Experimental study and qualitative and quantitative modelling of sustainable urban drainage systems (SUDS)

Abstract

Climate changes have become always more frequent, increasing the interest of researchers in finding the causes and, above all, the structural or non-structural solutions to solve the problem. Economic development together with rapid population growth constantly increase the demand of goods and services. As the same as drought, also precipitation became more intense and frequent, even with more ever short duration. These events for their heavy impact are called ‘extreme rainfall events’. The actual management of urban waters is unsustainable thus, foregoing reasons lead to an imperative need to develop new urban ecosystems, requiring a rethink of traditional development techniques. Traditional urban drainage systems are designed to rapidly collect and convey overland flows to the treatment plants, without taking into account of their qualitative characteristics. In order to reach the aim of the qualitative and quantitative control of stormwater in urban areas, a possible way is the widespread implementation in urban areas of ‘blue-green infrastructure’ that provide an holistic and integrated approach to the problem. They are one step beyond other ‘classic’ sustainable urban drainage measures such as LID (Low Impact Development), SUDS (Sustainable Urban Drainage Systems) or BMPs (Best Management Practices), allowing to emphasize their beneficial effects. Use of BGC as a part of sustainable drainage system concept is a winning approach, that allow managing and treatment of stormwater runoff within urban areas, using practices made of green and blue components. Generally green components are represented by any kind of existing vegetation (floral plants, grass, hedges) while the blue one by lakes, ponds, rivers and canals (natural or artificial). Together, these infrastructures allow to create a network between them at regional scale. The real behaviour of these structures is not yet properly modelled. Most of the software currently used in urban hydrology (SWMM by EPA, Music by eWater CRC, etc…) model in a reasonable way the hydraulic behaviour of infiltration practices (such as bioretention cells, infiltration trenches, vegetated filter strips, porous pavement) using a simple mass balance approach. Generation, inflow and transport of pollutants are, instead, determined by the land use assigned to each subcatchments, namely through buildup and washoff laws describing accumulation and washout by either a mass per unit of subcatchment area or per unit of curb length. This approach completely lack of quality algorithms within LID models that take into account of their quality performances as, for instance, reduction of efficiency due to the clogging effect. The clogging phenomenon, described as the decrease in infiltration rate of the soil due to the reduction in soil porosity and hydraulic conductivity, occurs for the majority within infiltration practices such as bioretention cells, infiltration trenches, vegetated swales and permeable pavers. Precisely these latter practices are one of the easiest to implement into urban environment, being aimed to reduce impervious areas and work as ‘link’ within BGCs networks. From these premises the research in the following thesis is developed, whose main objective is to study the implementation of 'blue and green' elements in urban areas and their effect on pollutant loads reduction. Initially, a study of common errors retrieved within a DTM (Digital Terrain Model) has been faced because, if not corrected, they will affect the overland flow network generation and the subsequent hydraulic modelling. DEMs (Digital Elevation Models) can include both terrain elevation data, which commands flow direction of floodwater, and land cover information, which dictates resistance to floodwater distribution. Very often DTMs originate from a variety of ground observations supplemented by various remote sensing techniques (aerial and satellite measurements, total stations, dGPS, aerial LiDAR, terrestrial laser scanning) thus, containing systematic or random errors to individuate and eliminate. A study were carried out to evaluate how DTM resolutions and presence of building affect overland flow network delineation in the Liguori Channel basin, situated in Cosenza (Italy). To achieve this aim, three different DEMs of the study area, generated from different sources, were used: two contour-based DTMs with contour interval respectively of 30 m (DTM 30) and 20 m (DTM 20), and one LiDAR-based DEM, with horizontal resolution of 1 m (LIDAR DTM). Moreover, for a more in depth analysis, LIDAR DTMb (with buildings) cell size has been down sampled from 1 to 5 meters coarse resolution, in order to evaluate also, how cell size affect ponds delineation. Individuation of likely flood areas (ponds) has been carried out using Arc Hydro Tools developed at Centre for Research in Water Resources at University of Texas at Austin. Research highlighted how the correction of DEM generated from LiDAR data and other sources overlapping the buildings (i.e. retrieved from cadas maps) help to diminish the total accumulated water volume into surface ponds, real or spurious, and also that their number does not depend by the raster cell size, but from the accuracy of the source data. Afterwards, a first attempt of best management practices implementation has been carried out within the Liguori Channel situated in Cosenza, Italy. The overland flow network of a highly urbanized sub area has been enhanced through the addition of a certain percentage of green roof and porous pavements. A series of simulations were carried out, using in input the historical annual rainfall series (between 2008 and 2011) and considering a first scenario without LIDs (reference case) and a second scenario with the new practices implemented. Moreover, the same simulation were repeated in continuous, namely considering a single time series composed by 4 years of precipitations (2008-2011) and taking into account, in addition to the two previous cases, of a third scenario where LIDs may deal with clogging phenomenon. In order to perform the EPA SWMM modelling, a ‘residential’ land use has been defined, characterised by build-up and wash off laws for the considered pollutant (Total Suspended Solids – TSS). As regards the green roof and porous pavement simulation parameters, currently these values has been gathered from literature. Within SWMM, the clogging phenomenon is taken into account through a parameter called ‘clogging factor’ that considers the possible decay of LID performance due to the fine material carried by infiltration waters. The empirical formulation is affected by some parameters such as the number of years it takes to fully clog the system (Yclog), the annual rainfall amount over the site (Pa), the pavement's capture ratio CR (area that contributes runoff to the pavement divided by area of the pavement itself), the system's void ratio (VR), the Impervious Surface Fraction (ISF) and the pavement layer thickness (T). The yearly simulation performed show how the percentage reduction of volumes into the network is around 35% on average each year, the mass of Total Suspended Solids is around 30% on average while the relative concentration undergoes an increment around 15%. The latter result can be explained looking at the SWMM runoff quality algorithm. In fact, currently SWMM takes into account of the reduction of pollutants only in terms of reduction of overland flow, due to the lacking of quality algorithms for LIDs simulation. Consequently, the presence of BMPs increases the amount of stormwater that infiltrates, decreasing runoff, therefore the mass of pollutants reaching the sewer outlet. The lower is the volumes of water reaching the sewer, keeping constant the total mass of pollutant over the catchment, the higher is the average outlet concentrations. The results of the continuous simulation are, also, very interesting. While during the annual simulations the trend of volumes for the scenario ‘LIDs with clogging’ ranges always between the other two cases, without and with LIDs, when the continuous simulation is considered, the volumes of the clogged LID are even higher than the volumes occurring without any BMP implemented. The efficiency tends to decrease during time, from 50% when simulation starts to almost 0% at the end of the second year, continuing then to swing around zero per cent for the remaining part of the simulation. In this case, in fact, during the first two simulated years the trend is similar to what it has been found during the annual simulation, while starting from the third year (January 2010), volumes generated for the case ‘LIDs with clogging’ are equal or even higher than those ones generated when no LIDs are used. Although EPA SWMM results are interesting and indicative of LID operation, they are not very accurate, especially concerning the qualitative simulation of the stormwater management practices. For this reason, later, the research has been focused on improving the qualitative simulation algorithms, with particular attention to porous pavements. Data collected into an experimental laboratory rig of three different and widely used permeable pavement types has been analysed. The investigated systems were: monolithic porous asphalt (PA), modular Hydrapave (HP), and monolithic Permapave (PP). The rig, made of three vertical compartments in which the three porous pavers stratigraphies has been rebuilt, has been subjected to a semi-synthetic hyetograph, made of five different rain intensities (wetting regime) plus several drying periods. From the frequency curve typical of Brisbane (AU), in correspondence of different percentile ranges four flow rates has been chosen (A, B, C, D). In addition, a 1 in 5 year storm of 5 min duration was selected; this represents the typical design storm where the porous pavers are likely to be developed. The accelerated laboratory test allowed to simulate 26 years of operation under Melbourne climate. About the water quality monitoring, an intense sampling regime has been conducted in which samples were collected from inflow and outflow and analysed for Total Suspended Solids (TSS), Total Phosphorus (TP) and Total Nitrogen (TN). Afterwards, a correlation analysis has been performed in order to individuate the key variables affecting the porous pavement functioning. According to these results, the key variables identified to affect the pollutant concentration values were: the cumulative flow every 6, 12 and 24 hours before the sampling time, the cumulative inflow volume in each time step and the cumulative trapped mass. Initially, it has been tried to analyse the phenomenon through the ‘k-C* model’, that is a conceptual model used to simulate the pollutant behaviour through the system, based on a first-order kinetic decay equation. Notwithstanding the wide popularity and tested applicability on various other treatment practices such as sand filters, wetlands, ponds, infiltration systems and vegetated swales, the model did not show satisfying results when applied to porous pavements, especially about heavy metal and total nitrogen modelling. The predictive power of the model has been assessed through the calculation of the Nash–Sutcliffe model efficiency coefficient, widely adopted in the Anglo-Saxon world to evaluate behaviour and performance of the hydrologic models. Nash-Sutcliffe coefficient is an indicator of the model’s ability to predict about the 1:1 line between observed and simulated data. NSE ranges between −∞ and 1.0 (1 inclusive), with NSE = 1 being the optimal value. Values between 0.0 and 1.0 are generally viewed as acceptable levels of performance, whereas values < 0.0 indicates that the mean observed value is a better predictor than the simulated value, which indicates unacceptable performance. Considering this, the concentration data collected has been processed, also taking into account of the correlation analysis previously carried out, which allowed to estimate the concentrations of the main pollutants such as TSS (Total Suspended Solids), TP (Total Phosphorous) and TN (Total Nitrogen) to the output section of the porous pavements. The reliability of the new proposed formulas has been demonstrated both by high values of the Nash- Sutcliffe coefficients, always positive, and also by very low errors (between 10% and 25%) among modelled and measured concentrations