This study introduces a geographical multi-layered clustering protocol for ad hoc wireless sensor networks (WSNs), where the size of clusters is variable so that the closer clusters to the base station (BS) have a smaller size than farther ones. Moreover, in each cluster, using some intelligent fuzzy rules and in a decentralised way, a novel subtree strategy is determined. In this way, some parent nodes are chosen that are responsible for collecting and aggregating data from their adjacent ordinary nodes and sending them to its cluster head, directly or via other parent nodes, which substantially decreases intra-cluster communication energy costs. Furthermore, these two compatible techniques can fairly mitigate the hot spot problem resulting from multi-hop communication with the BS. The simulation results demonstrate that the proposed protocol outperforms two energy-efficient protocols named DSBCA and LEACH in terms of functional network longevity for both small-scale and large-scale sensor networks.