This research applies a hierarchical k-means clustering method on the TF-IDF weighted 2019 cycling transactions from the Citi Bike bike-sharing system operating in New York City, with the primary goal of investigating the spatiotemporal usage pattern of its docking points. With a particular focus on bike-sharing stations in Manhattan, we classify 504 stations into four main clusters featuring heterogeneous dynamic usages, including leisure-oriented, residential- oriented, workplace-oriented, and off-peak oriented. We interpret each cluster based on their salient characteristics and anticipate possible future directions of this work.