On the use of mechanistic modeling for the numerical analysis of low impact developments techniques

Abstract

The increasing frequency of flooding events in urban catchments related to an increase in impervious surfaces highlights the inadequacy of traditional urban drainage systems. Low-impact developments (LIDs) techniques have proven to be valuable alternatives for stormwater management and hydrological restoration, by reducing stormwater runoff and increasing the infiltration and evapotranspiration capacity of urban areas. However, the lack of diffusion of adequate modelling tools represents a barrier in designing and constructing such systems. Mechanistic models are reliable and accurate tools for analysis of the hydrologic behaviour of LIDs, yet only a few studies provide a comprehensive numerical analysis of the hydrological processes involved and test their model predictions against field-scale data. Moreover, their widespread use among urban hydrologists suffers from some limitations, namely: complexity, model calibration and computational cost. This suggest that more research is needed to address these issues and examine the applicability of this kind of models. Thus, the main aim of this thesis was to investigate the benefits and the limitations in the use of mechanistic modelling for LIDs analysis. In this view, the mechanistic modelling approach has been used to simulate the hydraulic/hydrologic behaviour of three different LIDs installed at the University of Calabria: an extensive green roof, a permeable pavement and a stormwater filter. Each case study was used to examine a particular modelling aspect. The morphological and hydrological complexity of the green roof required the use of a three-dimensional mechanistic model, which was validated against experimental data with satisfactory results. The measured soil hydraulic properties of the soil substrate highlighted important characteristics, accounted in the simulation. The validated model was used to carry out a hydrological analysis of the green roof and its hydrological performance during the entire simulated period as well as during single precipitation events. Conversely, a one-dimensional mechanistic model was used to simulate the hydraulic behaviour of a permeable pavement, whose parameters were calibrated against experimental data. A Global Sensitivity Analysis (GSA) followed by a Monte Carlo filtering highlighted the influence of the wear layer on the hydraulic behaviour of the pavement and identified the ranges of parameters generating behavioural solutions in the optimization framework. Reduced ranges were then used in the calibration procedure conducted with the metaheuristic Particle swarm optimization (PSO) algorithm for the estimation of hydraulic parameters. The calibrated model was then validated against an independent set of data with good results. Finally, to address the issue of computational cost, the surrogate-based modelling technique has been applied to calibrate a two-dimensional mechanistic model used to simulate the hydraulic behaviour of a stormwater filter. The kriging technique was utilized to approximate the deterministic response of the mechanistic model. The validated kriging model was first used to carry out a Global Sensitivity Analysis of the unknown soil hydraulic parameters of the filter layer. Next, the Particle Swarm Optimization algorithm was used to estimate their values. Finally, the calibrated model was validated against an independent set of measured outflows with optimal results. Results of the present thesis confirmed the reliability of mechanistic models for LIDs analysis, and gave a new contribution towards a much broader diffusion of such modelling tools.