The forearc topography of the Cascadia subduction zone varies systematically along-strike, with high-relief in the north and south (the Olympic and Klamath ranges) separated by relatively low relief in its central region. This systematic topographic variability reflects the long-term pattern of uplift and erosion, however, the underlying cause of uplift and the mechanisms by which current topography is supported are unclear. Here, we synthesize results from seismic imaging, geodetic (decadal) and geomorphic (millennial) uplift rates, erosion rates, topographic analysis, and characteristics of the megathrust interface, with a mechanical model to infer that buoyancy in the subslab asthenosphere influences the development and longevity of Cascadia's forearc topography. The Cascadia margin can be divided into three broad segments, with the northern and southern segments characterized by rapid geodetic and geomorphic uplift rates, rapid erosion rates, high coseismic subsidence of great megathrust ruptures, shallower slab dip angles, and increased plate locking compared to the central segment. Tomographic images show low-velocity anomalies beneath the slab in the northern and southern segments, which are inferred to be regions of partial melt, resulting from localized upwellings, and regions of positive mantle buoyancy. Modeling suggests that these buoyant regions can locally increase the total shear force at the megathrust by either shallowing the slab dip — thereby increasing the area of the seismogenic zone — or increasing mechanical plate coupling along the megathrust, by increasing normal stress and/or the effective coefficient of friction. We propose that: 1) sub-slab buoyancy influences topographic development by modulating along-strike patterns of strain within the over-riding plate during the seismic cycle, 2) Permanent development of surface topography occurs due to unrecovered strain over thousands of seismic cycles, and 3) variations in Cascadia's forearc topography are laterally supported by changes in the total shear force at the megathrust interface. Using independent estimates for slab dip and plate locking, we predict the first-order variations in Cascadia forearc topography, with local maxima in the north and south as observed. Given our results, variations in subslab buoyancy may be critical to explaining forearc topography in other subduction systems.