Assimilation of Global Navigation Satellite System (GNSS) zenith total delays (ZTDs) into the numerical weather model is one of the focuses of GNSS meteorology and has demonstrated a positive impact on short-range precipitation forecasts. In this study, observations from 17 stations of the Hong Kong Continuously Operating Reference Stations (CORS) are applied to derive ZTDs based on the precise point positioning (PPP) method. A week period in August 2018 is selected, characterized by an advective weather regime and intense precipitation events. The 3-dimensional variational data assimilation (3DVAR) technique is used to produce the initial conditions for the forecast. We conduct two experiments to investigate the impact of assimilating GNSS ZTDs in the weather research and forecasting (WRF) model. In the baseline experiment, the GNSS ZTDs are withdrawn from the observation data, while in the second experiment, the GNSS ZTDs are included. In addition, the sensitivity of results to aspects of ZTDs with different temporary resolution and ZTDs derived from different navigation systems are analyzed. Results show benefits to forecast quality when using GNSS ZTDs as an additional observation type and a more positive effect is obtained by assimilating ZTDs derived from GPS. It is worth mentioning that assimilation of ZTDs with appropriate temporary resolution will enhance the forecast quality.