Global Navigation Satellite System (GNSS) Meteorology has emerged as a valuable tool for atmospheric monitoring, providing high-resolution, near-real-time data that can significantly enhance nowcasting applications. By analyzing GNSS signal delays caused by atmospheric water vapor, it is possible to retrieve accurate estimates of Precipitable Water Vapor (PWV), a crucial parameter in short-term weather forecasting. This study presents a novel two-step machine learning framework for precipitation nowcasting, integrating GNSS-derived PWV with meteorological observations.

In the first step, a Random Forest (RF) model estimates precipitation based on GNSS-derived PWV, surface weather parameters, and auxiliary atmospheric variables. In the second step, a Long Short-Term Memory (LSTM) network predicts precipitation for the next hour, leveraging temporal dependencies within the data to improve forecasting accuracy. This hybrid approach combines the ability of RF to capture nonlinear relationships with the strength of LSTM in modeling sequential patterns.

The proposed methodology demonstrates interesting performance as compared to traditional forecasting models, particularly for extreme weather events such as intense rainfall and thunderstorms. The integration of GNSS meteorology with advanced machine learning techniques enhances short-term precipitation forecasting, offering a reliable tool for meteorological services, disaster prevention agencies, and early warning systems. This study highlights the potential of GNSS-based nowcasting for real-time decision-making in weather-related risk management.