We show that the dipole anisotropy of CRs above 1 EeV measured by the Pierre Auger Observatory is well described as the sum of the contribution of an extragalactic population of CRs (ECRs) plus CRs from a Galactic source. Above 8 EeV the CRs are almost exclusively extragalactic and the ECR anisotropy is well constrained by data; the energy dependence of the dipole is weak and well-understood theoretically. Modeling the spectrum and composition of CRs above 1 EeV reveals two disjoint rigidity groups, attributable to ECR and GCR populations. This enables the relative contributions of GCRs and ECRs to the total flux in each energy bin to be determined. Thus the ECR dipole in the region containing both ECR and GCRs can be removed, allowing us to isolate the dipole anisotropy of the highest energy Galactic CRs.

The dipole of these highest energy GCRs is inconsistent at greater than 6-sigma with being toward or away from the Galactic center, disfavoring acceleration in the Galactic termination shock and preferring acceleration in a transient event whose longitude we constrain: L ~70 \pm 20 deg. The amplitude alpha of the high-energy GCR dipole is ~0.05, which leads to an estimate for the distance/time since the event, using alpha ~ r/(2 c t) ~ 0.05. A candidate remnant of the transient is identified: SNR G65.3+5.7. This SN occurred about 800 pc away, about 20 kyr ago. This single transient event can be responsible for the entire local population of CRs with energies ranging over the bins 0.25-0.5 EeV to 4-8 EeV, with only a tiny fraction of the overall energy of the transient explosion being in VHE CRs. The pulsar PSR J1931+30 may also be a relic of the event; if so, it would confirm the core-collapse nature of the SN. Most massive young stars are in binary systems. We thus propose that the highest energy GCRs were accelerated when the shock created by a core-collapse supernova collided with the wind of its massive binary companion. From the rate of core collapse SNe and the fact that massive stars are generically in binaries with other massive stars, we estimate the probability of seeing an event close enough in space and time to give the observed flux and an anisotropy as large or larger than observed to be O(0.1-1).