Discussion papers

CPM-02-92 - 2 September 2001

Policy Modelling with ABSS: The Case of Water Demand Management

Olivier Barthelemy, Scott Moss, Thomas Downing and Juliette Rouchier


A promising area of application for agent based social simulation is social policy analysis. This promise stems from the ability of ABSS models to describe key processes and relationships in actual societies. However, existing descriptive studies with ABSS have been too abstract convincingly to inform policy analysis. This paper describes the transition from one such study towards a usefully detailed and realistic model currently being developed to inform water demand management in the south of England during a period of climate change.

The FIRMA project, in which this work is being undertaken, was proposed specifically to use agent based modelling techniques to enable stakeholders to become involved in the model specification and validation process. The value of the agents approach in this context is that the stakeholders can assess whether agent behaviour, interaction among the agents and resulting system level properties are descriptively accurate.

Capturing water demand management issues with agent-based models is itself a highly non-trivial task. In the first place, stakeholders including both regulatory agencies and water supply companies have identified weather, drought or its absence, demographic factors and technology as essential elements in any description of the determinants of water demand. The models to be used in policy analysis will have to capture all of these phenomena. In practice, this requires an single model relating consumption to rainfall, temperature and humidity (weather), hydrology relating the weather to the level of groundwater (drought) and the effects of those on both water demand and public authorities’ measures for managing that demand. So in a single model, we have to capture the weather system, the hydrological system, the behaviour of households and the behaviour of policy agencies and, moreover, to do so in a way that engages stakeholders in the model development and validation.