Short-term domestic water demand: estimation, forecasting and management

In the UK, the water resource problems during the droughts of 1988-1992 and the well- publicized problems of 1995-1996, serve to highlight the finite nature of the potable water resource. Demand management is increasingly considered a fundamental tool in promoting a sustainable water resource strategy. However, of equal importance is the development of accurate water demand forecasts that work in parallel with demand management measures. These forecasts should predict all components ot water use on different planning horizons. Presently, water plcs have a very limited understanding of factors influencing short-term domestic water demands and rely upon crude methods of forecasting. However, no precise definition of short-term exists within the water industry. Previous research defined ‘short- term to be between twenty-four hours and seven days ahead. It was argued that effective weather prediction and operationally useful forecast times were the determining factor in the definition.
The domestic water consumption data used in this research is derived from two different methods, (i) zonal metering and (ii) individual household metering. Welsh Water and Yorkshire Water provided zonal metering data, which refers to the flow to a zone that has many households. Essex and Suffolk Water, Thames Water and Yorkshire Water provided individual household metering data, which is a measure of consumption in single households. These data were used in the determination of: (i) underlying factors that influence the demand for water in both the short- and the medium- to long-term and (ii) factors that influence short-term demands. The influential factors aided in the exploration of modelling strategies to forecast short-term domestic water demands. Approaches explored included a pragmatic approach, based on a form of accounting using a series of 'lookup tables’, and advanced approaches, including stepwise regression, both with and without k-means cluster analysis, and univariate and multivariate ARMA time-series modelling. The most successful approach was then used to determine how future scenarios such as changes in the population base, climate, culture and technology might influence the characteristics of short-term domestic water demands.

Household size and property type appears to exert the greatest underlying influence on medium- to long-term domestic water demands. In the short-term, domestic water demand appears to be influenced by the two days antecedent and the prevailing day’s weather conditions, day of the week, calendar effects, school holidays and demand management measures. No single approach provided the best overall prediction of short-term domestic water demands. However, the pragmatic approach emerged as one of the most promising techniques. The pragmatic approach, used to determine how future scenarios might change the characteristics of short-term domestic water demand, suggests that increases in demands are associated with changes in the population base, climate and culture. However, changes in technology associated with the widespread implementation of demand management measures have the potential to suppress the increases and indeed reduce demands to less than those of the present day.