1: \begin{abstract}

2: Effective stormwater treatment infrastructures are crucial for mitigating the adverse effects of runoff on urban water quality. However, designing cost-effective treatment systems can be challenging due to complex turbulent flow dynamics. This study presents an in-depth analysis of turbulent flow in hydrodynamic separators (HS) using time-resolving, high-resolution particle image velocimetry (PIV) and modal decomposition techniques. The examination of interrogation window sizes on PIV measurements highlights a trade-off between spatial resolution and measurement uncertainty. Additionally, the impact of sampling frequencies and durations on the convergence of turbulence statistics, such as mean flow and Reynolds stress, is quantified. Results indicate that higher-order statistics require significantly larger sampling sizes (5x) to achieve the same level of statistical convergence as mean flow statistics. Leveraging proper orthogonal decomposition (POD) and spectral proper orthogonal decomposition (SPOD), dominant turbulence structures and coherent flow features are identified, providing a foundation for the development of reduced-order models (ROM). These ROMs demonstrate improved computational efficiency and have potential for real-time control applications and integration into larger drainage-scale simulations. Furthermore, the outcome of this study establishes a high-quality, open-source turbulent flow database for HS systems, offering the civil and environmental engineering community valuable resources to guide the development and application of computational fluid dynamics (CFD) tools. This study represents a critical step toward the ultimate goal of facilitating scientifically guided HS design and optimization.

3: 

4: \noindent\textbf{Keyword}: water treatment; urban water infrastructure; POD; SPOD; modal analysis; CFD

5: \end{abstract}

6: