Session D7 - Hydrologic monitoring of SCMs

This study focuses on modeling and optimizing the Artificial Wetland/Reservoir Tank System (SHATR) at Javeriana University using machine learning to enhance resilience against climate extremes. Historical precipitation and runoff data (2014–2016) were processed into 98 rainfall events for predictive modeling. Among tested algorithms, Support Vector Regression (SVR) optimized via Bayesian search achieved the best performance (R² = 0.9024, RMSE = 1.59 cm). Key predictors included accumulated rainfall, peak intensity, and initial tank level. Climate change scenarios (2021–2100) projected a 6.1% rise in overflow events for 2021–2040 versus 3.1% historically. While outflow modeling showed limitations, inflow prediction proved feasible for early warning and adaptive management. Machine learning (ML) demonstrated efficiency and robustness for non-linear hydrological relationships, outperforming traditional hydrodynamic models. Findings highlight the need for proactive strategies: increasing storage capacity, optimizing spillway design, and integrating real-time alerts. Future work should incorporate longer monitoring, additional variables, and hybrid models. This research validates ML as a powerful tool for resilient urban water management under climate uncertainty.

As part of the European GreenStorm project, a conceptual hydrological model was developed to explore the performance of a nature-based solution (NbS) under a range of design and implementation contexts. This solution, Vertuo©, is an innovative NbS that combines a lined storage module and a natural soil area. The objective is to assess the hydrological behaviour of this NbS and to guide its design across different climatic contexts (temperate in Paris and Mediterranean in Athens), under both actual and future (RCP8.5) conditions. A workshop brought together around twenty experts, including researchers and municipal engineers from Paris, Copenhagen, and Athens, to test the tool and explore potential optimization strategies. The scenarios analysed revealed several key issues: the difficulties of increasing evapotranspiration due to limits imposed by potential evapotranspiration, exfiltrations that remains challenging to control, and a marked alternation between periods of dryness and saturation. Parameter modifications (soil type, Vertuo©/natural distribution, drain height) provided only limited improvements, particularly in terms of the overall water balance, and no clear compromise emerged. This work represents a first step toward an integrated framework for the design and management of multifunctional, climate-resilient NbS.

This study proposes a model to simulate the hydraulic/hydrological behaviour of modular tray-based blue roof systems during rainfall events. The model was developed using EPA-SWMM software and it was applied to a full-scale pilot of modular blue roof installed in south Italy. The paper firstly provides a description of the pilot installation. Then, the conceptualization of the modular blue roof system in the software is outlined. Available components and tools in the software were arranged and customized to reproduce the system behaviour. Model parameters were set based on the geometrical and hydraulic characteristics of the pilot. The model was validated by simulating the system behaviour during 9 rainfall events recorded on the pilot between 2018 and 2024. Results of the model application to the pilot show a good ability of the model to reproduce the behaviour of the system during precipitation events. The model opens perspective to the analysis of the broader benefits of implementing such systems at the scale of the urban catchment, providing valuable opportunities for sustainable urban water management.