The impact of drinking water network model spatial and temporal scale on hydraulic metrics indicating discolouration risk

Abstract

[EN] Matching model complexity to application and ensuring sufficient complexity to capture the emergent behaviour of interest is a perennial challenge. In this paper we define a model as the variables, parameters and factors that represent a particular place, time and situation, not the software or algorithms. Specifically, we explore the cross products of spatial and temporal scaling of water demands within extended period 1D network model simulations to predict the hydraulic conditions within individual drinking water pipes and the association of this with discolouration risk. High spatial scale hydraulic models investigated include mapping each customer with a unique demand node instead of the current practice of aggregated demand to nodes at the ends of pipe lengths. For demand profiling we compare top-down DMA inlet patterns at 15-minute resolution with bottom-up stochastic demand patterns down to 1 second timesteps. The value of the resulting increases in resolution of hydraulic model outputs are captured in a range of pipe specific metrics that are likely to be indicative of discolouration risk. Unlike water quality surrogate of water age or chlorine residual that aggregate time and pipe effects from source to point of interest, discolouration risk has been shown to be primarily a function of the hydraulic conditions in a specific length of pipe. Hence the additional effort to achieve high-resolution modelling simulations are perhaps warranted to manage discolouration. Results review proposed discolouration metrics by correlating with consumer reported discolouration events, showing how these change as a function of spatial and temporal resolution. For example, increasing temporal scale from 15 minutes to 1-minute results in a 15-fold increase in identifying flow reversal locations that can facilitate settling of network discolouration material and therefore pose a discolouration risk. High temporal scale is shown to capture the on/off nature of customer demands and the significant impact on peak velocities that are shown suppressed when using aggregated profiles. This work provides an indication of the optimal level of model resolution required to differentiate pipes according to discolouration risk and hence improve targeting of pro-active maintenance and discolouration management efficiency.