Siriwardene, Nilmini Rukma (2003) Parameter estimation of urban drainage models. Research Master thesis, Victoria University of Technology.

Abstract

Urbanisation is one of the key factors that contributes to urban flooding, which has caused major destruction to the environment and the human race. In particular, the increase in population and building density influence the change in hydrological characteristics in urban areas. Conversion of pervious areas into impervious areas increases the stormwater runoff quantity dramatically. One way of minimising urban flooding is to convey stormwater to receiving waters through stormwater drainage systems, which has been practised in the past. This practice is currently changing and the current stormwater management deals with the holistic management of the urban water cycle, which includes stormwater drainage, improvement of stormwater quality and use of stormwater as an alternative supply source (to meet increasing urban water demand). The most practical and economical way of designing the urban stormwater drainage systems is by the application of computer based mathematical software tools. These tools can be used to identify flood prone areas by modelling the catchment. Currently, there are several software tools available to develop urban drainage models, and to design and analyse stormwater drainage systems in urban areas. The widely used tools in Australia are SWMM, MOUSE, DRAINS and XP-UDD. The accuracy of these models depends on the correct selection of model parameter values. Some of these parameters can be physically measured, whereas the other parameters are impossible or difficult to measure. Therefore, these parameter values, which are impossible or difficult to measure physically, have to be estimated through model calibration. Model calibration is done through an iterative process by comparing model predictions with observations, until the two sets match with each other within a reasonable accuracy. There are several methods available to calibrate mathematical models ranging from trial and error to optimisation methods. Traditionally, model calibration is done through trial and error. With this method, the experienced modellers estimate the model parameter values starting with educated guesses and refining these guesses by comparing observations and predictions (due to these parameter values). However, this method is subjective, time consuming and can also miss the optimum parameter set. On the other hand, computer based automatic optimisation methods have proven to be robust and efficient. In this project, one of the most popular automatic calibration optimisation method known as genetic algorithms (GAs) are used to calibrate the urban drainage models. Recently, GAs have proven to be successful and efficient in identifying the optimal solution for water resource modelling applications. These applications include rainfallrunoff modelling, river water quality modelling, pipe system optimisation and reservoir optimisation. However, in order to produce efficient and robust solutions, proper selection of GAs operators is necessary for the application, before conducting the optimisation. These GA operators include population size, number of simulations, selection method, and crossover and mutation rates. There are some general guidelines available to choose GAs operators for standard GAs optimisation applications. However, there are no specific guidelines available for selecting GAs operators for urban drainage model parameter optimisation. Therefore, the sensitivity of these operators were analysed in this study through numerical experiments by repetitive simulation considering one GAs operator at a time, by integrating GAs and urban drainage modelling software tools. This produced appropriate GAs operators for use in urban drainage model parameter optimisation. XP-UDD urban stormwater drainage software and GENESIS GAs software tools were used in this study to model the urban drainage catchment(s) and model parameter optimisation. These two software tools were linked through their input and output files to conduct the model parameter optimisation. Two typical urban catchments in Victoria (Australia) were used in selecting the appropriate GAs operators. For each catchment, two design storm events (i.e. small and large) were considered. The small storm considered runoff only from the impervious areas, while the large storm produced runoff from both impervious and pervious areas. Seven parameters were identified in the urban drainage model (which required calibration), two related to impervious area and the other five related to pervious area. Typical parameter values were assumed and used in XP-UDD models of the study catchments to produce the hydrographs corresponding to these two design storms and these hydrographs were then considered in the integrated GENESIS/XP-UDD as observed hydrographs in optimising GAs operators. Numerical experiments produced consistent and robust GAs operators for parameter optimisation of urban drainage models. Although there is no mathematical basis for optimising parameter values through repetitive simulation, this is an acceptable practice for complex systems. Model calibration was carried out only for one of the two study catchments used for GAs operator study, because of the time constraints. Furthermore, one catchment was considered sufficient, since the purpose of this part of the study was to investigate and demonstrate the use of GAs for optimising parameter values of urban drainage models. Observed rainfall/runoff data were available for this catchment only for small storms, which produced runoff only from impervious areas. Therefore, only the impervious area parameter values were estimated. The results obtained from GAs optimisation were compared with previous studies and found to be satisfactory. The soil infiltration parameters, which represent a sub-set of pervious area parameters, were determined through soil infiitrometer tests, which were conducted at several sites in the study catchment, which was used for model calibration. Soil infiltration tests were conducted, because the soil infiltration parameter values could not be estimated through model calibration, due to unavailability of observed data related to large storms. A standard double-ring infiltrometer was used to estimate these parameter values through field measurements and these measurements were taken over a period of six hours. Rainfall was measured for five days prior to the field test using a pluviometer, to determine the antecedent rainfall depths at the study sites. Soil infiltration parameter values were estimated by fitting soil infiitrometer test data to Horton's infiltration equation, since the Horton's infiltration equation is built into XP-UDD and is widely used in urban drainage modelling applications in Australia. Soil samples were also tested and analysed to determine the soil particle size distribution of each site to determine the soil type. In order to understand different soil types and to determine the soil infiltration rates in different urban catchments, these soil infiltrometer tests were conducted at another nineteen sites of seven urban drainage catchments in four city councils in Victoria. The infiltration parameter values found in this study were in general significantly different to the values given in DRAINS and XP-UDD software user manuals.

Search Google Scholar

Repository staff login