HARPS-N solar RVs are dominated by large, bright magnetic regions

State of the art radial-velocity (RV) exoplanet searches are currently limited by RV signals arising from stellar magnetic activity. We analyze solar observations acquired over a 3-year period during the decline of Carrington Cycle 24 to test models of RV variation of Sun-like stars. A purpose-built solar telescope at the High Accuracy Radial velocity Planet Searcher for the Northern hemisphere (HARPS-N) provides disk-integrated solar spectra, from which we extract RVs and $\log{R'_{\rm HK}}$. The Solar Dynamics Observatory (SDO) provides disk-resolved images of magnetic activity. The Solar Radiation and Climate Experiment (SORCE) provides near-continuous solar photometry, analogous to a Kepler light curve. We verify that the SORCE photometry and HARPS-N $\log{R'_{\rm HK}}$ correlate strongly with the SDO-derived magnetic filling factor, while the HARPS-N RV variations do not. To explain this discrepancy, we test existing models of RV variations. We estimate the contributions of the suppression of convective blueshift and the rotational imbalance due to brightness inhomogeneities to the observed HARPS-N RVs. We investigate the time variation of these contributions over several rotation periods, and how these contributions depend on the area of active regions. We find that magnetic active regions smaller than $60 \ \rm Mm^2$ do not significantly suppress convective blueshift. Our area-dependent model reduces the amplitude of activity-induced RV variations by a factor of two. The present study highlights the need to identify a proxy that correlates specifically with large, bright magnetic regions on the surfaces of exoplanet-hosting stars.