The study aimed to characterize the molecular epidemiology, phylogenetic relationship, and population dynamics of the G protein gene in clinical respiratory syncytial virus (RSV) strains isolated from northern Taiwan. We analyzed a total of 160 and 116 G protein gene sequences of RSV-A and RSV-B representative strains, respectively, from 804 clinical viral stocks collected between July 2000 and June 2016. Population dynamic patterns of the RSV G protein gene were analyzed using Bayesian inference through the Markov chain Monte Carlo framework. A phylogenetic analysis revealed that RSV-A from Taiwan could be categorized into GA2, GA5, and GA7 lineages. GA2 of RSV-A could be further divided into NA1, NA2, NA4, and ON1 clades. These RSV-A lineages has been replaced over time, whereas RSV-B strains from Taiwan continually evolved from a single lineage with significant time-dependent waves. Four putative positive selection sites were observed in both RSV-A and RSV-B. The Bayesian skyline plot revealed that the local population dynamics of RSV were associated with lineage displacement events. Both circulating subtypes and population dynamics represented a unique local pattern. Our results affirm the necessity of continuing molecular surveillance of RSV to attain a more comprehensive understanding of epidemics.