Comparison between the Top-down and Bottom-up approach for the diffuse-dispersive phenomenon analysis

Abstract

[EN] In order to detect deliberate or accidental contamination in drinking water distribution systems (DWDS), typically water quality sensors need to be installed in this system, and the data need to be analysed in order to feed alert systems and prevent the harm of contamination. This requires numerical (hydraulic and water quality) models that are as realistic as possible to support monitoring systems. Currently, water quality models used in the literature adopt an advective approach and simplified reaction kinetics, such as EPANET, which neglect diffusion-dispersion phenomena that are relevant in the presence of laminar and transient flow regimes. Another important aspect providing relevant uncertainty is related to the simplified estimation of sub-daily water demands that are commonly estimated from highly aggregated consumption data.The present study aims to analyse diffusion-dispersive phenomena in a realistic DWDS model, which shows turbulent, transitional and laminar flows, and compare this to how such a DWDS would typically be modelled with a coarse estimate of demands. We are therefore considering two different demand allocation approaches (Top-down and Bottom-up).In this paper the EPANET advective model and the diffusive-dispersive model, developed in a previous study, were used to better understand what the effect using the latter approach has within the DWDS as a function of two different types of demand allocation. To do this, the models results were compared to numerical tests that were performed on the real network of Zandvoort (the Netherlands) using a conservative tracer. For the 4 locations considered, it was noted that the diffusive-dispersive model responds well when using the bottom-up approach compared to the top-down approach. We found that in order to predict the tracer pattern, the Top-Down approach of demand allocation does not work well, even when an optimized diffusive-dispersive model is used. The bottom-up approach of demand allocation leads to far better results in predicting the tracer patterns, and with the diffusive-dispersive model the prediction improves even more. This means that in order to model water quality in a DWDS the first step should be to improve water demand models for this DWDS. This leads to an improved representation of flow regimes, and will most likely include laminar flows.