Evaluating the Potential of Personal Weather Stations (PWS) for Semi-distributed Hydrological Modelling

 

Personal Weather Stations (PWS) have gained attention in recent years as a potential complement to operational meteorological networks, which are often sparse and may not adequately capture localized rain events, especially in areas with complex orography. PWS, on the other hand, can improve the spatial resolution of rainfall data due to their affordability and, thus, widespread distribution. However, their effectiveness and reliability depend on overcoming certain challenges. PWS often lack adherence to World Meteorological Organization standards, as they may not be properly placed nor regularly maintained, and there are no standardised approaches for data quality check. Frequent gaps in the series (mainly due to data transmission issues), and a constantly changing network layout further limit reliability and consistency of PWS data for hydrological modelling. Therefore, the application of PWS rain data for hydrological modelling is still in its infancy.  

This research focuses on evaluating PWS rainfall data for hydrological modelling in the peri-urban Lambro catchment in northern Italy, by comparing characteristics of hourly rainfall data obtained from the MeteoNetwork (Giazzi et al., 2022; https://doi.org/10.3390/atmos13060928, 2022) to those of the rain gauge data obtained from the  Regional Agency for the Protection of the Environment of Lombardy (ARPA). This study focuses on the characteristics of the subcatchment-averaged rainfall series are compared. The rain depths in each of the 15 subcatchments are calculated by using the inverse-distance weighting method with the power of 2, and with increasing maximum distance between the station and the centroid of a subcatchment (10km, 25km and 50km). The two subcatchment-averaged rainfall series are compared in terms of (1) accumulated rain depth, (2) maximum rainfall intensity, and (3) timing of the peak rainfall intensity during a rain event. 

Our results indicate that, compared to ARPA rainfall data, PWS data can both underestimate and overestimate rainfall values with similar frequency. Specifically, the magnitude of error in rain depths ranges from -44% to +56% across the subcatchments, and this range does not change significantly with increasing maximum distance. With the maximum distance of 10 km, in eight out of 15 subcatchments the absolute value of the error is smaller than 15%, while the median value amounts to 1.9%, and decreases to -17% and -19% with increasing maximum distance. The errors in maximum rainfall intensity are slightly larger, ranging from -67% to 76%, when compared to the official ARPA gauges with the maximum distance of 10 km. The median error amounts to 15.5%, -26% and -30% for the three maximum distance values. Concerning the timing of peak intensity, there are no discrepancies between the two datasets, and PWS data can be considered accurate in this regard. However, large errors in rain depths and intensities suggest that PWS rain data alone cannot be expected to yield accurate outputs in hydrological simulations. This conclusion will be tested by running a hydrological model with these datasets.  

 

Acknowledgement  

This research is part of the work within the COST Action “Opportunistic Precipitation Sensing Network” (OpenSense, CA20136)