Wet delay in Global Navigation Satellite System (GNSS), mainly caused by water vapor in the atmosphere, is difficult to be accurately modeled using empirical wet delay models as water vapor is highly variable in both space and time. In this paper we propose correcting the GNSS wet delay using water vapor data from Weather Research and Forecasting (WRF) model’s assimilation results. We conduct six consecutive 24-h WRF forecasts to model the three-dimension (3D) distribution of water vapor in the South China region 20° N–33° N and 108° E–123° E from 0 h UTC April 06, 2020 to 0 h UTC April 11, 2020. GNSS Precipitable Water Vapor (PWV) from 27 stations of the Crustal Movement Observation Network of China (CMONOC) and meteorological profiles from 22 radiosonde stations are assimilated into WRF model to improve the water vapor modeling performance of WRF. Totally, four WRF schemes are adopted, i.e. WRF scheme 0: WRF without water vapor data assimilation; WRF scheme 1: WRF with GNSS PWV assimilation only; WRF scheme 2: WRF with radiosonde profiles assimilation only; WRF scheme 3: WRF with both GNSS PWV and radiosonde profiles assimilation. The water vapor data from the four WRF schemes are used to augment Precise Point Positioning (PPP) by correcting GNSS wet delay at seven International GNSS Service (IGS) stations distributed in South China. The static PPP results show that using the water vapor correction data from different WRF schemes can improve PPP positioning accuracy by 29.5% to 42.3% in the vertical component of GNSS stations. In addition, the WRF-augmented PPP can shorten convergence time by 43.3% to 57.3% in the GNSS station vertical component, if using 10 cm positioning error as the convergence criterion. The kinematic PPP results show that WRF-augmented PPP can improve positioning accuracy in the vertical component by 20.0% to 33.6%.