Ground pressure and temperature are two key meteorological parameters for retrieving precipitable water vapor (PWV) from Global Navigation Satellite Systems (GNSS). The problem is that, many GNSS stations are either not equipped with meteorological sensors or the time resolution of meteorological data is relatively low. To improve the spatio-temporal resolution of meteorological parameters, a new spatio-temporal Kriging model based on an improved adaptive genetic algorithm (IAGA-STK) is proposed. The ERA5 (fifth-generation reanalysis dataset of the European Centre for Medium-range Weather Forecasting) dataset derived temperature and pressure with a time resolution of 1 h and horizontal resolution of 0.125° $\times 0.125°$ during the period of storm Mawar (31 Aug–4 Sep 2017), was used as inputs to the IAGA-STK model. Test data were from 13 GNSS stations equipped with meteorological sensors in the Hong Kong region, and the root mean square errors (RMSEs) of temperature and pressure from IAGA-STK model are reduced 1.5% and 5% in comparison with the traditional spatio-temporal interpolated model, respectively. GNSS-derived zenith total delays with the IAGA-STK were converted into PWV (GNSS-PWVIE), and the RMSE of the GNSS-PWVIE was less than 1.7 mm, which can meet the RMSE threshold requirements (3 mm) of PWVs as inputs to weather nowcasting. In addition, the relationship between the storm’s path and water vapor was discussed, and the PWV products with a high time resolution can be used to study the life cycle of a storm.