Identifying similar rainfall events across different periods to investigate temporal changes in hydrological performance of blue-green infrastructure

• BGI hydrological performance compared at two points in time. • Time series data mining to identify similar rainfall events. • K-means clustering to group months with similar rainfall patterns. • Dynamic time warping to quantify the similarity of rainfall events. Understanding how the performance of blue-green infrastructure (BGI), nature-based stormwater management systems such as bioretention cells and green roofs, evolves is essential to sustaining their hydrological function. However, traditional monitoring approaches typically compare BGI performance using rainfall events classified as similar based solely on total depth, which can obscure subtle differences in hydrological performance. This study presents a workflow that improves temporal assessment by (i) identifying statistically similar rainfall events across periods using rainfall characteristics such as intensity, duration, and peaking time, and (ii) analysing hydrological response using both volume- and time-based metrics. We evaluated the workflow under controlled but synthetic conditions using simulations of bioretention systems with underdrains. Two Swiss case studies (Kloten and Bern) were used to represent rainfall-only and mixed inflow BGI configurations, respectively. Results show that isolating temporally similar rainfall events reduces variability compared with volume-based grouping, enabling more reliable detection of performance change. Among the performance metrics tested, the volume-based indicator proved more robust, detecting temporal differences in 16–38 matched rainfall–response pairs for the two configurations. In the mixed inflow system, additional runoff increased variability, but the workflow still detected a consistent performance change over time. In contrast, time-based metrics such as P-UF lag and centroid shift required substantially larger samples to achieve comparable statistical power. These findings demonstrate the workflow’s value as a proof-of-concept for designing efficient BGI monitoring programmes. Applications to other BGI types, climatic regimes, and extended datasets remain future work.