Iron oxides are abundant in nature, cheap, and non-toxic, and are therefore frequently used as support materials for nanoparticle catalysts. In recent years, there has been a drive to reduce the nanoparticle size to the single-atom limit. This not only reduces the required amounts of expensive materials, but also improves the catalytic selectivity. However, the properties of supported single atoms depend strongly on the support material and the binding geometry, and so it is essential that we understand the atomic-scale structures of these support materials and their interactions with adatoms. This thesis presents results on four different iron oxide surfaces: The (001) and (111) facets of magnetite (Fe3O4), and the (1-102) and (0001) facets of hematite (α-Fe2O3). The four surfaces were evaluated in terms of their atomic-scale structure, their ability to stabilize single adatoms, and their suitability as model systems for single atom catalysis (SAC). The different facets were studied with surface science methods, including scanning tunnelling microscopy (STM), x-ray photoelectron spectroscopy (XPS), low energy ion scattering (LEIS), non-contact atomic force microscopy (ncAFM), temperature-programmed desorption (TPD), and quantitative low-energy electron diffraction (LEED), as well as computational modelling using density functional theory (DFT). Single crystal samples were studies in ultra-high vacuum (UHV), and their stability in ambient conditions was tested by UHV-compatible exposures to liquid and ambient pressure water. The magnetite (001) facet has been studied extensively in recent years, and its termination in UHV is known precisely. The surface stabilizes high coverages of single adatoms at room temperature, presenting an excellent template for a fundamental study of SAC mechanisms. After a brief review of the literature, results on the stability of this termination in ambient conditions are presented. When samples were exposed to liquid water for several minutes, the “subsurface cation vacancy” reconstruction of the sample was lifted, and a sub-monolayer coverage of an oxyhydroxide phase was formed. This affects the stabilization of single adatoms, which is based on the surface reconstruction. Results of the water exposure on nickel adatoms are presented, which show that the single adatom dispersion is not preserved in these conditions. Extensive literature also exists on the magnetite (111) facet, but is much more ambiguous than for Fe3O4(001). Several different terminations can coexist, but only one structure, the Fe-tet1 termination, can be prepared to cover the entire surface. Experimental and computational work on this Fe-tet1 structure are shown. Several possible modifications were found to be thermodynamically favourable at typical experimental conditions, which explains the abundance of defects found in experiment. Identification of the Fe-tet1 termination by STM or LEED was found to be unreliable, as all defectsare either aperiodic or highly mobile. Preliminary experiments were performed to test the stabilization of adatoms. Deposition of platinum yielded clusters at room temperatures, while low coverages of nickel appeared to be incorporated into the surface. Due to the ill-defined nature of the surface itself, however, the exact configurations of the adatom structures were not identified. The hematite (1-102) surface exhibits two different terminations, a stoichiometric (1×1) structure and a reduced (2×1) reconstruction. Both terminations can reliably be prepared in UHV to cover the entire surface. Extensive experiments were performed to determine the atomic-scale structures. The (1×1) termination was found to be a bulk-truncated structure, as confirmed by quantitative LEED measurements. A previously proposed model for the (2×1) termination was found to be incorrect, and a new model is presented here. Surface interactions with water were studied both in UHV and in ambient conditions. The (1×1) termination was stable under liquid water exposures, with equivalent results as for UHV-compatible water pressures. In contrast, the (2×1) reconstruction was changed to a (1×1) periodicity after exposure to liquid water. Titanium-doped films were grown to improve sample conductivity, and Ti-induced defects were studied experimentally and by DFT. Finally, the stabilization of rhodium adatoms was investigated on both terminations. In UHV, Rh sintered to clusters upon deposition at room temperature, but single rhodium adatoms could be stabilized on the α-Fe2O3(1-102)-(1×1) surface by depositing them in a background of water. A bulk-truncated termination has previously been reported for the hematite (0001) facet, and is generally assumed in theoretical modelling of single atom catalysis on hematite, but could not be reproduced here for clean samples. Instead, the “biphase” structure with (40±5) Å periodicity was found to be the most stable termination for both fully oxidized and slightly reduced surfaces, while a Fe3O4(111)-like termination is formed at more reducing conditions. Due to the high complexity of this surface structure, α-Fe2O3(0001) is likely not suitable as a model system for single atom catalysis, and deposition of adatoms was not attempted. Overall, the magnetite (001) and hematite (1-102) facets were found to be the most promising surfaces for studying single atom catalysis on iron oxide substrates. Extensive studies on Fe3O4(001) already exist, but its instability in ambient water pressures indicates that these results may not be transferable to real systems. The α-Fe2O3(1-102) facet exhibits a simple bulk-truncated termination that is stable in ambient conditions, and single adatoms could be stabilized by co-adsorbed water, making it a prime candidate for future studies.