Personal weather station rainfall data for semi-distributed flood modelling: Feasibility and limitations

Accurate representation of the spatial and temporal variability of precipitation is a fundamental requirement for reliable flood modelling, especially if semi-distributed/fully-distributed models are used. However, official rain gauge networks often exhibit limited spatial coverage and low density, leading to substantial uncertainty in the representation of rainfall at the sub-basin scale. Recently, opportunistic precipitation observations derived from personal weather stations (PWS) have attracted increasing attention as a potential complementary data source, offering unprecedented spatial coverage. At the same time, PWS networks are characterized by heterogeneous data quality, inconsistent maintenance, frequent outages, incomplete records, and a dynamically changing network structure. Despite the attention that PWS have gained, their applicability in hydrological modelling, especially within semi-distributed modelling frameworks, has been explored in only a limited number of studies.

This study evaluates the feasibility of PWS rainfall data for semi-distributed hydrological flood modelling and outlines the conditions under which their application is appropriate. The Lambro catchment in northern Italy is used as a case study. PWS rainfall observations obtained from the Meteonetwork platform (https://www.meteonetwork.it/, Giazzi et al., 2022) and official rainfall data provided by the Lombardy Regional Environmental Protection Agency (ARPA) are used in this study. Different PWS-based rainfall datasets are created: namely, raw PWS data (PWSraw), quality-controlled PWS data (PWSqc), and data from persistent PWS stations, implying those PWS that were active over all considered storm events (PWSqc_p), and their combinations with the ARPA observations (denoted by ARPA + PWSraw, ARPA + PWSqc, ARPA + PWSqc_p, respectively).

Each set is compared to the ARPA rain gauge measurements, which are used a reference dataset. The evaluation is performed by comparing rainfall features at the point- and sub-basin scales, as well as through semi-distributed hydrological flood simulations by analyzing the impact of the rainfall input on simulated peak discharge and timing of its occurrence, and runoff volume at the basin outlet. The hydrological modelling with every rainfall dataset is performed by using the semi-distributed model developed by Politecnico di Milano (Cazzaniga et al., 2022).

The results demonstrate that quality-controlled and persistent PWS datasets (PWSqc and PWSqc_p), as well as their combination with ARPA observations, generally enhance hydrological model performance. This indicates that PWS data can provide added value for semi-distributed flood modelling when appropriately controlled and integrated with reference datasets from the official networks.

 

References

Cazzaniga, G., De Michele, C., D’Amico, M., Deidda, C., Antonio Ghezzi, A., and Nebuloni, R.: Hydrological response of a peri-urban catchment exploiting conventional and unconventional rainfall observations:  the case study of Lambro Catchment, Hydrology and Earth Sysem. Sciences, 26, 2093–2111, https://doi.org/10.5194/hess-26-2093-2022, 2022.

Giazzi, M., Peressutti, G., Cerri, L., Fumi, M., Riva, I. F., Chini, A., Ferrari, G., Cioni, G., Franch, G., Tartari, G., Galbiati, F., Condemi, V., and Ceppi, A.: Meteonetwork: An Open Crowdsourced Weather Data System, Atmosphere, 13, 928, https://doi.org/10.3390/atmos13060928, 2022.

https://www.arpalombardia.it/   

 

Acknowledgments

The authors would like to thank the COST Action “OpenSense” (CA20136) for supporting collaboration opportunities among the co-authors through the STSM program.