Exploring the added value of low-cost sensors via citizen observatories for peak rainfall monitoring in cities (Case study: Brussels)

Rainfall is the driving force of hydrological events. In order to predict Pluvial Flooding in cities modelling approaches make use of rainfall data of various sources: radar-based observations and predictions, high-precision rain gauges (like the OTT Pluvio² types used in the Brussels monitoring network Flowbru.be). The first have the advantage of being area-covering and having predictive power, the latter providing more precise absolute and ground-based rainfall measurements but potentially lacking spatial representativity. In an urban setting , high-density rainfall measurements are important as a little shift in rainfall may lead to a significantly different hydrological response (peak flow at different location in sewer network). The main objective is to explore the potential of low-cost rain sensors as complement for extreme peak rainfall monitoring in Brussels, Belgium. Within the frame of the FloodCitiSense project (www.floodcitisense.eu) rainfall data has been collected during 2 years (2019-2021) using low-cost acoustic rain sensors, installed via citizen observatories. For the data analysis we focus mainly on convective rain storms typically occurring during summer time, which are most often very localized and challenging to measure and/or predict.

The research questions were as following: (1) What is the performance of the low-cost sensors compared to the existing high-precision rain gauges of the FLOWBRU monitoring network in network? (2) Can we improve the quantitative estimation of extreme rainfall distribution using the measurements of the low-cost sensors?

A comparative analysis, focusing on rainfall events with a return period of 10 years (T10), between a local low-cost acoustic rain sensor and a high-precision FLOWBRU rain gauge, installed at the same location (Royal Meteorological Institute) revealed a relative strong correlation between both rainfall timeseries, but a significant under estimation of cumulative rainfall during the events. A regression analysis enabled to develop a dynamic multiplier, varying in function of the rainfall intensity per 5-min timestep, improving the rainfall estimated by the low-cost sensor. Therefore the multiplier has been used to re-calibrate all low-cost measurements. In order to answer the second research question a spatial interpolation (Inverse Distance Weighted) using the cumulative rainfall per T10 event from FLOWBRU stations WITH and WITHOUT the low-cost stations has been applied. As a reference radar QPE images were used (cumulative rainfall per T10 event). Although yielding variable results, the use of the low-cost sensor data shows clearly an added value for (extreme) peak rainfall monitoring in Brussels.