Increasing Rainfall Observation Density in Urban Areas through the Integration of Crowdsourced Personal Weather Stations

Accurate rainfall monitoring is fundamental for hydrological modeling and flood-risk management, yet the spatial density of official rain-gauge networks remains insufficient to capture short-range variability, especially in urban areas. The growing availability of crowdsourced Personal Weather Stations (PWSs) provides an opportunity to increase the density of rainfall observations at minimal cost. This study evaluates the reliability of citizen-owned PWSs across eleven Italian cities and investigates their integration with Official Weather Stations (OWSs) to enhance rainfall observation density in Rome and Turin, validated against radar-derived accumulations. Rainfall data from Netatmo PWSs (5-minute resolution) and regional OWS networks (10- to 30-minute resolution) were collected and aggregated to hourly totals to ensure temporal consistency between datasets. After applying a 10 % missing-data threshold, PWS–OWS pairs were compared in terms of annual rainfall totals, zero-rainfall frequencies, and short-duration maxima. The analysis revealed that while PWSs reproduce rainfall occurrence with high temporal consistency, they tend to underestimate rainfall volumes, particularly during high-intensity events. Two flash-flood events—Rome (8 June 2021) and Turin (22 June 2021)—were analysed to assess whether integrating selected PWSs (within ±10 % and ±20 % of the OWS reference) improves spatial rainfall representation. Interpolation with Inverse Distance Weighting (IDW) and Ordinary Kriging (OK) was compared against radar-derived accumulations. The integration of PWSs led to varying outcomes: in Turin, the denser network improved the spatial agreement with radar, while in Rome, where OWS coverage was already high, the added PWSs resulted in a marginal reduction in agreement. These contrasting results highlight that the effectiveness of PWS integration is case-dependent, influenced by network density, rainfall characteristics, and local station geometry. Finally, the open-source pws-pyQC algorithm was applied to the PWS data, and the same interpolation–validation workflow was repeated. The quality control effectively removed inconsistent records and enhanced internal data reliability. Overall, this work demonstrates that crowdsourced PWSs substantially increase rainfall observation density and, when properly filtered, can complement official networks to support high-resolution urban rainfall monitoring.