Unpredictable variations in the ocean originate from both external atmospheric forcing and chaotic processes internal to the ocean itself, and are a crucial sink of predictability on interdecadal timescales. In a global ocean model, we present i.) an optimization framework to compute the most efficient noise patterns to generate uncertainty and ii.) a computationally inexpensive, dynamical method for attributing sources of ocean uncertainty to internal (mesoscale eddy-driven) and external (atmospherically driven) origins, sidestepping the more typical ensemble approach. These two methods are then applied to a range of metrics (heat content, volume transport, and heat transport) and time averages (monthly, yearly, and decadal) in the subtropical and subpolar North Atlantic. The optimal noise patterns create variability in integrated quantities of interest through features of the underlying circulation such as the North Atlantic Current and deep water formation regions. Meanwhile, noise forcing diagnosed from model representations of the actual climate system stimulates these theoretical patterns with various degrees of efficiency, ultimately leading to the growth of error. We reaffirm that higher frequency variations in meridional transports are primarily wind driven, while surface buoyancy forcing is the ultimately dominant source of uncertainty at lower frequencies. For year-averaged quantities in the subtropics, it is mesoscale eddies which contribute the most to oceanic uncertainty, accounting for up to 60% after 60 years of growth for volume transport at 25°N. The impact of eddies is greatly reduced in the subpolar region, which we suggest may be explained by overall lower sensitivity to small-scale noise there.