements. Sediment core analyses for reduced substances profiles allowed distinguishing between SOU caused by both oxic respiration and the flux of reduced substances out of the sediment. Long-term monitoring data were used to estimate the relative importance of SOU for the total oxygen depletion in the lake. Finally, one-dimensional numerical models were used to reproduce lake temperature and oxygen concentrations and to assess the impact of changing environments on the oxygen budget of the deep-water. The results of the microprofile measurements led to a new parameterization of turbulent diffusion close to the sediment and enabled a similarity scaling of diffusivity as well as oxygen close to the sediment. However, the comparison of microprofile measurements at different lake depths showed that SOU decreased consistently with depth from ~1 g m-2 d-1 at 40 m to ~0.2 g m-2 d-1 at 133 m independently from the small-scale variability due to hydrodynamic forcing. Similar vertical structures of SOU and total oxygen depletion have been found in other Swiss lakes. The decrease of SOU with depth was attributed to the greater amount of easily degradable organic matter available in the upper layers. The comparison between SOU and the reduced substances flux revealed that oxic respiration is by far the dominant pathway of organic matter mineralization. While the long-term monitoring data did not show a decreasing trend in either the areal hypolimnetic mineralization rate (1.34 g m-2 d-1) or the extent of hypoxia, a strong relationship between deep mixing in winter hypoxic conditions was found. Hence, deep-water oxygen concentrations were predominantly controlled by resupply during the unstratified period in winter. To assess the long-term changes of winter mixing in Lake Geneva, the one-dimensional model SIMSTRAT was used to reproduce lake temperature and stratification between 1981 and 2012 and was run afterwards under atmospheric conditions representative for the years 2045–2076 and 2070–2101, according to the IPPC scenario A1B. The simulations predicted (i) a decrease in winter mixing depth from an average of ~172 m to only ~127 m at the end of this century, and (ii) complete homogenization of temperature and oxygen in winter will decrease by ~50%. Hence, changes in mixing may have stronger impact than eutrophication on the deep-water oxygen. A simple oxygen model coupled to SIMSTRAT predicted an increase in hypoxic conditions in the deep part of Lake Geneva by ~25%. Additionally, a detailed oxygen model was developed based on the observational findings of this dissertation which takes the spatial variability of oxygen depletion and its dependency on lake turbulence into account. This model can be generalized to understand and predict climate-induced changes of future oxygen concentrations in other deep lak