Water resources model

A water resources model simulates the sources of water supplies and moving of water around a water network over time, typically using estimates of water demand and water availability in rivers and groundwater. The demands are most commonly given at water treatment works, but may also include abstractions taken directly from the environment (e.g. farmers). In cases where multiple water treatment works supply the same place (e.g. a city), then the point of demand will be given as the city. Currently the demands in our model are both provided by water companies and calculated from the Environment Agency’s abstraction licence database.

The flows in water resources models are most commonly those that flow into reservoirs (called “inflows”). In cases where rivers are abstracted from/released into multiple times, more flows will have to be included. To create flows, we start with rainfall generated from the Weather@Home climate modelling work and then the University of Bristol uses their national ‘dynamic topmodel’ to transform rainfall into runoff – and thus provide us with the flow inputs we require.

The first key part of our work is to create a conceptual model of the UK water network, i.e. understanding where different assets are, how they are connected, what regulations govern operation, when should assets be aggregated, etc. This conceptualisation is key to collaboration with other stakeholders – it both enables the model to be reproduced in any given water resource system simulation software and enables those with knowledge about the system to easily check that (for example) assets are currently located and connected. We give an example of this conceptualisation below.

As part of the MaRIUS project and in a collaborative project with Atkins for WaterUK a national water resource model was developed covering Thames Water, United Utilities, Yorkshire Severn Trent and Anglian Water. This model is based upon the water companies’ own models and typically includes all major reservoirs, treatment works, demand centres and important distribution junctions. In places it is aggregated (for example, the 100 odd reservoirs in the Pennines are aggregated into 6 nodes) to aid computational speed. The model is build using the WATHNET simulation modelling system, which was developed in the University of Newcastle, Australia, and is wide used for water supply planning in Australia. Wathnet is a model based on Network Flow Programming (NFP). Wathnet was selected for following reasons:

  1. The efficient computation time and capability of running on parallel nodes
  2. The scripting feature which facilitate introducing any rules or constraints; and
  3. Tts architecture facilitates the implementation of multi-objective optimisation and handling optionality

We see the conceptualisation of the water resources network surrounding Birmingham. It is important at this stage to note that the points of demand (yellow nodes) are aggregated and actually represent a distribution network that covers a wide region, rather than the points depicted. Since, in water resources modelling as a whole, it is typically unfeasible to couple distribution network models with supply network models, the demands are commonly aggregated to points.

By simulating England and Wales’ water resource system we can answer a range of important questions. Which cities are likely to face shortfalls in supply? How far do the impacts of specific assets spread throughout the system (e.g. can expanding a reservoir in Wales increase supply reliability in London?)? How resilient is our water supply to changes in climate and demand? What are the benefits of national scale infrastructure projects that cross the boundaries of existing water companies and regions?