This article addresses wireless sensor networks (WSN) whose nodes are organized in groups (i.e., clusters) and follow a duty-cycle. Each cluster is locally managed by a cluster head which employs a medium access control protocol to avoid interferences in all intra-cluster communications. Nevertheless, inter-cluster interferences can still occur. To this regard, we consider two clusters as interfering if their hop distance is at most t, with t ! 2, in the cluster connectivity graph. Under such a model, we target convergecast data collection of aggregated traffic and show that finding a minimum-latency interference-free convergecast schedule up to distance t is NP-hard for clusterbased WSNs with arbitrary topologies. Due to the hardness result, we restrict our attention to cluster-tree WSNs which can model ad hoc WSN deployments. We optimally solve the problem on trees for t ¼ 2 by minimizing both the latency and the schedule length. Then, for any t ! 2, we propose a minimum-latency interference-free algorithm that obtains a slot assignment with guaranteed approximated latency in OðntÞ time, where n is the number of clusters in the WSN. We also discuss a distributed implementation of such a scheduling algorithm that results in an exchange of OðntÞ messages. Moreover, we consider a minimum-latency data collection in complete trees of arbitrary degree as a special case. We finally validate our findings by a simulation study on synthetic tree topologies.