One major challenge of relying on Dedicated Short Range Communication (DSRC)-equipped vehicles for high precision cooperative localization is related to the high computational complexity and heavy communication loads of exhaustively considering links to all neighbors regardless of their quality. This paper addresses the problem of selecting the best subset of links to spatial neighbors, considering varying degradation first from GPS conditions and second from GPS positioning capabilities. We formulate a computationally efficient particle filter-based link selection algorithms based on Cramér-Rao Lower Bound (CRLB) indicators accounting for neighbors uncertainties. We show that selective fusion significantly reduces the computational complexity and required network traffic with a modest increase in the position error in most cases and an acceptable degradation in the worst-case long-term GPS-denied condition or under severe neighbor positions uncertainties.