The trajectories and stability of boundary currents, of mesoscale vortices, and of recirculations, are often largely imposed by ocean bottom topography. Here several related questions in the influence of topography on mesoscale ocean circulation are investigated, largely motivated by observed circulation features in the sub-polar North Atlantic ocean. Observations show that boundary currents tend to become highly variable and shed material near sharp topographic variations, such as peninsula edges or corners of underwater capes. Baroclinic instability is understood to be one of the main causes of internal variability of large scale ocean circulation. Therefore the influence of horizontally curving topography on baroclinic instability is studied, under the hypothesis that the curvature may cause a higher tendency towards instability. That is done within a minimum complexity model, a two-layer quasi-geostrophic model, and compared with the classic rectilinear model. First necessary conditions for instability as well as growth rate bounds are derived. Growth rates are calculated analytically or numerically for several flow and topography profiles. The growth rate in uniform azimuthal flow is similar to that in uniform rectilinear azimuthal flow, but decreases with increasing depth-averaged flow component amplitude. That is recognized as a generalization of the so called "barotropic governor" effect. Instability growth rate is nonetheless higher with uniform azimuthal flow when isopycnal slope is similar to the topographic slope magnitude, a common scenario in the ocean. Non-normal instability is studied as well, and is generally intensified with uniform azimuthal flow. Thus a complex picture emerges as to the influence of horizontal curvature on baroclinic instability. The Deep Western Boundary Current (DWBC) carries water masses formed in deep convection sites southward, as part of the Atlantic Overturning Meridional Circulation (AMOC), a circulation pattern of climatic importance. Observations show that the DWBC "leaks" material at an anomalously high rate in its path along two underwater capes in the Newfoundland Basin. The leakiness, resulting in water masses dilution, and in AMOC alternative (interior) pathways southward, has not been studied extensively from a dynamical perspective before. A high-resolution realistic regional numerical model configuration and a particle advection model are developed for this purpose. The numerical results, as well as two datasets of ocean float trajectories, are analyzed to determine the dynamical causes of leakiness and its phenomenology. It is found that leakiness is concentrated in three "hotspots", in which topography turns and steepens. Mean Lagrangian velocity is offshore at these locations, showing that leakiness occurs by mean separation. The mean velocity does not have a substantial eddy-rectified component at the two northern hotspots, where most of the mean leakiness happens. Likewise, energetic analysis shows eddies do not locally force the mean offshore flow. Furthermore, potential vorticity is not diluted substantially by eddies along mean separating streamlines. These results are consistent with mean leakiness occurring by inertial separation. A scaling analysis also suggests that bathymetric conditions near the leakiness hotspots are supportive of inertial separation. Eddy processes also contribute substantially to leakiness, partially through chaotic advection. In several North Atlantic basins semi-stationary anticyclonic vortices (ACs) have been repeatedly observed for decades, within areas with bowl-like topography. These basins play significant parts in AMOC transport and transformations, and previous evidence suggests these ACs contribute to these processes. Therefore the formation processes of ACs above topographic bowls is studied here using idealized free evolution simulations in one or two isopycnal layers. It is demonstrated that ACs readily form under different (bowl-like) topographies and initial conditions. A non-dimensional nonlinearity parameter (epsilon ~ ratio of vorticity to bowl PV gradient), or a potential vorticity (PV) inhomogeneity (PVI) parameter, largely determine if a trapped AC is formed from random mesoscale-like initial conditions. Trapped ACs form and stay close to bowl-center for epsilon ~0.5 (PVI ~ 1). For epsilon ~ 1 (PVI ~ 0) vortices freely cross the topography by mutual interactions. For intermediate epsilon or PVI values, trapped ACs can form at different bowl radii since the PV gradient is nullified by the presence of a slope current. Trapped ACs generally form by repeated mergers of ACs within the bowl, and have anomalously low PV. Tracer analysis shows that ACs which eventually merge into the trapped AC are sourced from within (outside) the bowl in low (high) energy cases. Two different cross-bowl propagation mechanisms are examined. Monopole beta drift as well as dipole self propagation can both contribute to cross-bowl AC material transport, but the latter appears faster in relevant cases. The vertical structure of the trapped AC is studied as well. It is shown that it is top (bottom) intensified for top (bottom) intensified domain-mean initial conditions. That is consistent with observational structure but in contrast with the common vertical structure in Taylor Caps and of the slope current in our simulations, which remain bottom-intensified in all cases. Scaling laws for vertical structures are suggested in several cases. The robustness of AC formation to topographic complexity is studied, as well as its long-term evolution, and the results are contrasted with topographic turbulence theories, which predict a slope current but not a bowl-trapped AC.